Using Support Vector Machines for Direct Marketing Models

Size: px
Start display at page:

Download "Using Support Vector Machines for Direct Marketing Models"

Transcription

1 Iteratioal Joural of Egieerig ad Advaced Techology (IJEAT) ISSN: , Volume-4 Issue-4, April 2015 Usig Support Vector Machies for Direct Marketig Models A. Nachev, T. Teodosiev Abstract This paper presets a case study of data miig modelig for direct marketig, based o support vector machies. We address some gaps i previous studies, amely: dealig with radomess ad 'lucky' set compositio; role of variable selectio, data saturatio, ad cotrollig the problem of uder-fittig ad over-fittig; ad selectio of kerel fuctio ad model hyper-parameters for optimal performace. I order to avoid overestimatio of the model performace, we applied a double-testig procedure, which combies cross-validatio, ad multiple rus. To illustrate the poits discussed, we built predictive models, which outperform those discussed i previous studies. Idex Terms classificatio, data miig, direct marketig, support vector machies. I. INTRODUCTION Today, baks are faced with various challeges offerig products ad service to their customers, such as icreasig competitio, cotiually risig marketig costs, decreased respose rates, ad at the same time ot havig a direct relatioship with their customers. I order to address these problems, baks aim to select those customers who are most likely to be potetial buyers of the ew product or service ad make a direct relatioship with them. I simple words, baks wat to select the customers who should be cotacted i the ext marketig campaigs. Respose modelig is usually formulated as a biary classificatio problem. The customers are divided ito two classes, respodets ad o-respodets. Various classificatio methods (classifiers) have bee used for respose modelig such as statistical ad machie learig methods. They use historical purchase data to trai ad the idetify customers who are likely to respod by purchasig a product. May data miig ad machie learig techiques have bee ivolved to build decisio support models capable of predictig the likelihood if a customer will respod to the offerig or ot. These models ca perform well or ot that well depedig o may factors, a importat of which is how traiig of the model has bee plaed ad executed. Recetly, eural etworks have bee studied i [9], [10], [11], [12], ad regarded as a efficiet modelig techique. Decisio trees have bee explored i [10] ad [11]. Support vector machies are also well performig models discussed i [9], [12], ad [13]. May other modelig techiques ad approaches, both statistical ad machie learig, have bee studied ad used i the domai. Mauscript Received o April A. Nachev, BIS, Caires Busiess School, NUI Galway, Galway, Irelad. T. Teodosiev, Departmet of Computer Sciece, Shume Uiversity, Shume, Bulgaria. This paper focuses to support vector machies as a modelig techique ad discuss factors, which affect their performace ad capabilities to predict. We exted the methodology used i [9], [10], ad [11], addressig certai gaps, which ifluece model performace. The remaider of the paper is orgaized as follows: sectio II provides a overview of the data miig techique used; sectio III discusses the dataset used i the study, its features, ad the preprocessig steps eeded to prepare the data for experimets; sectio IV presets ad discuses the experimetal results; ad sectio V gives coclusios. II. SUPPORT VECTOR MACHINES Support vector machies are commo machie learig techiques. They belog to the family of geeralized liear models, which achieve a classificatio or regressio decisio based o the value of the liear combiatio of iput features. Usig historical data alog with supervised learig algorithms, SVM geerate mathematical fuctios to map iput variables to desired outputs for classificatio or regressio predictio problems. SVM, origially itroduced by Vapik [1], provide a ew approach to the problem of patter recogitio with clear coectios to the uderlyig statistical learig theory. They differ radically from comparable approaches such as eural etworks because SVM traiig always fids a global miimum i cotrast to the eural etworks. SVM ca be formalized as follows: Traiig data is a set of poits of the form D = {(x i, c i ) x i R p,c i { 1,1}} i=1 where the c i is either 1 or -1, idicatig the class to which the poit x i belogs. Each data poit x i is a p-dimesioal real vector. Durig traiig a liear SVM costructs a p-1-dimesioal hyper-plae that separates the poits ito two classes (Fig. 1). Ay hyper-plae ca be represeted by w x b = 0, where w is a ormal vector ad deotes dot product. Amog all possible hyper-plaes that might classify the data, SVM selects oe with maximal distace (margi) to the earest data poits (support vectors). Whe the classes are ot liearly separable (there is o hyperplae that ca split the two classes), a variat of SVM, called soft-margi SVM, chooses a hyperplae that splits the poits as clealy as possible, while still maximizig the distace to the earest clealy split examples. The method itroduces slack variables, ξ i, which measure the degree of misclassificatio of the datum x i. Soft-margi SVM pealizes misclassificatio errors ad employs a parameter (the soft-margi cost costat C) to cotrol the cost of misclassificatio. Traiig a liear SVM classier solves the costraied optimizatio problem: (1) 183

2 Usig Support Vector Machies for Direct Marketig Models mi w,b,ξk 1 s.t. 2 w 2 + C w x + b 1 ξ i I dual form the optimizatio problem ca be represeted by 1 mi αi 2 i=1 j=1 α j y i y j x i x j s.t. 0 C, c i = 0 The resultig decisio fuctio f (x) = w x + b has a weight vector w = k=1 a k y k x k i=1 > 0 are called support vectors, si. Data poits x i for which defie the maximum margi hyperplae. Maximizig the margi allows oe to miimize bouds o geeralizatio error. If every dot product is replaced by a o-liear kerel fuctio, it trasforms the feature space ito a higher-dimesioal oe, thus though the classifier is a hyperplae i the high-dimesioal feature space (Fig. 2). The resultig classifier fits the maximum-margii the trasformed feature space. The kerel fuctio ca be hyperplae defied: k(x i, x j ) = Φ(x i )Φ(x ) j (4) where Φ(x) maps the vector x to some other Euclidea space. The dot product x i x j i the f replaced by k(x i, x j ) so that the S problem i its dual form ca be redefiedd as: maximize (i ) ɶL(α) = 1 2 i i j α j y i y j k(x i, x j ) s.t. y i = 0; 0; 1 i N i A o-liear SVM is largely characterized by the choice of its kerel, ad SVMs thus lik the problems they are desiged for with a large body of existig work o kerel-based methods. Some commo kerels fuctios iclude: Liear kerel: k(x, x') = x x' Polyomial kerel: k(x, x') = (scale x x'+ offs Gaussia RBF kerel: k(x, x') = exp( σ x x' 2 ) (8) Hyperbolic taget kerel: k(x, x') = tah(scale x x'+ + offset) (9) Laplacia kerel: ξ i=1 i (2) k(x, x') = exp( σ x x' ) (10) i=1 (3) ce they uiquely formulae above is SVM optimizatio (5) (6) set) degree (7) Figure 1. Maximum-margi hyperplae for a SVM traied with samples from two classes. Samples o the margi are support vectors. The choice of kerel strogly depeds o the task specifics ad is usually made after empirical survey. The kerel parameters appear hyper-parameters for the model ad their tuig is a impotrat for the classifier performace. The SVM s major advatage lies with their ability to map variables oto a extremely high feature space. Because the size of the margi does ot deped o the data dimesio, SVM are robust with respect to data with high iput dimesio, however, it has bee discovered they do ot favor large datasets, due to the demads imposed o virtual memory, ad the traiig complexity resultat from the use of such a scaled collectio of data [2]. Figure 2. Kerel fuctio: a liearly iseparable iput space ca be mapped to a liearly separable higher-dimmesioal space. Work from Fei, Li, ad Yog [3] highlighted three crucial problems i the use of support vector machies. These are: attaiig the optimal iput subset, correct kerel fuctio, 184

3 Iteratioal Joural of Egieerig ad Advaced Techology (IJEAT) ISSN: , Volume-4 Issue-4, April 2015 ad the optimal parameters of the selected kerel, all of which are prime cosideratios withi this study. III. DATA A. Dataset The direct marketig dataset used i this study was provided by Moro, Laureao, ad Cortez [9], also available i [8]. It cosists of 45,211 samples, each havig 17 attributes, oe of which is the class label. The attributes are both categorical ad umeric ad ca be grouped as: demographical (age, educatio, job, marital status); bak iformatio (balace, prior defaults, loas); direct marketig campaig iformatio (cotact type, duratio, days sice last cotact, outcome of the prior campaig for that cliet, etc.) The dataset is ubalaced, because the successful samples correspodig to the class 'yes' are 5,289, which is 11.7% of all samples. There are o missig values. Further details about data collectio, uderstadig, ad iitial preprocessig steps ca be foud i [9]. With referece to the stadard for data miig projects CRISP-DM [4], we did two data pre-processig trasformatios: mappig o-umeric data ito biary dummies ad ormalizatio. No-umeric categorical variables were decomposed ito a series of dummy biary variables. For example, a sigle variable, such as educatio havig possible values of "ukow", "primary", "secodary", ad "tertiary" would be decomposed ito four separate variables: ukow - 0/1; primary - 0/1; secodary - 0/1; ad tertiary - 0/1. This is a full set of dummy variables, which umber correspods to the umber of possible values. However, i this example oly three of the dummy variables are eed - if values of three are kow, the fourth is also kow. For example, give that these four values are the oly possible oes, we ca kow that if the educatio is either ukow, primary, or secodary, it must be tertiary. Thus we map a categorical variable ito dummies, which are oe less tha the umber of possible values. Usig reduced umber of dummies we coverted the origial dataset variables ito 42 umeric variables altogether, which is 6 less tha the 48 variables used i [10] ad [11]. There are two beefits of that: first, the model becomes simpler ad faster; secodly, avoidig redudacy i data aliviates the SVM problem with demads imposed o virtual memory, ad the traiig complexity with huge umber of support vectors. The model buildig utility we used coverts categorical variables to biary dummies without redudacy. The secod data trasformatio we did is related to ormalizatio/scalig. This procedure attempts to have all iput variables x a with cosistet values, regardless of their origial scale of ad/or differet measuremet uits used, e.g. day (1-31) vs. duratio i secods (0-4918). If the data are left as they are, the traiig process gets iflueced ad biased by some domiatig variables with large values. I order to address this, we did ormalizatio (z-scorig) by: x a,i ew = x µ a,i σ i {1N} (11) where µ is the mea ad σ is the stadard deviatio of x a. After the trasformatio, each iput variable has zero mea ad uit stadard deviatio. B. Variable Importace Referrig to the data preparatio stage of the CRISP-DM project model for data miig [4], we explore how presece or absece of the iput variables preseted to the model for traiig ad testig affects the classifier performace. Removig most irrelevat ad redudat variables from the data may help to alleviate the effect of the curse of dimesioality, ehace the model geeralizatio capabilities, speed up the learig process, ad to improve the model iterpretability. The variable selectio also helps to acquire better uderstadig about data ad how they are related with each other. This work uses Sesitivity Aalysis (SA) for rakig the variable importace to the model by measurig the effects o the output whe the iputs are varied through their rage of values [5]. While iitially proposed for eural ets, SA is curretly used with virtually ay supervised learig techique, such as SVM [6]. The SA varies a iput variable x a through its rage with L levels, uder a regular sequece from the miimum to the maximum value. Let x a, j deotes the j-th level of iput x a. Let ŷ deote the value predicted by the model for oe data sample (x) ad let ŷ = P(x) is the fuctio of model resposes. Kewley, Embrechts, ad Breema propose i [7] three sesitivity measures, amely rage ( S r ), gradiet ( S g ) ad variace ( S v ): S r = max(ŷ aj : j {1,, L}) mi(ŷ aj : j {1,, L}) L S g = ŷ aj ŷ a j 1 / (L 1) j=2 L S v = (ŷ aj y a ) 2 / (L 1) j=2 (12) where y a deotes the mea of the resposes. The gradiet is the oly measure that is depedet o the order of the sesitivity resposes. For all measures, the higher the value, the more relevat is the iput x a. The relative importace r a ca be give by: r a = ς a / (13) M ς i=1 i where ς a is the sesitivity measure for x a (e.g., rage) [5]. IV. EXPERIMENTS AND DISCUSSION I order to explore the SVM performace for task outlied ad compare the model characteristics with those discussed i studies [9], [10], [11], we used the same dataset ad did experimets cosistetly. Further to that, we exteded the methodology addressig the followig gaps: Validatio ad testig. Usig validatio ad test sets i a double-testig procedure helps to avoid overestimatio of the model performace. Radomess ad 'lucky' set compositio. Radom samplig is a fair way to select traiig ad testig sets, but some 'lucky' draws ca trai the model much better tha others. Usig rigorous testig ad validatio procedures we solidify the coclusios made. 185

4 Usig Support Vector Machies for Direct Marketig Models Choice of kerel fuctio. We explore the SVM performace usig the five of the most commom kerel fuctios discussed above. Optimizatio of the model hyper-parameters. We tested the SVM performace with differe hyper-parameters, some of which are specific for the kerel fuctio used. Variable selectio. Further to idetifyig importace of variables ad their cotributio to the classificatio task o the basis of SA, we applied backward selectio procedure to elimiate some iput variables. Data saturatio. We also explored the capacity of the SVM to act i early stages of data collectio where lack of sufficiet data may lead to uderfitted models. All experimets were coducted usig R eviromet [15], [16], ad [17]. I order to select iput variables for elimiatio, we did SA usig three sesitivity measures: rage, gradiet, ad variace by 10 rus of the model per measure. Fig. 3-5 show the iput variable importace, usig a bar plot for each r a i equatio (13), sorted i descedig order. The whiskers i the figures represet cofidece itervals. Two of the measures, rage ad variace, fid loa as the least sigificat iput, while the gradiet measure fids default the oe. Ayway, both iput variables show similar isigificace to the classificatio task. Applyig backward variable selectio procedure by elimiatig first loa, we re-evaluated the iput sigificaces ad further elimiated cotact ad campaig to obtai best results. For the sake of cosistecy with the previous studies, we first used 98% of the origial dataset, which was further split radomly ito traiig ad validatio sets i ratio 2:1. The rest of 2% was used for fial ad idepedet test set. Usig test set i additio to the validatio set solidifies the performace estimatio as the validatio set specifics ca ifluece the search for best hyper- parameters values. Thus, estimatio based o validatio set oly ca get biased. I order to provide more realistic performace results ad reduce the effect of lucky set compositio, each versio of the model was ru 10 times with differet radomly selected traiig ad validatio sets. For each ru, a 3-fold cross-validatio creates 3 model istaces ad averages their results. We iterated all those procedures 10 times per model, recordig ad averagig accuracy ad AUC. Aother part of our experimets was to test how differet levels of data saturatio affect the SVM model performace. I a realistic situatio, buildig a dataset ca be a ogoig process, startig with a small dataset, which grow gradually over the time duratio pdays poutcome age balace previous campaig moth educatio job day marital housig cotact default loa Figure 3. Iput importaces usig rage sesitivity measure duratio pdays poutcome age balace previous moth campaig housig educatio marital job cotact day default loa Figure 4. Iput importaces usig variace sesitivity measure duratio pdays poutcome age balace previous moth campaig educatio job day housig marital cotact loa default Figure 5. Iput importaces usig gradiet sesitivity measure. 186

5 Iteratioal Joural of Egieerig ad Advaced Techology (IJEAT) ISSN: , Volume-4 Issue-4, April 2015 Performace of a classifier traied at differet stages of the dataset lifetime is a importat characteristic, as some modellig techiques may show better results tha other i differet data saturatios. Table 1 summarizes the SVM performace i terms of accuracy ad AUC with differet levels of data saturatio, ragig from 10% to 98% of the origial dataset. Results show that the 20% saturatio yields best average accuracy of % with some lucky sets achievig %. The table also shows a droppig performace whe data saturatio gets higher / lower. This ca be iterpreted as havig the size icreasig / decreasig makes the model to over-fit / uder-fit to the traiig set. The best model here outperforms the best models reported i previous studies [10], [11] with 90.09% max accuracy. Table 1. SVM performace with fractios of the origial dataset used for traiig. Merit 98%set 80%set 60%set 40%set 20%set 10%set ACC% AUC I data miig, classificatio performace is ofte measured usig accuracy as the figure of merit. For a give operatig poit of a classifier, the accuracy is the total umber of correctly classified istaces divided by the total umber of all available istaces. Accuracy, however, varies dramatically depedig o class prevalece. It ca be a misleadig estimator i cases where the most importat class is typically uderrepreseted, such as the class of 'yes' of those who respod positively to the direct marketig. For these applicatios, sesitivity ad specificity ca be more relevat performace estimators. I order to address the accuracy deficiecies, we did Receiver Operatig Characteristics (ROC) aalysis [14]. I a ROC curve, the true positive rate (TPR), a.k.a. sesitivity, is plotted as a fuctio of the false positive rate (FPR), a.k.a. 1-specificity, for differet cut-off poits. Each poit o the ROC plot represets a sesitivity/specificity pair correspodig to a particular decisio threshold. A test with perfect discrimiatio betwee the two classes has a ROC plot that passes through the upper left corer (100% sesitivity, 100% specificity). Therefore the closer the ROC plot is to the upper left corer, the higher the overall accuracy of the test. The area uder the ROC curve (AUC) is a commo measure for the evaluatio of discrimiative power. AUC represets classifier performace over all possible threshold values, i.e. it is threshold idepedet. We used the best performig 20% dataset for traiig ad validatio, iterally split i ito traiig ad validatio sets i ratio 2:1. The fit algorithm rus 10 times with differet radom selectio of traiig ad validatio sets. For each of those set compositios, the 3-fold cross-validatio creates 3 model istaces ad average results. Fig. 6 shows the results by 10 colored lies ad a tick black curve, which is average of the 10 curves. Stadard deviatio bars, aalogous to the whiskers, depict the variace of TPR. True positive rate ROC of SVM traied by 20% of Data False positive rate SVM AUC=0.891 baselie Figure 6. ROC curves of 10 SVM models. Black lie represets average performace. Stadard deviatio bars measure variace. Lift is aother metric, ofte used to measure performace of marketig models. A good performace is whe the respose withi the target is much better tha average for the populatio as a whole. I a cumulative lift chart (gais chart), the y-axis shows the percetage of true positive resposes (TPR). Formally, TPR = sesitivity = TP / (TP + FN) (14) where TP ad FN are true positive ad false egative predictios, respectively. Fig. 7 shows the cumulative lift charts of the 10 SVM models, ru uder the ROC aalysis. The colors ad whiskers i the curves have the same purpose as above. Aother way to characterize performace of a classifier is to look at how precisio ad recall chage as threshold chages. This ca be visualized by precisio-recall curve (Fig. 8). The better the classifier, the closer its curve is to the top-right corer of the graph. Formally, precisio = TP / (TP + FP) (15) recall = TP / (TP + FN) (16) I terms of a direct marketig task, precisio is the percet of correctly idetified 'yes' customers (who purchase the product) amog all reported as 'yes'; recall is the percet of correctly idetified 'yes' customers amog those who are 'yes' i the test set. Recall ad precisio are iversely related: as recall icreases, precisio decreases ad visa versa. Aother factor that affects the SVM performace is the choice of kerel fuctio ad selectig proper values for its parameters, which alog with the misclassificatio cost C, hyper-parameters of the model. We explored empirically the SVM performace with the five kerels outlied i equatios (6)-(10)

6 Usig Support Vector Machies for Direct Marketig Models Cumulative LIFT of SVM True positive rate SVM ALIFT= Rate of positive predictios Figure 7. Cumulative LIFT curves of 10 SVM models. Black lie represets the average. Stadard deviatio bars measure variace. Figure 9. SVM performace with Gaussia RBF kerel ad hyper-parameters C ad sigma (σ). Table 2. SVM optimal hyper-parameters usig differet kerel fuctios. Precisio - Recall of SVM Precisio SVM Hyperparameter liear RBF poly tah Laplacia C sigma /a /a /a degree /a /a 2 /a /a scale /a /a /a offset /a /a /a ACC max % Fially, we built Variable Effect Characteristic (VEC) curves [6] to explore the average impact of the four most importat iput variables x a, which plot the x aj values (x-axis) versus Recall Figure 8. Precisio-Recall curves of 10 SVM models. Black lie represets the average. Stadard deviatio bars measure variace. Table 2 summarizes outcomes. We foud that Gaussia RBF is the best performig kerel i two sets of hyper-parameter values: C=3, sigma=0.089; C=3.5, sigma=0.091, both yeidig max ACC=91.108%. Fig. 9 illustrates grid search for optimal Gaussia RBF hyper-parameters i a rage where the SVM provides the best discrimiatory power. The highest two peaks correspod to the max accuracy obtaied, while multiple rages of both C ad sigma obtai a very good accuracy above 90.7% outperformig the SVM models discussed i the previous studies. The traiig set here is the 20% of the origial oe ad the three iput variables loa, cotact, ad campaig were elimiated. the ŷ aj resposes (y-axis). Betwee two cosecutive x aj values, the VEC plot uses a lie (iterpolatio) for cotiuous values ad a horizotal segmet for categorical data. We ru the model 10 times ad plotted the average values vertically. The whiskers added represet the cofidece itervals. Fig show how duratio, pdays, poutcome, ad age cotribute the model performace. From the duratio VEC is evidet that the last cotact with shortest ad logest duratio cotribute mostly to the positive outcome, whilst a moderate duratio, typically about 2000 sec cotributes to a egative outcome. Similarly, the pdays VEC shows that the sooer the customer is cotacted after the last cotact, the better. The gap betwee the cotacts ca be exteded up to oe year, but ay over-delayed cotact is useless ad cotributes to egative outcome. 188

7 Iteratioal Joural of Egieerig ad Advaced Techology (IJEAT) ISSN: , Volume-4 Issue-4, April Figure 10. VEC curve for the 'duratio' iput Figure 11. VEC curve for the 'pdays' iput. success ukow failure other Figure 12. VEC curve for the 'poutcome' iput. I relatio to the poutcome iput, the VEC curve shows that customers who purchased the product or service are ot likely to purchase it agai ad should t be ivolved i the ew direct marketig campaig, but there is a high chace to sell the product to ay other customers. Fially, the age VEC curve shows that the marketig campaig is better to target mid-age customers betwee 40-50; there is a egligible chace to sell the product to elderly people, particularly above 70. Figure 13. VEC curve for the 'age' iput. V. CONCLUSION This paper presets a case study of data miig modelig techiques for direct marketig. We address some issues which we fid as gaps i previous studies, amely: The most commo partitioig procedure for traiig, validatio, ad test sets uses radom samplig. Although, this is a fair way to select a sample, some 'lucky' draws trai the model much better tha others. Thus, the model istaces show variace i behavior ad characteristics, iflueced by the radomess. I order to address this issue ad further to [9], [10], [11], we used a methodology, which combies cross-validatio (CV), multiple rus over radom selectio of the folds, ad multiple rus over radom selectio of partitios. Each model was tested may times ivolvig 3-fold cross-validatio, radom partitioig ad iteratios. We also applied double-testig procedure with both validatio ad test sets. Aother cotributio is exploratio of the SVM with differet kerels ad differet values of hyper-parameters. The empirical results show that the best performig kerel is the Gaussia RBF with C=3, sigma=0.089; C=3.5, sigma=0.091, both yeidig max ACC=91.108%. We also aalysed how SVM performs with differet levels of data saturatio ad foud that the 20% dataset is best for traiig. We also did aalysis o how iput variable importace affects the model performace ad foud that elimiatig three iputs improve the SVM discrimiatory power. Importace metrics were based o sesitivity aalysis. I coclusio, we believe that a rigorous model aalysis, ivolvig the issues discussed i the paper, lead to solid ad better results. REFERENCES [1] V. Vapik, The Nature of Statistical Learig Theory, Spriger, New York, [2] S. Horg, M. Su, Y. Che, T. Kao, R. Che, J. Lai, ad C. Perkasa, A ovel itrusio detectio system based o hierarchical clusterig ad support vector machies, Expert Systems with Applicatios, vol.38, 2010, pp [3] L. Fei, W. Li, ad H. Yog, Applicatio of least squares support vector machies for discrimiatio of red wie usig visible ad ear ifrared spectroscopy, Itelliget System ad Kowledge Egieerig, vol. 1, 2008, pp [4] P. Chapma, J. Clito, R. Kerber, T. Khabaza, T. Reiartz, C. Shearer, ad R. Wirth, CRISP-DM Step-by-step data miig guide, CRISP-DM Cosortium,

8 Usig Support Vector Machies for Direct Marketig Models [5] P. Cortez, M. Embrechts. Usig sesitivity aalysis ad visualizatio techiques to ope black box data miig models. Iformatio Scieces vol. 225, 2013, pp [6] P. Cortez, A. Cerdeira, F. Almeida, T. Matos, ad J. Reis, Modelig wie prefereces by data miig from physicochemical properties, Decisio Support Systems, vol. 47, o. 4, 2009, pp [7] R. Kewley, M. Embrechts, C. Breema Data strip miig for the virtual desig of pharmaceuticals with eural etworks, IEEE Trasactios o Neural Networks, vol. 11 (3), 2000, pp [8] A. Asucio ad D. Newma, UCI Machie Learig Repository, Uiv. of Califoria Irvie," [Olie], Available: edu/ mlear/mlrepository.html. [9] S. Moro, R. Laureao, P Cortez, "Usig Data Miig for Bak Direct Marketig: A Applicatio of the CRISP-DM Methodology," P. Novais (Ed.), Proceedigs of the Europea Simulatio ad Modellig Coferece - ESM'2011, 2011, pp [10] H. Elsalamoy ad A. Elsayad, Bak Direct Marketig Based o Neural Network, Iteratioal Joural of Egieerig ad Advaced Techology, vol. 2 o. 6, 2013, pp [11] H. Elsalamoy, Bak Direct Marketig Aalysis of Data Miig Techiques, Iteratioal Joural of Computer Applicatios, vol. 85 o. 7, 2014, pp [12] E. Yu ad S. Cho, Costructig respose model usig esemble based o feature subset selectio, Expert Systems with Applicatios, vol. 30 o. 2, 2006, pp [13] H. Shi ad S. Cho, Respose modelig with support vector machies, Expert Systems with Applicatios, vol. 30 o. 4, 2006, pp [14] T. Fawcett, A itroductio to ROC aalysis, Patter Recogitio Letters, vol. 27, o.8, 2005, pp [15] P. Cortez, Data Miig with Neural Networks ad Support Vector Machies usig the R/rmier Tool. Proceedigs of the 10th Idustrial Coferece o Data Miig, Spriger, LNAI 6171, 2010, pp [16] R Developmet Core Team. R: A laguage ad eviromet for statistical computig. R Foudatio for Statistical Computig, [Olie]. Available: [17] T. Sig, O. Sader, N. Beerewikel, ad T. Legauer, ROCR: visualizig classifier performace i R, Bioiformatics vol. 21, o. 20, 2005, pp

Euclidean Distance Based Feature Selection for Fault Detection Prediction Model in Semiconductor Manufacturing Process

Euclidean Distance Based Feature Selection for Fault Detection Prediction Model in Semiconductor Manufacturing Process Vol.133 (Iformatio Techology ad Computer Sciece 016), pp.85-89 http://dx.doi.org/10.1457/astl.016. Euclidea Distace Based Feature Selectio for Fault Detectio Predictio Model i Semicoductor Maufacturig

More information

Designing a learning system

Designing a learning system CS 75 Machie Learig Lecture Desigig a learig system Milos Hauskrecht milos@cs.pitt.edu 539 Seott Square, x-5 people.cs.pitt.edu/~milos/courses/cs75/ Admiistrivia No homework assigmet this week Please try

More information

Improving Template Based Spike Detection

Improving Template Based Spike Detection Improvig Template Based Spike Detectio Kirk Smith, Member - IEEE Portlad State Uiversity petra@ee.pdx.edu Abstract Template matchig algorithms like SSE, Covolutio ad Maximum Likelihood are well kow for

More information

Designing a learning system

Designing a learning system CS 75 Itro to Machie Learig Lecture Desigig a learig system Milos Hauskrecht milos@pitt.edu 539 Seott Square, -5 people.cs.pitt.edu/~milos/courses/cs75/ Admiistrivia No homework assigmet this week Please

More information

3D Model Retrieval Method Based on Sample Prediction

3D Model Retrieval Method Based on Sample Prediction 20 Iteratioal Coferece o Computer Commuicatio ad Maagemet Proc.of CSIT vol.5 (20) (20) IACSIT Press, Sigapore 3D Model Retrieval Method Based o Sample Predictio Qigche Zhag, Ya Tag* School of Computer

More information

ANN WHICH COVERS MLP AND RBF

ANN WHICH COVERS MLP AND RBF ANN WHICH COVERS MLP AND RBF Josef Boští, Jaromír Kual Faculty of Nuclear Scieces ad Physical Egieerig, CTU i Prague Departmet of Software Egieerig Abstract Two basic types of artificial eural etwors Multi

More information

A SOFTWARE MODEL FOR THE MULTILAYER PERCEPTRON

A SOFTWARE MODEL FOR THE MULTILAYER PERCEPTRON A SOFTWARE MODEL FOR THE MULTILAYER PERCEPTRON Roberto Lopez ad Eugeio Oñate Iteratioal Ceter for Numerical Methods i Egieerig (CIMNE) Edificio C1, Gra Capitá s/, 08034 Barceloa, Spai ABSTRACT I this work

More information

Evaluation of Support Vector Machine Kernels for Detecting Network Anomalies

Evaluation of Support Vector Machine Kernels for Detecting Network Anomalies Evaluatio of Support Vector Machie Kerels for Detectig Network Aomalies Prera Batta, Maider Sigh, Zhida Li, Qigye Dig, ad Ljiljaa Trajković Commuicatio Networks Laboratory http://www.esc.sfu.ca/~ljilja/cl/

More information

Cubic Polynomial Curves with a Shape Parameter

Cubic Polynomial Curves with a Shape Parameter roceedigs of the th WSEAS Iteratioal Coferece o Robotics Cotrol ad Maufacturig Techology Hagzhou Chia April -8 00 (pp5-70) Cubic olyomial Curves with a Shape arameter MO GUOLIANG ZHAO YANAN Iformatio ad

More information

Solving Fuzzy Assignment Problem Using Fourier Elimination Method

Solving Fuzzy Assignment Problem Using Fourier Elimination Method Global Joural of Pure ad Applied Mathematics. ISSN 0973-768 Volume 3, Number 2 (207), pp. 453-462 Research Idia Publicatios http://www.ripublicatio.com Solvig Fuzzy Assigmet Problem Usig Fourier Elimiatio

More information

Ones Assignment Method for Solving Traveling Salesman Problem

Ones Assignment Method for Solving Traveling Salesman Problem Joural of mathematics ad computer sciece 0 (0), 58-65 Oes Assigmet Method for Solvig Travelig Salesma Problem Hadi Basirzadeh Departmet of Mathematics, Shahid Chamra Uiversity, Ahvaz, Ira Article history:

More information

DATA MINING II - 1DL460

DATA MINING II - 1DL460 DATA MINING II - 1DL460 Sprig 2017 A secod course i data miig http://www.it.uu.se/edu/course/homepage/ifoutv2/vt17/ Kjell Orsbor Uppsala Database Laboratory Departmet of Iformatio Techology, Uppsala Uiversity,

More information

Pruning and Summarizing the Discovered Time Series Association Rules from Mechanical Sensor Data Qing YANG1,a,*, Shao-Yu WANG1,b, Ting-Ting ZHANG2,c

Pruning and Summarizing the Discovered Time Series Association Rules from Mechanical Sensor Data Qing YANG1,a,*, Shao-Yu WANG1,b, Ting-Ting ZHANG2,c Advaces i Egieerig Research (AER), volume 131 3rd Aual Iteratioal Coferece o Electroics, Electrical Egieerig ad Iformatio Sciece (EEEIS 2017) Pruig ad Summarizig the Discovered Time Series Associatio Rules

More information

Fundamentals of Media Processing. Shin'ichi Satoh Kazuya Kodama Hiroshi Mo Duy-Dinh Le

Fundamentals of Media Processing. Shin'ichi Satoh Kazuya Kodama Hiroshi Mo Duy-Dinh Le Fudametals of Media Processig Shi'ichi Satoh Kazuya Kodama Hiroshi Mo Duy-Dih Le Today's topics Noparametric Methods Parze Widow k-nearest Neighbor Estimatio Clusterig Techiques k-meas Agglomerative Hierarchical

More information

Copyright 2016 Ramez Elmasri and Shamkant B. Navathe

Copyright 2016 Ramez Elmasri and Shamkant B. Navathe Copyright 2016 Ramez Elmasri ad Shamkat B. Navathe CHAPTER 18 Strategies for Query Processig Copyright 2016 Ramez Elmasri ad Shamkat B. Navathe Itroductio DBMS techiques to process a query Scaer idetifies

More information

UNIT 4 Section 8 Estimating Population Parameters using Confidence Intervals

UNIT 4 Section 8 Estimating Population Parameters using Confidence Intervals UNIT 4 Sectio 8 Estimatig Populatio Parameters usig Cofidece Itervals To make ifereces about a populatio that caot be surveyed etirely, sample statistics ca be take from a SRS of the populatio ad used

More information

Math 10C Long Range Plans

Math 10C Long Range Plans Math 10C Log Rage Plas Uits: Evaluatio: Homework, projects ad assigmets 10% Uit Tests. 70% Fial Examiatio.. 20% Ay Uit Test may be rewritte for a higher mark. If the retest mark is higher, that mark will

More information

Pattern Recognition Systems Lab 1 Least Mean Squares

Pattern Recognition Systems Lab 1 Least Mean Squares Patter Recogitio Systems Lab 1 Least Mea Squares 1. Objectives This laboratory work itroduces the OpeCV-based framework used throughout the course. I this assigmet a lie is fitted to a set of poits usig

More information

A New Morphological 3D Shape Decomposition: Grayscale Interframe Interpolation Method

A New Morphological 3D Shape Decomposition: Grayscale Interframe Interpolation Method A ew Morphological 3D Shape Decompositio: Grayscale Iterframe Iterpolatio Method D.. Vizireau Politehica Uiversity Bucharest, Romaia ae@comm.pub.ro R. M. Udrea Politehica Uiversity Bucharest, Romaia mihea@comm.pub.ro

More information

Bayesian approach to reliability modelling for a probability of failure on demand parameter

Bayesian approach to reliability modelling for a probability of failure on demand parameter Bayesia approach to reliability modellig for a probability of failure o demad parameter BÖRCSÖK J., SCHAEFER S. Departmet of Computer Architecture ad System Programmig Uiversity Kassel, Wilhelmshöher Allee

More information

Dynamic Programming and Curve Fitting Based Road Boundary Detection

Dynamic Programming and Curve Fitting Based Road Boundary Detection Dyamic Programmig ad Curve Fittig Based Road Boudary Detectio SHYAM PRASAD ADHIKARI, HYONGSUK KIM, Divisio of Electroics ad Iformatio Egieerig Chobuk Natioal Uiversity 664-4 Ga Deokji-Dog Jeoju-City Jeobuk

More information

Research on Identification Model of Financial Fraud of Listed Company Based on Data Mining Technology

Research on Identification Model of Financial Fraud of Listed Company Based on Data Mining Technology 208 2d Iteratioal Coferece o Systems, Computig, ad Applicatios (SYSTCA 208) Research o Idetificatio Model of Fiacial Fraud of Listed Compay Based o Data Miig Techology Jiaqi Hu, Xiao Che School of Busiess,

More information

Administrative UNSUPERVISED LEARNING. Unsupervised learning. Supervised learning 11/25/13. Final project. No office hours today

Administrative UNSUPERVISED LEARNING. Unsupervised learning. Supervised learning 11/25/13. Final project. No office hours today Admiistrative Fial project No office hours today UNSUPERVISED LEARNING David Kauchak CS 451 Fall 2013 Supervised learig Usupervised learig label label 1 label 3 model/ predictor label 4 label 5 Supervised

More information

( n+1 2 ) , position=(7+1)/2 =4,(median is observation #4) Median=10lb

( n+1 2 ) , position=(7+1)/2 =4,(median is observation #4) Median=10lb Chapter 3 Descriptive Measures Measures of Ceter (Cetral Tedecy) These measures will tell us where is the ceter of our data or where most typical value of a data set lies Mode the value that occurs most

More information

Investigating methods for improving Bagged k-nn classifiers

Investigating methods for improving Bagged k-nn classifiers Ivestigatig methods for improvig Bagged k-nn classifiers Fuad M. Alkoot Telecommuicatio & Navigatio Istitute, P.A.A.E.T. P.O.Box 4575, Alsalmia, 22046 Kuwait Abstract- We experimet with baggig knn classifiers

More information

Algorithms for Disk Covering Problems with the Most Points

Algorithms for Disk Covering Problems with the Most Points Algorithms for Disk Coverig Problems with the Most Poits Bi Xiao Departmet of Computig Hog Kog Polytechic Uiversity Hug Hom, Kowloo, Hog Kog csbxiao@comp.polyu.edu.hk Qigfeg Zhuge, Yi He, Zili Shao, Edwi

More information

EMPIRICAL ANALYSIS OF FAULT PREDICATION TECHNIQUES FOR IMPROVING SOFTWARE PROCESS CONTROL

EMPIRICAL ANALYSIS OF FAULT PREDICATION TECHNIQUES FOR IMPROVING SOFTWARE PROCESS CONTROL Iteratioal Joural of Iformatio Techology ad Kowledge Maagemet July-December 2012, Volume 5, No. 2, pp. 371-375 EMPIRICAL ANALYSIS OF FAULT PREDICATION TECHNIQUES FOR IMPROVING SOFTWARE PROCESS CONTROL

More information

Analysis of Documents Clustering Using Sampled Agglomerative Technique

Analysis of Documents Clustering Using Sampled Agglomerative Technique Aalysis of Documets Clusterig Usig Sampled Agglomerative Techique Omar H. Karam, Ahmed M. Hamad, ad Sheri M. Moussa Abstract I this paper a clusterig algorithm for documets is proposed that adapts a samplig-based

More information

Criterion in selecting the clustering algorithm in Radial Basis Functional Link Nets

Criterion in selecting the clustering algorithm in Radial Basis Functional Link Nets WSEAS TRANSACTIONS o SYSTEMS Ag Sau Loog, Og Hog Choo, Low Heg Chi Criterio i selectig the clusterig algorithm i Radial Basis Fuctioal Lik Nets ANG SAU LOONG 1, ONG HONG CHOON 2 & LOW HENG CHIN 3 Departmet

More information

Improvement of the Orthogonal Code Convolution Capabilities Using FPGA Implementation

Improvement of the Orthogonal Code Convolution Capabilities Using FPGA Implementation Improvemet of the Orthogoal Code Covolutio Capabilities Usig FPGA Implemetatio Naima Kaabouch, Member, IEEE, Apara Dhirde, Member, IEEE, Saleh Faruque, Member, IEEE Departmet of Electrical Egieerig, Uiversity

More information

South Slave Divisional Education Council. Math 10C

South Slave Divisional Education Council. Math 10C South Slave Divisioal Educatio Coucil Math 10C Curriculum Package February 2012 12 Strad: Measuremet Geeral Outcome: Develop spatial sese ad proportioal reasoig It is expected that studets will: 1. Solve

More information

Image Segmentation EEE 508

Image Segmentation EEE 508 Image Segmetatio Objective: to determie (etract) object boudaries. It is a process of partitioig a image ito distict regios by groupig together eighborig piels based o some predefied similarity criterio.

More information

Journal of Chemical and Pharmaceutical Research, 2013, 5(12): Research Article

Journal of Chemical and Pharmaceutical Research, 2013, 5(12): Research Article Available olie www.jocpr.com Joural of Chemical ad Pharmaceutical Research, 2013, 5(12):745-749 Research Article ISSN : 0975-7384 CODEN(USA) : JCPRC5 K-meas algorithm i the optimal iitial cetroids based

More information

New HSL Distance Based Colour Clustering Algorithm

New HSL Distance Based Colour Clustering Algorithm The 4th Midwest Artificial Itelligece ad Cogitive Scieces Coferece (MAICS 03 pp 85-9 New Albay Idiaa USA April 3-4 03 New HSL Distace Based Colour Clusterig Algorithm Vasile Patrascu Departemet of Iformatics

More information

Data Analysis. Concepts and Techniques. Chapter 2. Chapter 2: Getting to Know Your Data. Data Objects and Attribute Types

Data Analysis. Concepts and Techniques. Chapter 2. Chapter 2: Getting to Know Your Data. Data Objects and Attribute Types Data Aalysis Cocepts ad Techiques Chapter 2 1 Chapter 2: Gettig to Kow Your Data Data Objects ad Attribute Types Basic Statistical Descriptios of Data Data Visualizatio Measurig Data Similarity ad Dissimilarity

More information

A new algorithm to build feed forward neural networks.

A new algorithm to build feed forward neural networks. A ew algorithm to build feed forward eural etworks. Amit Thombre Cetre of Excellece, Software Techologies ad Kowledge Maagemet, Tech Mahidra, Pue, Idia Abstract The paper presets a ew algorithm to build

More information

Creating Exact Bezier Representations of CST Shapes. David D. Marshall. California Polytechnic State University, San Luis Obispo, CA , USA

Creating Exact Bezier Representations of CST Shapes. David D. Marshall. California Polytechnic State University, San Luis Obispo, CA , USA Creatig Exact Bezier Represetatios of CST Shapes David D. Marshall Califoria Polytechic State Uiversity, Sa Luis Obispo, CA 93407-035, USA The paper presets a method of expressig CST shapes pioeered by

More information

Neural Networks A Model of Boolean Functions

Neural Networks A Model of Boolean Functions Neural Networks A Model of Boolea Fuctios Berd Steibach, Roma Kohut Freiberg Uiversity of Miig ad Techology Istitute of Computer Sciece D-09596 Freiberg, Germay e-mails: steib@iformatik.tu-freiberg.de

More information

An Efficient Algorithm for Graph Bisection of Triangularizations

An Efficient Algorithm for Graph Bisection of Triangularizations A Efficiet Algorithm for Graph Bisectio of Triagularizatios Gerold Jäger Departmet of Computer Sciece Washigto Uiversity Campus Box 1045 Oe Brookigs Drive St. Louis, Missouri 63130-4899, USA jaegerg@cse.wustl.edu

More information

Feature Selection for Change Detection in Multivariate Time-Series

Feature Selection for Change Detection in Multivariate Time-Series Feature Selectio for Chage Detectio i Multivariate Time-Series Michael Botsch Istitute for Circuit Theory ad Sigal Processig Techical Uiversity Muich 80333 Muich, Germay Email: botsch@tum.de Josef A. Nossek

More information

Morgan Kaufmann Publishers 26 February, COMPUTER ORGANIZATION AND DESIGN The Hardware/Software Interface. Chapter 5

Morgan Kaufmann Publishers 26 February, COMPUTER ORGANIZATION AND DESIGN The Hardware/Software Interface. Chapter 5 Morga Kaufma Publishers 26 February, 28 COMPUTER ORGANIZATION AND DESIGN The Hardware/Software Iterface 5 th Editio Chapter 5 Set-Associative Cache Architecture Performace Summary Whe CPU performace icreases:

More information

Software Fault Prediction of Unlabeled Program Modules

Software Fault Prediction of Unlabeled Program Modules Software Fault Predictio of Ulabeled Program Modules C. Catal, U. Sevim, ad B. Diri, Member, IAENG Abstract Software metrics ad fault data belogig to a previous software versio are used to build the software

More information

Normal Distributions

Normal Distributions Normal Distributios Stacey Hacock Look at these three differet data sets Each histogram is overlaid with a curve : A B C A) Weights (g) of ewly bor lab rat pups B) Mea aual temperatures ( F ) i A Arbor,

More information

Lecture 13: Validation

Lecture 13: Validation Lecture 3: Validatio Resampli methods Holdout Cross Validatio Radom Subsampli -Fold Cross-Validatio Leave-oe-out The Bootstrap Bias ad variace estimatio Three-way data partitioi Itroductio to Patter Recoitio

More information

Polynomial Functions and Models. Learning Objectives. Polynomials. P (x) = a n x n + a n 1 x n a 1 x + a 0, a n 0

Polynomial Functions and Models. Learning Objectives. Polynomials. P (x) = a n x n + a n 1 x n a 1 x + a 0, a n 0 Polyomial Fuctios ad Models 1 Learig Objectives 1. Idetify polyomial fuctios ad their degree 2. Graph polyomial fuctios usig trasformatios 3. Idetify the real zeros of a polyomial fuctio ad their multiplicity

More information

Analysis of Server Resource Consumption of Meteorological Satellite Application System Based on Contour Curve

Analysis of Server Resource Consumption of Meteorological Satellite Application System Based on Contour Curve Advaces i Computer, Sigals ad Systems (2018) 2: 19-25 Clausius Scietific Press, Caada Aalysis of Server Resource Cosumptio of Meteorological Satellite Applicatio System Based o Cotour Curve Xiagag Zhao

More information

Accuracy Improvement in Camera Calibration

Accuracy Improvement in Camera Calibration Accuracy Improvemet i Camera Calibratio FaJie L Qi Zag ad Reihard Klette CITR, Computer Sciece Departmet The Uiversity of Aucklad Tamaki Campus, Aucklad, New Zealad fli006, qza001@ec.aucklad.ac.z r.klette@aucklad.ac.z

More information

IMP: Superposer Integrated Morphometrics Package Superposition Tool

IMP: Superposer Integrated Morphometrics Package Superposition Tool IMP: Superposer Itegrated Morphometrics Package Superpositio Tool Programmig by: David Lieber ( 03) Caisius College 200 Mai St. Buffalo, NY 4208 Cocept by: H. David Sheets, Dept. of Physics, Caisius College

More information

A Note on Least-norm Solution of Global WireWarping

A Note on Least-norm Solution of Global WireWarping A Note o Least-orm Solutio of Global WireWarpig Charlie C. L. Wag Departmet of Mechaical ad Automatio Egieerig The Chiese Uiversity of Hog Kog Shati, N.T., Hog Kog E-mail: cwag@mae.cuhk.edu.hk Abstract

More information

An Efficient Algorithm for Graph Bisection of Triangularizations

An Efficient Algorithm for Graph Bisection of Triangularizations Applied Mathematical Scieces, Vol. 1, 2007, o. 25, 1203-1215 A Efficiet Algorithm for Graph Bisectio of Triagularizatios Gerold Jäger Departmet of Computer Sciece Washigto Uiversity Campus Box 1045, Oe

More information

New Fuzzy Color Clustering Algorithm Based on hsl Similarity

New Fuzzy Color Clustering Algorithm Based on hsl Similarity IFSA-EUSFLAT 009 New Fuzzy Color Clusterig Algorithm Based o hsl Similarity Vasile Ptracu Departmet of Iformatics Techology Tarom Compay Bucharest Romaia Email: patrascu.v@gmail.com Abstract I this paper

More information

Mobile terminal 3D image reconstruction program development based on Android Lin Qinhua

Mobile terminal 3D image reconstruction program development based on Android Lin Qinhua Iteratioal Coferece o Automatio, Mechaical Cotrol ad Computatioal Egieerig (AMCCE 05) Mobile termial 3D image recostructio program developmet based o Adroid Li Qihua Sichua Iformatio Techology College

More information

Performance Plus Software Parameter Definitions

Performance Plus Software Parameter Definitions Performace Plus+ Software Parameter Defiitios/ Performace Plus Software Parameter Defiitios Chapma Techical Note-TG-5 paramete.doc ev-0-03 Performace Plus+ Software Parameter Defiitios/2 Backgroud ad Defiitios

More information

BASED ON ITERATIVE ERROR-CORRECTION

BASED ON ITERATIVE ERROR-CORRECTION A COHPARISO OF CRYPTAALYTIC PRICIPLES BASED O ITERATIVE ERROR-CORRECTIO Miodrag J. MihaljeviC ad Jova Dj. GoliC Istitute of Applied Mathematics ad Electroics. Belgrade School of Electrical Egieerig. Uiversity

More information

Learning to Shoot a Goal Lecture 8: Learning Models and Skills

Learning to Shoot a Goal Lecture 8: Learning Models and Skills Learig to Shoot a Goal Lecture 8: Learig Models ad Skills How do we acquire skill at shootig goals? CS 344R/393R: Robotics Bejami Kuipers Learig to Shoot a Goal The robot eeds to shoot the ball i the goal.

More information

1 Enterprise Modeler

1 Enterprise Modeler 1 Eterprise Modeler Itroductio I BaaERP, a Busiess Cotrol Model ad a Eterprise Structure Model for multi-site cofiguratios are itroduced. Eterprise Structure Model Busiess Cotrol Models Busiess Fuctio

More information

SOFTWARE usually does not work alone. It must have

SOFTWARE usually does not work alone. It must have Proceedigs of the 203 Federated Coferece o Computer Sciece ad Iformatio Systems pp. 343 348 A method for selectig eviromets for software compatibility testig Łukasz Pobereżik AGH Uiversity of Sciece ad

More information

An Improved Shuffled Frog-Leaping Algorithm for Knapsack Problem

An Improved Shuffled Frog-Leaping Algorithm for Knapsack Problem A Improved Shuffled Frog-Leapig Algorithm for Kapsack Problem Zhoufag Li, Ya Zhou, ad Peg Cheg School of Iformatio Sciece ad Egieerig Hea Uiversity of Techology ZhegZhou, Chia lzhf1978@126.com Abstract.

More information

EM375 STATISTICS AND MEASUREMENT UNCERTAINTY LEAST SQUARES LINEAR REGRESSION ANALYSIS

EM375 STATISTICS AND MEASUREMENT UNCERTAINTY LEAST SQUARES LINEAR REGRESSION ANALYSIS EM375 STATISTICS AND MEASUREMENT UNCERTAINTY LEAST SQUARES LINEAR REGRESSION ANALYSIS I this uit of the course we ivestigate fittig a straight lie to measured (x, y) data pairs. The equatio we wat to fit

More information

Chapter 1. Introduction to Computers and C++ Programming. Copyright 2015 Pearson Education, Ltd.. All rights reserved.

Chapter 1. Introduction to Computers and C++ Programming. Copyright 2015 Pearson Education, Ltd.. All rights reserved. Chapter 1 Itroductio to Computers ad C++ Programmig Copyright 2015 Pearso Educatio, Ltd.. All rights reserved. Overview 1.1 Computer Systems 1.2 Programmig ad Problem Solvig 1.3 Itroductio to C++ 1.4 Testig

More information

Numerical Methods Lecture 6 - Curve Fitting Techniques

Numerical Methods Lecture 6 - Curve Fitting Techniques Numerical Methods Lecture 6 - Curve Fittig Techiques Topics motivatio iterpolatio liear regressio higher order polyomial form expoetial form Curve fittig - motivatio For root fidig, we used a give fuctio

More information

Our second algorithm. Comp 135 Machine Learning Computer Science Tufts University. Decision Trees. Decision Trees. Decision Trees.

Our second algorithm. Comp 135 Machine Learning Computer Science Tufts University. Decision Trees. Decision Trees. Decision Trees. Comp 135 Machie Learig Computer Sciece Tufts Uiversity Fall 2017 Roi Khardo Some of these slides were adapted from previous slides by Carla Brodley Our secod algorithm Let s look at a simple dataset for

More information

Evaluation scheme for Tracking in AMI

Evaluation scheme for Tracking in AMI A M I C o m m u i c a t i o A U G M E N T E D M U L T I - P A R T Y I N T E R A C T I O N http://www.amiproject.org/ Evaluatio scheme for Trackig i AMI S. Schreiber a D. Gatica-Perez b AMI WP4 Trackig:

More information

Assignment Problems with fuzzy costs using Ones Assignment Method

Assignment Problems with fuzzy costs using Ones Assignment Method IOSR Joural of Mathematics (IOSR-JM) e-issn: 8-8, p-issn: 9-6. Volume, Issue Ver. V (Sep. - Oct.06), PP 8-89 www.iosrjourals.org Assigmet Problems with fuzzy costs usig Oes Assigmet Method S.Vimala, S.Krisha

More information

Empirical Validate C&K Suite for Predict Fault-Proneness of Object-Oriented Classes Developed Using Fuzzy Logic.

Empirical Validate C&K Suite for Predict Fault-Proneness of Object-Oriented Classes Developed Using Fuzzy Logic. Empirical Validate C&K Suite for Predict Fault-Proeess of Object-Orieted Classes Developed Usig Fuzzy Logic. Mohammad Amro 1, Moataz Ahmed 1, Kaaa Faisal 2 1 Iformatio ad Computer Sciece Departmet, Kig

More information

9.1. Sequences and Series. Sequences. What you should learn. Why you should learn it. Definition of Sequence

9.1. Sequences and Series. Sequences. What you should learn. Why you should learn it. Definition of Sequence _9.qxd // : AM Page Chapter 9 Sequeces, Series, ad Probability 9. Sequeces ad Series What you should lear Use sequece otatio to write the terms of sequeces. Use factorial otatio. Use summatio otatio to

More information

Identification of the Swiss Z24 Highway Bridge by Frequency Domain Decomposition Brincker, Rune; Andersen, P.

Identification of the Swiss Z24 Highway Bridge by Frequency Domain Decomposition Brincker, Rune; Andersen, P. Aalborg Uiversitet Idetificatio of the Swiss Z24 Highway Bridge by Frequecy Domai Decompositio Bricker, Rue; Aderse, P. Published i: Proceedigs of IMAC 2 Publicatio date: 22 Documet Versio Publisher's

More information

Probabilistic Fuzzy Time Series Method Based on Artificial Neural Network

Probabilistic Fuzzy Time Series Method Based on Artificial Neural Network America Joural of Itelliget Systems 206, 6(2): 42-47 DOI: 0.5923/j.ajis.2060602.02 Probabilistic Fuzzy Time Series Method Based o Artificial Neural Network Erol Egrioglu,*, Ere Bas, Cagdas Haka Aladag

More information

Harris Corner Detection Algorithm at Sub-pixel Level and Its Application Yuanfeng Han a, Peijiang Chen b * and Tian Meng c

Harris Corner Detection Algorithm at Sub-pixel Level and Its Application Yuanfeng Han a, Peijiang Chen b * and Tian Meng c Iteratioal Coferece o Computatioal Sciece ad Egieerig (ICCSE 015) Harris Corer Detectio Algorithm at Sub-pixel Level ad Its Applicatio Yuafeg Ha a, Peijiag Che b * ad Tia Meg c School of Automobile, Liyi

More information

Mining from Quantitative Data with Linguistic Minimum Supports and Confidences

Mining from Quantitative Data with Linguistic Minimum Supports and Confidences Miig from Quatitative Data with Liguistic Miimum Supports ad Cofideces Tzug-Pei Hog, Mig-Jer Chiag ad Shyue-Liag Wag Departmet of Electrical Egieerig Natioal Uiversity of Kaohsiug Kaohsiug, 8, Taiwa, R.O.C.

More information

An improved support vector machine based on particle swarm optimization in laser ultrasonic defect detection

An improved support vector machine based on particle swarm optimization in laser ultrasonic defect detection A improved support vector machie based o particle swarm optimizatio i laser ultrasoic defect detectio School of Sciece, North Uiversity of Chia, aiyua, Shaxi 35, Chia xut98@63.com,hhp9@63.com,xywag@6.com,43497@qq.com

More information

Fuzzy Rule Selection by Data Mining Criteria and Genetic Algorithms

Fuzzy Rule Selection by Data Mining Criteria and Genetic Algorithms Fuzzy Rule Selectio by Data Miig Criteria ad Geetic Algorithms Hisao Ishibuchi Dept. of Idustrial Egieerig Osaka Prefecture Uiversity 1-1 Gakue-cho, Sakai, Osaka 599-8531, JAPAN E-mail: hisaoi@ie.osakafu-u.ac.jp

More information

Bezier curves. Figure 2 shows cubic Bezier curves for various control points. In a Bezier curve, only

Bezier curves. Figure 2 shows cubic Bezier curves for various control points. In a Bezier curve, only Edited: Yeh-Liag Hsu (998--; recommeded: Yeh-Liag Hsu (--9; last updated: Yeh-Liag Hsu (9--7. Note: This is the course material for ME55 Geometric modelig ad computer graphics, Yua Ze Uiversity. art of

More information

arxiv: v2 [cs.ds] 24 Mar 2018

arxiv: v2 [cs.ds] 24 Mar 2018 Similar Elemets ad Metric Labelig o Complete Graphs arxiv:1803.08037v [cs.ds] 4 Mar 018 Pedro F. Felzeszwalb Brow Uiversity Providece, RI, USA pff@brow.edu March 8, 018 We cosider a problem that ivolves

More information

Lip Contour Extraction Based on Support Vector Machine

Lip Contour Extraction Based on Support Vector Machine Lip Cotour Extractio Based o Support Vector Machie Author Pa, Xiaosheg, Kog, Jiagpig, Liew, Ala Wee-Chug Published 008 Coferece Title CISP 008 : Proceedigs, First Iteratioal Cogress o Image ad Sigal Processig

More information

Stone Images Retrieval Based on Color Histogram

Stone Images Retrieval Based on Color Histogram Stoe Images Retrieval Based o Color Histogram Qiag Zhao, Jie Yag, Jigyi Yag, Hogxig Liu School of Iformatio Egieerig, Wuha Uiversity of Techology Wuha, Chia Abstract Stoe images color features are chose

More information

x x 2 x Iput layer = quatity of classificatio mode X T = traspositio matrix The core of such coditioal probability estimatig method is calculatig the

x x 2 x Iput layer = quatity of classificatio mode X T = traspositio matrix The core of such coditioal probability estimatig method is calculatig the COMPARATIVE RESEARCHES ON PROBABILISTIC NEURAL NETWORKS AND MULTI-LAYER PERCEPTRON NETWORKS FOR REMOTE SENSING IMAGE SEGMENTATION Liu Gag a, b, * a School of Electroic Iformatio, Wuha Uiversity, 430079,

More information

Image based Cats and Possums Identification for Intelligent Trapping Systems

Image based Cats and Possums Identification for Intelligent Trapping Systems Volume 159 No, February 017 Image based Cats ad Possums Idetificatio for Itelliget Trappig Systems T. A. S. Achala Perera School of Egieerig Aucklad Uiversity of Techology New Zealad Joh Collis School

More information

Intrusion Detection Method Using Protocol Classification and Rough Set Based Support Vector Machine

Intrusion Detection Method Using Protocol Classification and Rough Set Based Support Vector Machine Computer ad formatio Sciece trusio Detectio Method Usig Protocol Classificatio ad Rough Set Based Support Vector Machie Xuyi Re College of Computer Sciece, Najig Uiversity of Post & Telecommuicatios Najig

More information

Reversible Realization of Quaternary Decoder, Multiplexer, and Demultiplexer Circuits

Reversible Realization of Quaternary Decoder, Multiplexer, and Demultiplexer Circuits Egieerig Letters, :, EL Reversible Realizatio of Quaterary Decoder, Multiplexer, ad Demultiplexer Circuits Mozammel H.. Kha, Member, ENG bstract quaterary reversible circuit is more compact tha the correspodig

More information

Redundancy Allocation for Series Parallel Systems with Multiple Constraints and Sensitivity Analysis

Redundancy Allocation for Series Parallel Systems with Multiple Constraints and Sensitivity Analysis IOSR Joural of Egieerig Redudacy Allocatio for Series Parallel Systems with Multiple Costraits ad Sesitivity Aalysis S. V. Suresh Babu, D.Maheswar 2, G. Ragaath 3 Y.Viaya Kumar d G.Sakaraiah e (Mechaical

More information

Clustering and Classifying Diabetic Data Sets Using K-Means Algorithm

Clustering and Classifying Diabetic Data Sets Using K-Means Algorithm Article ca be accessed olie at http://www.publishigidia.com Clusterig ad Classifyig Diabetic Data Sets Usig K-Meas Algorithm M. Kothaiayaki*, P. Thagaraj** Abstract The k-meas algorithm is well kow for

More information

Image Analysis. Segmentation by Fitting a Model

Image Analysis. Segmentation by Fitting a Model Image Aalysis Segmetatio by Fittig a Model Christophoros Nikou cikou@cs.uoi.gr Images take from: D. Forsyth ad J. Poce. Computer Visio: A Moder Approach, Pretice Hall, 2003. Computer Visio course by Svetlaa

More information

Text Feature Selection based on Feature Dispersion Degree and Feature Concentration Degree

Text Feature Selection based on Feature Dispersion Degree and Feature Concentration Degree Available olie at www.ijpe-olie.com vol. 13, o. 7, November 017, pp. 1159-1164 DOI: 10.3940/ijpe.17.07.p19.11591164 Text Feature Selectio based o Feature Dispersio Degree ad Feature Cocetratio Degree Zhifeg

More information

A Parallel DFA Minimization Algorithm

A Parallel DFA Minimization Algorithm A Parallel DFA Miimizatio Algorithm Ambuj Tewari, Utkarsh Srivastava, ad P. Gupta Departmet of Computer Sciece & Egieerig Idia Istitute of Techology Kapur Kapur 208 016,INDIA pg@iitk.ac.i Abstract. I this

More information

Counting the Number of Minimum Roman Dominating Functions of a Graph

Counting the Number of Minimum Roman Dominating Functions of a Graph Coutig the Number of Miimum Roma Domiatig Fuctios of a Graph SHI ZHENG ad KOH KHEE MENG, Natioal Uiversity of Sigapore We provide two algorithms coutig the umber of miimum Roma domiatig fuctios of a graph

More information

CSCI 5090/7090- Machine Learning. Spring Mehdi Allahyari Georgia Southern University

CSCI 5090/7090- Machine Learning. Spring Mehdi Allahyari Georgia Southern University CSCI 5090/7090- Machie Learig Sprig 018 Mehdi Allahyari Georgia Souther Uiversity Clusterig (slides borrowed from Tom Mitchell, Maria Floria Balca, Ali Borji, Ke Che) 1 Clusterig, Iformal Goals Goal: Automatically

More information

Load balanced Parallel Prime Number Generator with Sieve of Eratosthenes on Cluster Computers *

Load balanced Parallel Prime Number Generator with Sieve of Eratosthenes on Cluster Computers * Load balaced Parallel Prime umber Geerator with Sieve of Eratosthees o luster omputers * Soowook Hwag*, Kyusik hug**, ad Dogseug Kim* *Departmet of Electrical Egieerig Korea Uiversity Seoul, -, Rep. of

More information

Performance Comparisons of PSO based Clustering

Performance Comparisons of PSO based Clustering Performace Comparisos of PSO based Clusterig Suresh Chadra Satapathy, 2 Guaidhi Pradha, 3 Sabyasachi Pattai, 4 JVR Murthy, 5 PVGD Prasad Reddy Ail Neeruoda Istitute of Techology ad Scieces, Sagivalas,Vishaapatam

More information

A Kernel Density Based Approach for Large Scale Image Retrieval

A Kernel Density Based Approach for Large Scale Image Retrieval A Kerel Desity Based Approach for Large Scale Image Retrieval Wei Tog Departmet of Computer Sciece ad Egieerig Michiga State Uiversity East Lasig, MI, USA togwei@cse.msu.edu Rog Ji Departmet of Computer

More information

BAYESIAN WITH FULL CONDITIONAL POSTERIOR DISTRIBUTION APPROACH FOR SOLUTION OF COMPLEX MODELS. Pudji Ismartini

BAYESIAN WITH FULL CONDITIONAL POSTERIOR DISTRIBUTION APPROACH FOR SOLUTION OF COMPLEX MODELS. Pudji Ismartini Proceedig of Iteratioal Coferece O Research, Implemetatio Ad Educatio Of Mathematics Ad Scieces 014, Yogyakarta State Uiversity, 18-0 May 014 BAYESIAN WIH FULL CONDIIONAL POSERIOR DISRIBUION APPROACH FOR

More information

Parabolic Path to a Best Best-Fit Line:

Parabolic Path to a Best Best-Fit Line: Studet Activity : Fidig the Least Squares Regressio Lie By Explorig the Relatioship betwee Slope ad Residuals Objective: How does oe determie a best best-fit lie for a set of data? Eyeballig it may be

More information

Copyright 2016 Ramez Elmasri and Shamkant B. Navathe

Copyright 2016 Ramez Elmasri and Shamkant B. Navathe Copyright 2016 Ramez Elmasri ad Shamkat B. Navathe CHAPTER 19 Query Optimizatio Copyright 2016 Ramez Elmasri ad Shamkat B. Navathe Itroductio Query optimizatio Coducted by a query optimizer i a DBMS Goal:

More information

A Novel Hybrid Algorithm for Software Cost Estimation Based on Cuckoo Optimization and K-Nearest Neighbors Algorithms

A Novel Hybrid Algorithm for Software Cost Estimation Based on Cuckoo Optimization and K-Nearest Neighbors Algorithms Egieerig, Techology & Applied Sciece Research Vol. 6, No. 3, 26, 8-22 8 A Novel Hybrid Algorithm for Software Cost Estimatio Based o Optimizatio ad K-Nearest Neighbors Algorithms Elaz Eskadaria Miadoab

More information

Optimization for framework design of new product introduction management system Ma Ying, Wu Hongcui

Optimization for framework design of new product introduction management system Ma Ying, Wu Hongcui 2d Iteratioal Coferece o Electrical, Computer Egieerig ad Electroics (ICECEE 2015) Optimizatio for framework desig of ew product itroductio maagemet system Ma Yig, Wu Hogcui Tiaji Electroic Iformatio Vocatioal

More information

Data-Driven Nonlinear Hebbian Learning Method for Fuzzy Cognitive Maps

Data-Driven Nonlinear Hebbian Learning Method for Fuzzy Cognitive Maps Data-Drive Noliear Hebbia Learig Method for Fuzzy Cogitive Maps Wociech Stach, Lukasz Kurga, ad Witold Pedrycz Abstract Fuzzy Cogitive Maps (FCMs) are a coveiet tool for modelig of dyamic systems by meas

More information

One advantage that SONAR has over any other music-sequencing product I ve worked

One advantage that SONAR has over any other music-sequencing product I ve worked *gajedra* D:/Thomso_Learig_Projects/Garrigus_163132/z_productio/z_3B2_3D_files/Garrigus_163132_ch17.3d, 14/11/08/16:26:39, 16:26, page: 647 17 CAL 101 Oe advatage that SONAR has over ay other music-sequecig

More information

Python Programming: An Introduction to Computer Science

Python Programming: An Introduction to Computer Science Pytho Programmig: A Itroductio to Computer Sciece Chapter 1 Computers ad Programs 1 Objectives To uderstad the respective roles of hardware ad software i a computig system. To lear what computer scietists

More information

9 x and g(x) = 4. x. Find (x) 3.6. I. Combining Functions. A. From Equations. Example: Let f(x) = and its domain. Example: Let f(x) = and g(x) = x x 4

9 x and g(x) = 4. x. Find (x) 3.6. I. Combining Functions. A. From Equations. Example: Let f(x) = and its domain. Example: Let f(x) = and g(x) = x x 4 1 3.6 I. Combiig Fuctios A. From Equatios Example: Let f(x) = 9 x ad g(x) = 4 f x. Fid (x) g ad its domai. 4 Example: Let f(x) = ad g(x) = x x 4. Fid (f-g)(x) B. From Graphs: Graphical Additio. Example:

More information

Comparison of classification algorithms in the task of object recognition on radar images of the MSTAR base

Comparison of classification algorithms in the task of object recognition on radar images of the MSTAR base Compariso of classificatio algorithms i the task of object recogitio o radar images of the MSTAR base A.A. Borodiov 1, V.V. Myasikov 1,2 1 Samara Natioal Research Uiversity, 34 Moskovskoe Shosse, 443086,

More information