RHUIET : Discovery of Rare High Utility Itemsets using Enumeration Tree

Transcription

1 International Journal for Research in Engineering Application & Management (IJREAM) ISSN : Vol-4, Issue-3, June 218 RHUIET : Discovery of Rare High Utility Itemsets using Enumeration Tree Mrs. C.Sivamathi, Ph.D Research Scholar, Department of Computer Science, Bharathiar, University, Coimbatore, India, c.sivamathi@gmail.com Dr. S.Vijayarani, Assistant Professor, Department of Computer Science, Bharathiar University, Coimbatore, India, vijimohan_2@yahoo.com Abstract: Utility mining has various forms, i.e. frequent utility mining, sequential utility pattern mining, utility based association rule mining and rare high utility mining. Among this, utility based rare itemset mining has attracted many researchers because utility based rare itemset mining considers semantic importance of the items and its occurrence. In this research work, a novel algorithm RHUIET (Rare High Utility Itemset using Enumeration Tree), was proposed to retrieve rare high utility itemsets. The itemsets which yields more profit, with fewer occurrences are termed as rare high utility itemsets. Hence, rare high utility itemsets have their utility is greater than minimum utility threshold and its support is less than minimum support threshold. This algorithm RHUIET, uses enumeration tree structure for generating candidate itemsets. Experimental results were conducted using Chess and Foodmart datasets. Execution time, Memory space and number of itemsets retrieved are analyzed. Key Words: Rare itemset mining, High Utility Rare itemset, Minimum utility threshold, Minimum support threshold. I. INTRODUCTION Utility itemset mining plays an important role in many applications like sales of items in retail marketing, purchasing behavior of customers, web log sequence analysis, analysis in protein sequences, analysis in XML query access patterns etc [1]. But traditional frequent based mining considers only the occurrence of sequence, not the semantic importance of the sequence. Hence utility mining was introduced, it considers the significance of the itemsets [2] [3]. This mining results the itemsets whose utility value is greater than minimum utility threshold value and its support less than minimum support threshold. In real time applications, there is a need to retrieve patterns both with high utility and rare occurrences. For example in retail marketing to promote cross marketing, we need to identify such high utility rare itemsets. It can also be applicable in web log sequence analysis to identify which websites were least frequently visited with high utility value. For such cases we need to retrieve patterns by considering both support and utility values. These types of patterns are termed as rare high utility patterns. These patterns must satisfy two properties: one is, the support of the pattern must be less than the minimum support threshold and the other is, the utility of the pattern must be greater than minimum utility threshold. The remaining section of the paper is organized as follows: Section 2 describes related works, proposed work was given under section 3, Section 4 explains the working of algorithm with example. Scetion 5 illustrates experimental results and section 6 concludes the work. II. RELATED WORKS In 26, H. Yao et al. proposed UMining [6] algorithm to find almost all the high utility itemsets from an original database. But it suffers to capture a complete set of high utility itemsets. Another algorithm named is Two-Phase; it able to find high utility itemsets. The Two-Phase algorithm is used to prune down the number of candidates and obtain the complete set of high utility itemsets. In the first phase, transaction-weighted downward closure property of search space is used to expedite the identification of candidates. In the second phase, one extra database scan is performed to identify high utility itemsets. However, this algorithm cannot deal with negative item values in utility mining. Yin et al [3] proposed USPAN algorithm to mine high utility sequential patterns. The authors introduced the lexicographic quantitative sequence tree to extract the complete set of high utility sequences. They also introduced two effective pruning strategies. Experimental results on both synthetic and real datasets show that USPAN works efficiently, even in a large scale data. Zihayat et al. [4] proposed an algorithm to retrieve high utility sequence in data streams. The authors proposed two efficient data structures namely ItemUtilLists (Item Utility Lists) and HUSP-Tree (High Utility Sequential Pattern Tree). In addition, a novel utility model called Sequence- Suffix Utility is proposed for effectively pruning the search space in HUSP mining. The authors named the algorithm as HUSP-Miner (High Utility Sequential Pattern 42 IJREAMV4I DOI : / , IJREAM All Rights Reserved.

2 International Journal for Research in Engineering Application & Management (IJREAM) ISSN : Vol-4, Issue-3, June 218 Miner) to find HUSPs in static databases efficiently. Then, a one-pass algorithm namely HUSP-Stream (High Utility Sequential Pattern mining over Data Streams) was proposed to incrementally update ItemUtilLists and HUSP-Tree online and find HUSPs over data streams. Experimental results on both real and synthetic datasets show that HUSP-Miner performs well. Dave [5] proposed an algorithm to retrieve high utility sequences in from Incremental Sequential Dataset. In the process of mining HUS, when new sequences are added into the existing database the whole procedure of mining HUS starts from the scratch, in spite of mining HUS only from incremental sequences. This result in excess of time as well as needs more efforts. So the authors proposed an incremental algorithm to mine high utility sequences from the Incremental Database. Lan et al. [7] proposed a projection-based average-utility itemset mining (PAI) algorithm to reveal HAUIs using a level-wise approach. The proposed upper-bound model reduces number of unpromising candidate items. Lin et al [8]. first developed the HAUP-tree structure and the HAUP-growth algorithm for mining HAUIs. In the HAUP-tree, each node at the end of a path stores the average-utility upper bound of the corresponding item as well as the quantities of the preceding items in the same path. This approach can thus be used to speed up the discovery of HAUIs [1]. Sivamathi et.al [16] compared existing utility mining algorithms and found the best one. The authors used execution time and memory space occupied by the algorithms as performance measures. Based on these performance metrics, experimentation was carried out. III. PROPOSED WORK: RHUIET This section illustrates the proposed RHUIET (Rare High Utility Itemset using Enumeration Tree) algorithm to retrieve rare high utility itemsets. Utility Mining algorithms generate candidates for itemset generation. Most of the algorithms use any standard data structure to sore and generate items. The proposed algorithm adopted enumeration tree structure to generate candidate itemsets. From the literature, it was found that no existing algorithm uses enumeration tree structure. The algorithm works as follows: Initially transaction weighted utility (TWU) and occurrence of each item (Support) is calculated [5] [6] [7]. Now select the items, whose TWU greater than minimum utility threshold [8] [9] [1] and Support less than minimum support threshold. These items are referred as promising items. Now construct enumeration tree with these promising items. Algorithm: High utility rare itemset mining. Input: Transaction database, Min. Support, Min. Utility. Output: High utility Rare itemsets. 1. Let D be a database. T1, T2, T3.Tn be a set of transactions.i1, I2, I3.In be the set of items in transactions. 2. Scan Database and calculate TWU and support of items.//transaction weighted Utility 2.1. TWU(I)= TU( Ti).//TU- Transaction Utility of item I. 2.2 support (I) = No. of occurrence of (I) in D. 3.HUI= {I.TWU > Min. Utility & I.Support < Min.Support} 4. Construct EnumTree(HUI); 5. Retrieve candidate itemsets from EnumTree(); 6. For each candidate item, calculate TWU, Support 7. HURI = {I.TWU > Min. Utility & I.Support < Min.Support} 8. Display HURI; Pseudo code of RHUIET An enumeration tree is constructed as follows: The root node of the tree is NULL. All the promising items are child nodes of root node. Now candidate items are generated by combining adjacent nodes. These combined nodes form as child of promising node. The process is repeated until no candidate generation is possible. Once candidate itemsets are found, utility and support of items are calculated. Finally retrieve rare high utility itemsets. This can be illustrated with the following example. IV. EXAMPLE The proposed work can be illustrated with the following example. Table 1 represents database. Here T1, T2 T3, T4, T5, T6 are Transaction ids. A,B,C,D,E,F are items in transactions. The number along with the item in parentheses represents the quantity of item purchased. In utility mining, this quantity is termed as internal utility of item. Table 2 shows the unit profit of items. This unit profit is termed as external utility of items. The product of internal and external utility gives utility item in a transaction. Similarly Transaction utility is the sum of utility of items in those transactions. Transaction weighted utility is the sum of transaction utilities in which the item present. More detailed definition about basics of utility mining can be found in [11][12][13][14][15]. Table 1. An Example transaction Trans. Id Transactions TU T1 (A,1) (B,3) (C,2) 19 T2 (A,2) (C,3) (D,1) 21 T3 (B,2) (C,1) 1 T4 (B,3) (C,1) (D,3) (E,2) 32 T5 (A,1) (B,3) (C,2) (D,1) (E,2) 28 T6 (F,1) 7 Initially TWU and support of items are calculated and is shown in table 3. Consider minimum utility threshold as 65 and minimum support as 4. From the example, the items A, D, E, F is termed as high utility rare items. Now an enumeration tree is constructed with these items for candidate generation. The root node is NULL. The items A, D, E, F are child of root node. Then the items in next level are combined to form candidate items. Hence, AD, AE, AF are children of A. Similarly DE and DF are children of D, EF are child of E and F has no child. Now 43 IJREAMV4I DOI : / , IJREAM All Rights Reserved.

3 International Journal for Research in Engineering Application & Management (IJREAM) ISSN : Vol-4, Issue-3, June 218 utility and support of these candidate itemsets are calculated. Then, it retrieves high utility rare itemsets, with support less than minimum support and utility greater than minimum utility. Table2. Profit table Item Unit profit A 2 B 3 C 4 D 5 E 2 F 7 able3.twu, support Item TWU Support A 68 3 B 89 4 C 11 5 D 81 3 E 6 2 F 7 1 Figure 1. Enumeration Tree V. EXPERIMENTAL RESULTS The algorithm is implemented in Java. The software tool used is NetBeans IDE 8.. The datasets used in the experiment are Chess and Foodmart. The experiment is conducted with various minimum support threshold and minimum utility threshold. At each different support utility threshold combination, execution time, memory space occupied and number of items retrieved was analyzed. Table 1 shows the execution time, memory space and number of itemset retrieved with different support and utility values in chess dataset and table 2 shows the performance metrics of the same in Foodmart dataset. Figure 2 and figure 3 represents graphical representation of the same. From the figure it was found that, the execution time and number of items retrieved were decreased, with increase in utility threshold value. Also it was found that, the memory space also decreases with increase in utility threshold values. Table 4. Performance of the RHUIET algorithm with different support and utility values in Chess dataset. Min_support Min_utility Exe. Time in ms Memory space in MB No. of itemsets retrieved IJREAMV4I DOI : / , IJREAM All Rights Reserved.

4 No. of items retrieved Exe. time in MS Mem. space in MB International Journal for Research in Engineering Application & Management (IJREAM) ISSN : Vol-4, Issue-3, June 218 Execution time in Chess dataset Memory space in Chess dataset Exe. Time in ms Memory space in MB Min. sup=2 Min. sup=3 Min.sup=4 Min.sup=5 Utility/ Support Threshold Min. sup=2 Min. sup=3 Min.sup=4 Min.sup=5 Utility/ Support Threshold No. of items retrieved in Chess dataset No. of itemsets retrieved Min. sup=2 Min. sup=3 Min.sup=4 Min.sup=5 Utility/ Support Threshold Figure 2. Performance analysis in Chess dataset. Table 5. Performance of the RHUIET algorithm with different support and utility values in Foodmart dataset. Min_support 2 Min_utility Exe. Time in ms Memory space in MB No. of itemsets retrieved IJREAMV4I DOI : / , IJREAM All Rights Reserved.

5 No. of items Exe. time in MS Mem. space in MB International Journal for Research in Engineering Application & Management (IJREAM) ISSN : Vol-4, Issue-3, June 218 Execution time in Foodmart dataset Memory space in Foodmart dataset Exe. Time in ms Memory space in MB Min. sup=2 Min. sup=3 Min. sup=4 Min. sup=5 Utility/Support threshold Min. sup=2 Min. sup=3 Min. sup=4 Min. sup=5 Utility/Support threshold No. of items retrieved in Foodmart dataset No. of itemsets retrieved Min. sup=2 Min. sup=3 Min. sup=4 Min. sup=5 Utility/Support threshold Figure 3. Performance of the algorithm in Foodmart dataset. VI. CONCLUSION In recent years utility based rare itemset mining has emerged as an interesting research area. This is because utility based rare itemset mining considers semantic significance of the items and its rare occurrence. In this paper an algorithm RHUIET was presented to retrieve rare high utility itemsets. Experimental results were conducted using Chess and Foodmart datasets. Execution time, Memory space and number of itemsets retrieved are analyzed with different support and utility values. REFERENCES [1] Vincent S. Tseng, Bai-En Shie, Cheng-Wei Wu, and Philip S. Yu, Fellow, Efficient Algorithms for Mining High Utility Itemsets from Transactional Databases, IEEE Transaction on knowledge. e and data engineering, vol. 25, no. 8, Aug 213. [2] C.F. Ahmed, S.K. Tanbeer, B.-S. Jeong, and Y.-K. Lee, Efficient Tree Structures for High Utility Pattern Mining in Incremental Databases, IEEE Trans. Knowledge and Data Eng., vol. 21, no. 12, pp , Dec. 29 [3] Junfu Yin, Zhigang Zheng & Longbing Cao, USpan: An Efficient Algorithm for Mining High Utility Sequential Patterns, in proceeding of KDD 12, August 12 16, 212, [4] Morteza Zihayat, Cheng-Wei WU, Aijun AN and Vincent S. Tseng, Efficiently Mining High Utility Sequential Patterns in Static and Streaming Data. [5] J Han, J.Pei, Y.Yin,R. Mao Mining frequent Patterns without candidate generation:a frequent -pattern tree approach, Data Mining and Knowledge Discovery 8(1)(24) [6] Liu. Y, Liao. W,A. Choudhary, A fast high utility itemsets mining algorithm, in: Proceedings of the Utility-Based Data Mining Workshp, August 25 [7] Y.-C. Li, J.-S. Yeh, and C.-C. Chang, Isolated Items Discarding Strategy for Discovering High Utility Itemsets, Data and Knowledge Eng., vol. 64, no. 1, Jan. 28. [8] C.H. Cai, A.W.C. Fu, C.H. Cheng, and W.W. Kwong, Mining Association Rules with Weighted Items, Proc. Int l Database Eng. and Applications Symp. (IDEAS 98), [9] R. Chan, Q. Yang, and Y. Shen, Mining High Utility Itemsets, Proc. IEEE Third Int l Conf. Data Mining, pp , Nov. 23. [1] 1 V.S. Tseng, C.-W. Wu, B.-E. Shie, and P.S. Yu, UP-Growth: An Efficient Algorithm for High Utility Itemsets Mining, Proc. 16th ACM SIGKDD Conf. Knowledge Discovery and Data Mining (KDD 1), IJREAMV4I DOI : / , IJREAM All Rights Reserved.

6 [11] H. Yao, H.J. Hamilton, and L. Geng, A Unified Framework for Utility-Based Measures for Mining Itemsets, Proc. ACM SIGKDD Second Workshop Utility-Based Data Mining, Aug. 26. [12] Jiawei Han, Hong Cheng, Dong Xin and Xifeng Yan, Frequent pattern mining: current status and future directions, Data Mining Knowledge Discovery, January 27. [13] M. J. Zaki, SPADE: An Efficient Algorithm for Mining Frequent Sequences, Machine Learning, 21,vol. 42, pp [14] 14. S. J. Yen and Y. S. Lee. 27. Mining high utility quantitative association rules. In DaWaK 27, LNCS 4654, pp [15] Z. H. Deng and Z. H. Wang. 21. A New Fast Vertical Method for Mining Frequent Itemsets. International Jour-nal of Computational Intelligence Systems, 3(6): [16] Sivamathi, C. and Vijayarani, S. (216, August). Performance analysis of utility mining algorithms. In Inventive Computation Technologies (ICICT), International Conference on (Vol. 3, pp. 1-4). IEEE. International Journal for Research in Engineering Application & Management (IJREAM) ISSN : Vol-4, Issue-3, June IJREAMV4I DOI : / , IJREAM All Rights Reserved.