CHAPTER 5 ANT-FUZZY META HEURISTIC GENETIC SENSOR NETWORK SYSTEM FOR MULTI - SINK AGGREGATED DATA TRANSMISSION 5.1 INTRODUCTION Generally, deployment of Wireless Sensor Network (WSN) is based on a many to - one communication models, where a single sink node gathers data from a number of data resources. In recent times, developments with multiple sinks are gradually emerging, However, they struggle with a few demerits especially in the successful delivery of data to target region. The outcome of multiple communications formulates the existing works proposed for single-sink model useless. Hence data aggregation process will support the transmission of data to multiple sink nodes in an efficient manner. Wireless sensor network with the hierarchical organization of sensors aggregate the tasks into groups. The sensor nodes broadcast the aggregated data directly to the distant base station. The existing Mixed Integer Programming (MIP) formulation obtain the optimal solutions for multi-action processes using the genetic algorithm. However, MIP only deals with a resource scheduling problem in the field of Business Process execution plan and faces certain mean absolute error with lower transmission rate. MIP is not effectual in emerging hybrid genetic algorithms with the Tabu search meta-heuristics ant colony optimization. Another existing work 130
developed for security purpose termed as Dynamic secure end-to-end Data Aggregation with Privacy function (DyDAP) decreases the network load. DyDAP secures the data aggregation, but topological configuration is inappropriate when packet needs are to be transferred to multiple sink nodes. In order to develop the hybrid genetic algorithm on ant-fuzzy system, Hybrid (i.e.) ant-fuzzy Meta-heuristic Genetic method (HMG) based on the Tabu search is proposed in this chapter. Ant-fuzzy Meta heuristic Genetic method carries out the classification process on the aggregated data. The classification based on the genetic method uses the Tabu search - based mathematical operation to attain the feasible solution on multiple sinks. Initially, Ant-fuzzy Meta-heuristic Genetic method classifies the data record based on the weighted meta-heuristic distance. Genetic method combines the ant and fuzzy rule to optimize the classification capability with the weighted meta-heuristic distance. The classified records perform the Tabu search operation to transmit the aggregated data to the multiple sink nodes. HMG method achieves approximately 19 % improved transmitted message rate. Experiment is conducted in the network simulator on the factor such as transmission message rate, classification time, buffer level and the mean absolute relative error rate. 5.2 MULTIPLE SINKS IN WIRELESS SENSOR NETWORKS A single sink node gathers ecological data from a huge range of sensing devices. Hence communication routing is devised in order to offer secured and successful transmission to a single sink node. But most of the latest establishments gradually provide a better framework where the sensed data must be transmitted to 131
multiple sinks. Similarly, the framework definitely needs the same WSN to serve multiple applications, each operating on different resources. Moreover, the requirement for multiple sinks node occurs in the other cases. A sample multi - sink node is depicted in Figure 5.1. Figure 5.1 A Model representing multiple sink nodes The multiple sink nodes delivery as shown in Figure 5.1 involves two main sources or origins indicated in dark circles. The three different arrow marks denote the various path to the destination end or multiple sink nodes. Multiple sinks are increasingly and essentially needed to run latent applications and programming objectives. In general, data aggregation supports multiple sink node delivery as per the recent research reviews. 5.2.1 Issues related to Single Sink Based Data Aggregation Major difficulty faced in single sink - based data aggregation is elaborated below in detail, 132
i. In sensor data collection, multiple communication faces traffic overhead resulting in imbalance as well as high energy utilization in the complete network. Moreover, it leads to prior extinction of the network lifetime. Hence the open dispute is to prolong the network lifetime by minimizing the energy utilization and at the same time preserving energy efficiency. ii. Scalability trouble is vast in the single sink network framework. In single sink framework, the data aggregated at the sink may lead to more than its communication power due to the huge range of sensors usage in the network. Additionally, the radio channel capacity nearer to the sink may turn into excess if the typical number of hops between the source and the sink becomes overload. iii. In single sink network framework, if the sink is heavily loaded, then the data will not arrive at target region causing transmission failure. 5.2.2 Multiple Sink - based Data Aggregation for WSN A standard operation of the sensor network is data collection. Here the modeled data at each of the sensor nodes need to be delivered for the purpose of advanced processing and analysis for the sink node. In WSN, data aggregation is expected as the most important shared data, processing feature which consumes 133
energy and minimizes routing conflict. For wireless access connection in sensor networks, data aggregation is taken as a significant model. Data aggregation performs the process of data combination from different sources along the route discarding the repetitions as well as minimizing the transmission rate, thus consuming energy. Therefore the concentration is made on achieving energy consumption on efficient use of data aggregation function with successful delivery of data to multiple sink nodes. The information received from the sources is aggregated to reduce the number of transmission consuming energy. Due to limited resource utilization in sensor network, number of transmissions needs to be reduced in order to enhance the bandwidth utilization as well as to prolong network lifetime. WSN based on data aggregation protocol should be built with secure and energy efficient basis. The work in [79] investigates the association between energy security and data aggregation process in WSN. Classification of secure data aggregation protocols is made based on network topology and security. However, the research addressed many problems of data aggregation especially from the energy point of view. The scalability of sensor networks rises with the presence of the multiple sinks. By a simulated connection, all sink nodes are linked to the simulated sink node, in multiple sink networks which consecutively vary to a single destination turn around multiple destinations. Efficient data collection task is performed by the nodes, and then the optimal sink is selected for delivering the data in multiple sink networks. The mean length between the source nodes and the sink will be reduced in 134
the multi sink network resulting in energy consumption and prolonged network lifetime. The sink node behaves as an access in multi sink network, transmitting the sensed data to the storage system in the network by means of internet. Only the data generated by a specific set of resources is gathered by the sink and,then the complete event observed is rebuilt at the data storage system. The work in [51] develops a methodology for adopting business process semantics and other paths in the Business Process Management structure in order to address the problems related to achieve accurate execution of processes. The approach employs mixed integer programming (MIP) formulation for a businessprocess execution plan and a meta-heuristic algorithm to achieve best solutions for multi-activity processes. The purpose is to get the best task that gives the most efficient performance under resource restriction. However, an optimization technique of generating is not openly applicable to Business Process, which allows for substitute paths. In addition, MIP is unable to handle mean absolute error increasing transmission rate. MIP is still in the progress of cresting a hybrid genetic algorithm with the Tabu search meta-heuristics ant colony optimization. In addition to energy efficiency issues, security also plays a vital role in sensor network. An approach for dynamic secure end-to-end data aggregation with privacy function, named DyDAP, presented in [52] tackles the security management issues like anonymity and data integrity. Moreover, DyDAP incorporates a new aggregation 135
algorithm using a discrete-time control loop and is able to energetically manage innetwork data fusion to minimize the communication load. However, DyDAP needs extra attention to improve efficiency more specifically on considering multimedia application where the nodes exchange signals as data in sensor networks. The main drawback of DyDAP is that the system is not applicable to a set of topological configurations, i.e. containing multiple sink nodes. Keeping the above demerits in mind, the proposed work aims to address the issues related to energy efficiency. 5.3 HYBRID META-HEURISTIC GENETIC METHOD ON WIRELESS SENSOR NETWORK The main objective of the proposed work is to classify the aggregated data and perform the path searching process for effective transmission of data packets on the multiple sink nodes. The hybrid meta-heuristic genetic method involves two phases. The first phase performs the classification operation with the hybrid (i.e.) ant-fuzzy rule aggregated data information. The aggregated information is classified with the meta-heuristic genetic algorithm. After the classification, the second phase carries out the search process using the Tabu search for the effective transmission of the data packets on the sensor network. Architecture Diagram of the Hybrid Meta-heuristic Genetic (HMG) method is illustrated in Figure 5.2. 136
Sensor network with Aggregated Data Classification operation Hybrid Meta-Heuristic Genetic method Weighted metaheuristic distance Tabu Search Path Selection Efficiency Transmit data packet to multi sink node Figure 5.2 Architecture Diagram of HMG method As illustrated in Figure 5.2, HMG method shows overall process briefly through the diagram. The energy effective data aggregation is performed on the sensor network using the fuzzy ant colony optimized clustering rule. The aggregated data through the ant-fuzzy rule perform the classification operation using the Hybrid Meta-Heuristic Genetic system. The classification process based on weighted metaheuristic form helps to increase the classification accuracy rate. Search process in identification of classification data is done using Tabu search in HMG method just to improve the path selection efficiency. The selection of path helps to transmit the data packet to multi-sink nodes. 137
5.3.1 Weighted Meta-Heuristic Distance for Classification Hybrid Meta-Heuristic Genetic system uses the ant-fuzzy aggregated data rule to classify the data packets. The genetic function uses the ant-fuzzy weights to classify the aggregate data. The hybrid genetic concept in HMG supports appropriate classification based on meta-heuristic chromosome weight distance computation in wireless sensor network. Meta-heuristic in HMG method represents the high level procedure elaborating the task of classification, searching and transmission of data packets to the destination sink nodes. Chromosome Weighted meta-heuristic is derived as ( )= 1 ( ), 2 ( ),. ( ) (1) Ant-Fuzzy rule as derived in (1) is represented as ( ). 1 ( ), 2 ( ) ( ) are the fuzzy set rules of 1,2 up to n respectively where the classification relies on weighted procedure. The weighted procedure in HMG method is computed as, g i (x) 0 (i=1,2,3...n) (2) ( ) 0 ( = 1, 2, 3 ) (3) 138
( ) and ( ) denotes the classified part in HMG method using the metaheuristic chromosome weighted distance. The positive and negative values of chromosome weighted in HMG method vary based on the axes distance in sensor network. The meta-heuristic weighted chromosome distance concerns with the more generalized result using the ant-fuzzy rule. Chromosome Weighted Meta-heuristics is formularized as = ( ), ( ), = 0, =1 (4) where ( ), ( ) are the classified parts. If they are Equal to i, then the result will have accurate output on classification. Otherwise, the result is set to zero in Chromosome Weighted Meta-heuristics (CWM) form. Chromosome Weighted Meta-heuristics in HMG method makes a few assumptions about the classification process so that they are usable for a variety of multi - sink packet transmission. 5.3.2 Hybrid genetic Algorithm Chromosomes are vectors of real-valued weights from the hybrid (i.e.) antfuzzy rule. Each chromosome is a vector with the weight count of decimal numbers. Moreover, vector value of the chromosome is associated with each classification attribute in HMG method. The relation between the chromosome and classification 139
attribute supports a better way to increase the classification accuracy such that data is delivered perfectly. Hybrid genetic Algorithm - based classification is demonstrated in Figure 5.3 Network Environment Aggregated Data Hybrid GA Environment Initial population Parent Chromosome Fitness Evaluation Child Chromosome (offspring) Performs Classification Figure 5.3 Hybrid Genetic Algorithm - based Classification 140
Figure 5.3 describes the classification of aggregated data using the hybrid genetic algorithm. The initial population is measured in hybrid genetic method to easily forecast the classification time. The populated chromosome measures the weight using the fitness function. The fitness function is measured for each parent chromosome selection. The selected parent chromosome performs the weighted meta-heuristic measure to classify the child chromosome (i.e.) offspring in sensor network. The goal of the hybrid genetic algorithm is to have an attribute fitness vector from the ant-fuzzy weights and to improve the classification accuracy. These computations are used in a number of ways to evaluate the functions as explained in the hybrid genetic algorithmic steps. //Sensor Network Environment Step 1: Step 2: Begin with aggregated Data Initialize a population P of n data packets from sensor network Step 3: Step 4: Step 5: Step 6: For every population P Fitness function is evaluated on n data packets If data packet passes the weighted criteria test, Then the classification process is preceded, until the specified termination condition 141
Step 7: Step 8: Step 9: End if For each Parent Chromosome selection Child Chromosome (i.e.) offspring is classified based on parent attribute vector value Step 10: Step 11: Step 12: End For End For End The classified attribute is associated with initial population of chromosome to generate effective aggregated data. On repeated iteration, chromosomes are evaluated through each data aggregated group classifying each data based on its chromosome meta-heuristic weights in HMG method. The parent and offspring chromosomes are effectively selected using the hybrid genetic algorithm for the ant-fuzzy rule - based data packet classification. 5.3.3 Tabu Search in HMG method The classified Ant-fuzzy rule data aggregation in HMG method performs the searching process through Tabu mathematical operation. The Tabu search in HMG method refers to the set of classification rules to formulate the binary programming. The ant-fuzzy rule with the meta-heuristic genetic method aims to reduce the delay 142
time. The delay time is minimized in HMG method by judging the result of the individual users and formularized as =1 =1 =1 (5) = [, ] Tabu search time minimizes the delay count by computing (5) for each of the =1 =1 =1 users., [, ] are weight results of the meta-heuristics. The Tabu search reduces the memory structure in HMG method so that the delay strategic point is reduced. The classified records perform the Tabu search operation to transmit the aggregated data to the multiple sink nodes without any strategic delay. The metaheuristic weight guides to search the best result and explores solution space. Initially, Tabu search assumes i*=i and k=0, where k=0 at the root node. Tabu search in HMG method is described as, Step 1: Choose a data packet i from the population set P Step 2: K=k+1, when travelled through each node with the data packets Step 3: Binary programming choose best path from chromosome set Step 4: If ( )< ( ), then i* = i. Step 5: Update Tabu and goal conditions for transmit data packet to multi-sinks Step 6: User request satisfied, then stop, else go to Step 2 Step 7: End If 143
Tabu search in HMG method offers the bilinear path to transmit the ant-fuzzy rule - based data packets to the multiple sink nodes. The size of the data packets is measured in order to easily identify the transmission time using the HMG method. The program specifies the sink nodes and connected objective function. The first term denotes parent or root attribute values and the second term denotes the offspring attribute values. The Tabu search - based data packet transmission in HMG is demonstrated in Figure 5.4. Root Attribute Off Spring Attribute Figure 5.4 Multi Sink Data Packet Transmission Tabu search is used to transmit the ant-fuzzy data packets to the multiple sink nodes with the feasible solution. In the first stage of data packet transmission in HMG method, a quality solution is generated with the root path selection. After the root path selection, the offspring (i.e.) forwarding nodes are evaluated to improve the 144
diversity of result. Finally, a diversification strategy is used in HMG method to generate a new initial solution to reinitialize the procedure and perform data aggregation process for the other set of features. 5.4 HMG EXPERIMENTAL WORK EVALUATION Hybrid ant fuzzy Meta-heuristic Genetic method (HMG) based on Tabu search is experimented on ns-2 network simulator. The sensor network taken for the experiment purpose performs the classification process on the aggregated data. The network range taken for the experimental work is about 1000*1000 m. The Random Waypoint model is developed to randomly classify and move another sensed node location point. The RWM model relocates to a randomly chosen location to perform the effective transmission on the multiple sink nodes. HMG takes 25 milliseconds on each simulation, and averagely 80 sensor nodes are taken for the experimental evaluation. The chosen nodes randomly move with a selected velocity and speed. The minimum moving speed of the sensor node is about 4.0 m/s of each sensed node. The random movement of sensor node uses Dynamic Source Routing (DSR) Protocol to perform the experimental evaluation in the HMG method. The experiment is conducted on the parameters such as overall transmitted message rate, classification time, buffer level and mean absolute relative error. 145
5.5 RESULT ANALYSIS ON HMG MODEL HMG method performs experimental work and compares the result percentage on the existing Mixed Integer Programming (MIP) formulation in [51] and Dynamic secure end-to-end Data Aggregation with Privacy function (DyDAP) in [52]. The existing and the proposed HMG functional parameters are analyzed through table values and graph points. This algorithm is implemented in ns-2 network simulator 5.5.1 Transmitted Message rate Transmission message rate is defined as the degree of accuracy in transferring the data packets to multiple sink nodes in sensor network. Transmission message rate is measured in terms of meter per second (m/s). ( / ) = ( ) Message size and travelling distance and time vary on each request. Higher the transmission message rates, then better the throughput of the system. The values are taken from the trace files of ns-2 network simulator. With message size of 5120 KB, the overall transmitted message rate (m/s) using the existing methods MIP and DyDAP were measured to be 48.29 m/s and 50.22 m/s, whereas the proposed HMG recorded 56.88 m/s. 146
Example: Transmitted Message Rate =Distance Travelled/Time required to reach multiple sink node Time required to reach multiple sink node = 1.4064 milli sec Overall Transmitted Message Rate = 80/1.4064 Overall Transmitted Message Rate = 56.88 m/ s Table 5.1 Tabulation of Overall Transmitted Message Rate Message Size (KB) Overall Transmitted Message Rate (m/s) DyDAP HMG MIP method function method 5120 48.29 50.22 56.88 10240 90.75 101.45 113.77 15360 140.69 155.76 170.66 20480 189.17 200.65 227.55 25600 250.91 265.19 284.44 30720 280.43 300.78 341.33 35840 350.19 369.72 398.22 The experimental work is conducted for different message sizes with network throughput maintained at 90 as tabulated in Table 5.1. 147
Overall Transmitted Message Rate (m/s) 450 400 350 300 250 200 150 100 50 0 MIP method DyDAP function HMG method Message Size (KB) Figure 5.5 Measure of Overall Transmitted Message Rate Figure 5.5 describes the transmitted message rate of HMG method, MIP method [51] and DyDAP function [52]. In HMG, hybrid genetic concept is adopted to provide better transmission with the help of meta-heuristic chromosome weight distance computation in wireless sensor network. Meta-heuristic in HMG method improves the transmission rate by 13 25 % when compared with the MIP method [51]. Existing MIP [51] drops certain packets during transmission at the time of resource assignment by changing the currently assigned resource into one of its alternatives, which leads to lower transmission rate. HMG method transmits the data packets from 7 13 % improved when compared with the DyDAP function [52]. As the power supply is limited in DyDAP function [52], data to be transferred is stuck in-between paths transmission causing lower transmission message rate. 148
5.5.2 Classification Time Classification time is defined as the amount of time taken to classify the aggregated data based on the hybrid genetic algorithm. Classification time is evaluated as = Duration of start and end time gives the actual classification time period which actually represents the classification data accuracy. The values are taken from the trace files of ns-2 network simulator. In terms of classification time (sec), with respect to data record size (KB), MIP and DyDAP recorded 440 and 420 sec whereas using the proposed HMG the classification time was measured to be 400 sec. Example: Data Record Size = 50 KB Classification Time = End time for classification -Start time for classification Classification Time = 600 sec -200 sec = 400 sec Table 5.2 Tabulation of Classification Time Data Record Classification Time (sec) Size (KB) MIP method DyDAP function HMG method 50 440 420 400 100 585 528 500 150 910 880 850 200 1530 1495 1462 250 1925 1885 1724 300 2199 2176 2125 350 2452 2398 2256 149
Classification time measures are shown in Table 5.2 in comparison with the existing MIP method [51], DyDAP function [52] and the proposed HMG method. Classification Time (sec) 3000 2500 2000 1500 1000 500 MIP method DyDAP function HMG method 0 50 100 150 200 250 300 350 Data Record Size (KB) Figure 5.6 Performance of Classification Time Figure 5.6 presented the classification time based on the varied data record sizes in terms of kilo bytes (KB). In HMG method, ( ) and ( ) denote the classified parts using the meta-heuristic chromosome so that the classification rate is reduced up to 3 14 % when compared with the MIP method [51] as decoding is done with the designed chromosome procedure with higher time instead of classification. The positive and negative values of chromosome weight value provide easy way to classification with axes distance measure so that the HMG method takes 2 8 % less time to classify when compared with the DyDAP function [52], while 150
DyDAP function requires additional time in handling with node reputation information. 5.5.3 Buffer Level In general, the buffer is considered as storage space in need to load the data temporarily so the nodes buffer size is the most important aspect within the network. If the amount of the buffer is filled, then packet will be dropped causing congestion in the sensor node. Buffer level denotes the higher memory space provided by the system to store data. ( )= Buffered rate is defined as the region of memory storage measurement. The memory storage is maintained high for storage of more messages in HMG method, and it is measured in terms of MB. The values are taken from the trace files of ns-2 network simulator. The buffer level (KB) attained 3495 KB using the proposed HMG whereas the existing MIP and DyDAP recorded 3198 KB and 3070 KB with respect to time. Example: Data Size = 20 KB Buffer Level = Actual memory space available/ Total Data size to be stored = 600/0.1716732 Buffer Level =3495 KB 151
Table 5.3 Tabulation of Buffer Level Buffer Level (KB) DyDAP Data Size MIP method function HMG method 20 3198 3070 3495 40 6001 5989 6601 60 8895 8775 9798 80 12010 11980 12650 100 14980 14940 15700 120 18020 18000 18818 140 21030 20980 21970 Table 5.3 demonstrates the buffer level value of the HMG method, MIP method [51] and DyDAP function [52]. As per the tabulation result, the proposed HMG method offers a higher space for data storage temporarily. 25000 Buffer Level (KB) 20000 15000 10000 5000 0 20 40 60 80 100 120 140 Data Size MIP method DyDAP function HMG method Figure 5.7 Measure of Buffer Level 152
Figure 5.7 measures the buffer level rate or memory rate of the HMG method based on the simulation time. As the simulation time varies, the buffer level for storing the information is also strengthened (i.e.) expanded. The effective memory structure development in HMG method improves buffer level by 4 10 % when compared with the MIP method [51] and 4 13 % when compared with the DyDAP function [52]. Tabu search performs the effective transmission with effective utilization of the buffer level of the sensor network. The crossover operation in MIP [51] requires additional memory to store the combined parent, and offspring chromosomes result in ineffective buffer level. At the same time, the buffer needs to store the result of data anonymity mechanisms based on data cloaking and privacy policy values resulting in higher requirement of buffer memory space. 5.5.4 Mean Absolute Error rate Mean Absolute Error rate is the number of imperfectly established data packets from the total number of received data packets at the different time intervals. =.. Mean Absolute Error rate is measured in terms of the percentage, and the expected value of error rate is measured in decimal points.the values are taken from the trace files of ns-2 network simulator. The mean absolute error rate (%) was 153
observed to be 0.0933 % and 0.0794 % using MIP and DyDAP when compared to the proposed HMG which recorded 0.0748 %. Example: No of Nodes =10 Mean Absolute Error rate = No. of incorrectly received data packets/ Total no. of data packets received = (6/80.2)*100 % Mean Absolute Error rate = 0.0748*100 % Table 5.4 Tabulation of Mean Absolute Error rate No. of nodes Mean Absolute Error rate *100(%) MIP method DyDAP function HMG method 10 0.0933 0.0794 0.0748 20 0.0815 0.0654 0.0623 30 0.0745 0.0606 0.0588 40 0.0689 0.0548 0.0504 50 0.0532 0.0462 0.0414 60 0.0468 0.0378 0.0347 70 0.0342 0.0295 0.0269 Table 5.4 depicts the tabulation for error rate found at the receiver side on prediction of faulty data packet. 154
Mean Absolute Error Rate (%) 0.1 0.08 0.06 0.04 0.02 0 10 20 30 40 50 60 70 MIP method DyDAP function HMG method No.of nodes Figure 5.8 Measure of Mean Absolute Error rate Figure 5.8 depicts the mean absolute error rate based on the node count of the sensor network. The classified attribute in accordance with initial population of chromosome easily predicts the faults minimizing the relative error. Chromosomes are evaluated by iterating through each data aggregated group so that the mean absolute error rate is reduced up to 19 26 % when compared with the MIP method [51]. Whereas the objective function performs a merging operation to split activity in transmission, so error rate is high in MIP method [51]. In HMG method, the parent and offspring chromosomes are effectively selected using the hybrid genetic algorithm providing a better way to reduce error rate for about 2 10 % when compared with the DyDAP function [52]. Because, DyDAP [52] alters the real position of the nodes at data cloaking anonymity approaches, the chance of mean error rate is high. 155
Finally, Hybrid Meta-heuristic Genetic method facilitates the process of classification on using ant-fuzzy rule - based data aggregation set. Hybrid genetic algorithm is more appropriate even to a wide variety of real number to achieve the best result in the wireless sensor network. The classified data are transmitted with the aid of Tabu search - based data path selection and transmission of information. Moreover, HMG mode is able to provide a successful transmission rate with sufficient buffer level in a minimum classification time discarding the error rate at the receiver side. 5.6 SUMMARY The algorithm Hybrid (i.e.,) ant-fuzzy Meta-heuristic Genetic (HMG) is the best as it develops hybrid genetic algorithm on ant fuzzy system for multi sink data aggregated transmission based on the Tabu search. The Ant-fuzzy Meta heuristic Genetic method is the best as it carries out the classification process on the aggregated data using the Tabu search based mathematical operation to attain the feasible solution on multiple sinks improving the buffer level. Further, the Ant-fuzzy Meta-heuristic Genetic method classifies the data record based on the weighted metaheuristic distance. Genetic method combines the ant and fuzzy rule to optimize the classification capability with the weighted meta-heuristic distance and therefore reducing the classification time. The classified records perform the Tabu search operation to transmit the aggregated data to the multiple sink nodes improving the overall transmitted message rate and reducing the mean absolute error rate. The proposed HMG is comparatively better than the existing methods Mixed Integer 156
Programming (MIP) formulation and Dynamic secure end-to-end Data Aggregation with Privacy function (DyDAP), in terms of parameters namely, overall transmitted message rate, classification time, buffer level and mean absolute error rate. With message size of 5120 KB, the overall transmitted message rate (m/s) using the existing methods MIP and DyDAP were measured to be 48.29 m/s and 50.22 m/s, whereas the proposed HMG recorded 56.88 m/s. In terms of classification time (sec), with respect to data record size (KB), MIP and DyDAP recorded 440 and 420 sec whereas using the proposed HMG the classification time was measured to be 400 sec. The buffer level (KB) attained 3495 KB using the proposed HMG whereas the existing MIP and DyDAP recorded 3198 KB and 3070 KB with respect to time. Finally, the mean absolute error rate (%) was observed to be 0.0933 % and 0.0794 % using MIP and DyDAP when compared to the proposed HMG which recorded 0.0748 %. The Tabu search improves the path efficiency rate to 11.34 % in contrast to the other existing works. Hybrid genetic algorithm reduces the mean absolute error rate predicting the faulty packets received. In addition, HMG model achieves improved classification performance in a realistic time. HMG method is more useful with higher benefits to different contexts of sensor network for transmitting the data packet successfully to the multiple sink nodes. Network simulator justified the better performance of HMG model by offering higher memory space increasing the buffer level memory rate on balancing the memory with Tabu search. Tabu search employed in HMG method gives an effective transmission of the ant-fuzzy data packets in wireless sensor network. 157