Development of Efficient Data Mining Techniques for Cancer Genomic Patterns Classification and Prediction

Abstract

Research Scholars are concentrating Microarray Technologies and its applications such as various Patterns of Classifications. It is facilitating Research Scholars to focus different Cancers Patterns for analysis. This is one of the major applications of Bioinformatics. Recently proposed Classifiers, that used for predicting various Cancer Patterns were identified for analysis. Those identified classifiers are i. Multi-Objective Particle Swarm Optimization (MPSO), ii. Kernelized Fuzzy Rough Set Based SemiSupervised Support Vector Machine (KFRS-S3VM) and iii. Hybrid Ant Bee Algorithm (HABA). The identified classifiers were implemented and studied thoroughly with various Patterns of Cancers in regard to Accuracy, Execution Time, FScore, Memory Utilization, Sensitivity, and Specificity. The result demonstrated that the performances of the above specified Models were relied on the Gene Patterns. It was also noted that the Multiobjective Particle Swarm Optimization (MPSO) is relatively outperforming other two classifiers. This research work noticed that the position and parameter values of MPSO, particularly needed to determine optimized center values to achieve higher classification accuracy. Thus, this work proposed an Enhanced Multi-Objective Pswarm Based Classifier (EMOPS) by computing optimized center values named Global Best Position gbest. The proposed Classifier EMOPS was implemented, and this work revealed from its results obtained from simulation that the introduced Classifier newlinexix newlineperforms better than that of the above mentioned three classifiers. For improving the efficiency of the system beyond exiting level in regard to achieve better Time Complexity, the introduced model, Enhanced MultiObjective Pswarm (EMOPS) is implemented under the Parallel Framework with Even numbers of Parallel Processors 9say 2,4,8,16) and evaluated and confirmed from the results that the Time Complexity was relatively decreasing under the execution of Parallel Framework. To improve the classification accuracy of EMOPS further by considering Optimization quality and convergence speed, this work developed a Smart Multi-Objective Particle Swarm Optimizer (SMOPSO) and it is noticed that this method used for updating Particles based on Pairwise Competitions instead of updating Global best particles. This model is simulated and analyzed thoroughly. The simulated work outperforms the EMOPS classifiers in regard to Diversity and Convergence on ZDT 1 and ZDT 3, Execution Time (Processing Time), and Classification Accuracy. To improve the Classification Accuracy further, An Enhanced CancerAssociation based Gene Selection Technique (ECAGS) was proposed by identifying, prioritizing and selecting genes through gene-association based ranking. From our experimental results of the proposed ECAGS, it was noticed that the ECAGS outperforms our previous model EMOPS in regard to Accuracy, Execution Time, FScore, Memory Utilization, Specificity and Sensitivity. newline newline