Improving Medical Dataset Accuracy with PSO Feature Selection and Machine Learning Classification Techniques

Abstract

The redundancy in the dataset usually effects the performance of the model in the sence of accuracy, computational time, cost of storage and rilaiability of the data analysis. A cleaned and noise free data can help achieving satisfactory performance of the model. The one-vs-rest approach serves as the fitness function for Particle Swarm Optimisation (PSO) to resolve the classification problem. The global optimization in machine learning reduces or removes irrelevant redundant data to provide accuracy. A good method of feature selection includes investigation, sample classification to avoid incomprehensibility. In this paper swarm optimization is used to implement feature selection. The support vector machines and one verses rest method prove to be the fitness function of PSO classification problem. Moreover, In applications such as data mining, machine learning, medical data processing, and pattern classification, feature selection is vital. Furthermore, our test results show that PSO-based feature selection improves machine learning model performance on medical datasets accuracy of 98.13%, thereby enabling successful and accurate diagnosis and forecast