Various biological diseases like heart diseases, breast cancer etc. are exceedingly increasing and becoming a major public health crisis and due to these millions of deaths happen globally each year. Earlier study revealed the conventional methods which solely depend on the patient’s case study are insufficient to correctly diagnose any particular disease. Manually analyzing various attributes of different diseases are challenging and long medical data are required which accurately determines the proper attribute selection which justifies the selection of the correct learning model. In this context the importance of evolutionary approach plays a crucial role about feature selection to train a model. Therefore, in this work the main focus is to train machine learning (ML) technique that has been proved valuable for classification task. In ML research, more suitable frameworks are required for better classification to solve such kind of problems. More specifically, in this paper, at the first phase, ensemble filter feature selection technique has been used to generate feature subset and in the second phase, enhanced PSO with adaptive inertia weight strategy has been applied for more accurate feature subset generation and finally it is validated using by K-Nearest Neighbor (KNN) learning models. Two datasets of biological diseases have been tested and results show a promising outcome for disease prediction.