Enhanced Fuzzy Logic Pre-Processing Technique Using Hybridized Bat and Particle Swarm Optimization Algorithm for Feature Selection.

In the modern era, the number of diseases is increasing and it is very crucial to diagnose them. One among them is the hepatitis virus in which the proper identification of symptoms is tedious. Machine learning algorithms are essential in recognizing the disease in the earlier stages. In this paper, a novel pre-processing technique and feature selection (FS) method are proposed to achieve improved accuracy with less time complexity. The proposed methodology is divided into three phases: In the first phase, the pre-processing technique such as Multiple Imputation with weight based fuzzy logic (MI-WFL) is applied to decrease the classification error. In the second phase, the correlation-based hybridized bat and particle swarm optimization (CHBPSO) FS algorithm is used to minimize the time complexity and maximize the accuracy rate. In the third phase, the poly kernel support vector machine (PK-SVM) classifier is implemented to avoid overfitting issues. According to the results of the experiments, the CHBPSO algorithm outperforms in the accuracy rate of 96.15%, and classification error for the following metrics mean absolute error (MAE) is 0.17, root mean squared error (RMSE) is 0.23, relative absolute error (RAE) is 49.66, and root relative squared error (RRSE) is 54.62 are lower than the existing algorithms. Furthermore, the proposed solution has attained less processing time of 1.6 sec due to the reduced optimal features. [ABSTRACT FROM AUTHOR]