HFSA: hybrid feature selection approach to improve medical diagnostic system

Thanks to the presence of artificial intelligence methods, the diagnosis of patients can be done quickly and accurately. This article introduces a new diagnostic system (DS) that includes three main layers called the rejection layer (RL), selection layer (SL), and diagnostic layer (DL) to accurately diagnose cases suffering from various diseases. In RL, outliers can be removed using the genetic algorithm (GA). At the same time, the best features can be selected by using a new feature selection method called the hybrid feature selection approach (HFSA) in SL. In the next step, the filtered data is passed to the naive Bayes (NB) classifier in DL to give accurate diagnoses. In this work, the main contribution is represented in introducing HFSA as a new selection approach that is composed of two main stages; fast stage (FS) and accurate stage (AS). In FS, chi-square, as a filtering methodology, is applied to quickly select the best features while Hybrid Optimization Algorithm (HOA), as a wrapper methodology, is applied in AS to accurately select features. It is concluded that HFSA is better than other selection methods based on experimental results because HFSA can enable three different classifiers called NB, K-nearest neighbors (KNN), and artificial neural network (ANN) to provide the maximum accuracy, precision, and recall values and the minimum error value. Additionally, experimental results proved that DS, including GA as an outlier rejection method, HFSA as feature selection, and NB as diagnostic mode, outperformed other diagnosis models.