Diabetic Prediction Model Based on BOX-COX Transformation and Optimized Autoencoder with SVM Classification

One of the most common illness in the rest of the world and country is diabetes, which causes many fatalities if treated late. Therefore, it is to create a system that uses medical data for promptly as well as accurately diagnose diabetic patients. In this designed model, the diabetic patient’s data are collected and gathered from the dataset. These data sets are pre-processed using MICE imputation and Box-cox transformation for missing value replacement and normalizing the data. The pre-processed data are reduced or selected using the optimized auto encoder algorithm for minimizing the complexity of the prediction model. Northern goashawk optimization is used for optimally selecting the parameters of auto encoder algorithm. Finally, the reduced features are trained using SVM classification for predicting diabetics. Accuracy, precision, recall, error, and specificity performance parameters are achieved for this developed model at 96.04%, 96.02%, 95.69%, 3.96%, and 95.21%, respectively. Comparisons are made between the evaluated values and existing approaches, such as KNN, NB and RF. Thus, the diabetic prediction model based on box-cox transformation and optimized auto encoder with SVM classification performs better than the existing model.