A Genetic Neuro-Fuzzy System for Diagnosing Clinical Depression

Abstract

Depression is a serious illness that affects millions each year and if left untreated, it may lead to the deaths of many. It comes in many flavors that can be very different among people and this makes diagnosing it very difficult. A lot of artificial intelligence systems have been designed to diagnosing depression but they failed to perform feature selection and extraction on the dataset used in training the systems and this has a huge implication on the classification accuracy of the system. The objective of this research work is to develop a depression diagnosis system, that takes into consideration feature selection and extraction of dataset using Genetic-Neuro-Fuzzy techniques. Feature selection and extraction, will enable identification of key symptoms and hidden traits which are vital in diagnosis of depression. In this work, a Genetic Neuro-Fuzzy Model which is capable of handling feature selection and extraction on depression dataset was proposed and designed for diagnosing clinical depression. The GA component optimizes the clinical dataset which consist of series of diagnosed depression cases by performing feature selection and extraction, while ANFIS is used in training the optimized dataset obtained from the GA. The system had 92.5% prediction accuracy. This is a significant improvement over the best related model in literature that achieved a prediction accuracy of 92.4%. The system is recommended for psychiatrist hospital to aid in depression diagnosis. The research is limited to the diagnosis of clinical depression; future work should focus on the other forms of depression and treatments. The model has incorporated feature selection and feature extraction for the prediction of clinical depression with significant results established with performance indicators.