Thesis:

Presenting a method to diagnose and optimize diabetes by using data mining methods and Tabu search algorithm

This study aims to propose a novel method for the diagnosis and optimization of diabetes by leveraging data mining techniques and the tabu search algorithm. Diabetes, as one of the most common chronic diseases worldwide, holds significant importance in public health; timely and accurate diagnosis plays a crucial role in preventing complications and reducing treatment costs. Therefore, developing efficient and optimized predictive models to identify diabetic patients is essential. The research methodology is based on real data from patients visiting healthcare centers affiliated with the Sabzevar University of Medical Sciences during 2023–2024, encompassing a wide range of clinical, behavioral, and biological features (62 features in total). Data quality was enhanced through preprocessing techniques and clustering based on an improved decision tree. Ultimately, the tabu search algorithm, as a metaheuristic optimization method, was employed to tune the parameters (C and gamma) of the support vector machine model to find the best parameters and establish an optimal decision boundary for distinguishing diabetic from non-diabetic patients. Evaluations showed that the proposed model achieved an accuracy of 85.61%, recall of 73.24%, precision of 82.17%, and an F1-score of 84.31%, demonstrating a reasonable balance between precision and recall and significant improvement in identifying positive cases compared to several reputable references. However, it performed weaker relative to other metaheuristic methods reported in the literature.

Keywords: Diabetes, Data Mining, Improved Decision Tree, Support Vector Machine, Tabu Search Algorithm