Diagnosis of Infertility in Women with Thalassemia Using a Hybrid Particle Swarm Optimization and Multilayer Perceptron Approach

One of the most common diseases in today’s world is thalassemia, and its prevalence is increasing globally every year. The use data mining techniques to develop predictive models for identifying individuals at risk of thalassemia is highly beneficial in reducing complications associated with the disease. These techniques utilize statistical methods and artificial intelligence to identify patterns and relationships among variables. In this study, a data mining approach based on the C۴.۵ decision tree, multilayer perceptron (MLP) neural network, and a hybrid model combining particle swarm optimization (PSO) with MLP was used to analyze criteria and predict infertility in thalassemia patients. A total of ۵۵ criteria were considered for thalassemia evaluation, which were used as input neurons in the neural network. The results indicate that the root mean square error (RMSE) for the C۴.۵ decision tree is ۰.۰۲۱۶, for the MLP neural network is ۰.۰۲۴۸, and for the hybrid PSO-MLP model is ۰.۰۲۷۶. The classification accuracy for the C۴.۵ decision tree, MLP neural network, and hybrid PSO-MLP model on test data is ۱۰۰%, ۹۶.۸۲۵۴%, and ۱۰۰%, respectively. It can be concluded that, for the given dataset and based on the clustering performed on the data, the C۴.۵ decision tree and the hybrid PSO-MLP model outperform the standalone MLP neural network in terms of predictive performance.

Keywords:

Thalassemia , Infertility , Decision Tree , Data Mining , Particle Swarm Optimization