Optimizing diabetes prediction with MLP neural networks and feature selection algorithm

In this research, the goal was to improve diabetes prediction by combining Multilayer Perceptron Neural Network (MLPNN) with Memetic Algorithm (MA) and Arithmetic Optimization Algorithm (AOA). The method suggested used a preprocessing step to choose a representative subset of attributes from the initial set. Next, the method suggested utilized a combination of the MA and AOA algorithms to optimize feature selection, resulting in a refined dataset that served as input for the Neural Network. Ultimately, the suggested approach utilized the multilayer perceptron neural network (MLPNN) to train the network with hidden layer neurons. The experimental findings indicated a 95% high accuracy rate was achieved. Machine learning classifiers achieved better accuracy compared to classifiers in previous studies, with Decision Tree and Logistic Regression classifiers each reaching 93.57% and 93.33% accuracy, respectively.