Mean Harris Hawks Optimization (MHHO) Based Feature Selection and FFNN-LBAAA for Semen Quality Predictive Model

Assess fertility potential of a male partner plays a significant role in semen analysis. Changes in people lifestyles, such as alcohol consumption, smoking, eating habits, etc., are main causes. According to recent scientific investigations, environmental and lifestyle factors have a considerable negative impact on men seminal quality. One crucial part in influencing the possibility of semen for incidence of pregnancy is a Machine Learning (ML) algorithm. Previous studies used unbalanced datasets with performance results that tended to favour the majority class and local training methods were easily prone to local minima. In this paper, start with dataset collection, and pre-processing the dataset by normalization. Then, Synthetic Minority Oversampling Technique (SMOTE) data balancing approach has been developed for balancing normal and aberrant occurrences. Afterwards subset of features is chosen using Mean Harris Hawks Optimization (MHHO). The next step is to feed features into a Feed-Forward Neural Network and a Learning-Based Artificial Algae Algorithm (FFNN-LBAAA) classifier. University of California, Irvine (UCI) provided semen quality prediction dataset. Performance evaluation metrics are precision, sensitivity, specificity, f-measure, accuracy. The proposed system is contrasted with various systems are Multi-Layer Perceptron (MLP), Back-Propagation Neural Network (BPNN), Artificial Neural Network with Sperm Whale Optimization Algorithm (ANN-SWA).