Male fertility is significantly influenced by the quality of the semen. The quality of semen has declined recently, and this unfavorable trend has caused significant concern. Moreover, lifestyle complexity poses a challenge in conducting the pertinent research needed for this aim (defined by elements such as frequent changes or the participation of multiple features and confounding variables). Previous research has shown that the accuracy of using machine learning (ML) to predict semen quality is lower due to irrelevant number of features presented in the dataset, and it may produces increased error rate. In this paper, proposed system majorly focuses on feature selection, and classification methods. Initially, dataset is collected using University of California, Irvine (UCI), and dataset is normalized by data normalization. Secondly, Synthetic Minority Oversampling Technique (SMOTE) is a method of oversampling. Thirdly from the balanced dataset, irrelevant, and redundant features are removed by Sparrow Search Algorithm (SSA). SSA is inspired by the sparrow population foraging behaviors for selecting most important and significant features from the dataset. Finally, Optimized Feed-Forward Neural Network (OFFNN) classifier is introduced for semen quality analysis. Various evaluation metrics like precision, sensitivity/recall, specificity, F-measure, and accuracy has been used to evaluate the results of classifiers.