Quantitative transfer learning- based students sports interest prediction using deep spectral multi-perceptron neural network

Abstract

Sports performance predictions are essential in understanding student interest rates. Early indications of student progress facilitate athletic departments to improve their learning interests and make students perform better. Interests in sports involve understanding key physical factors that significantly impact students’ sports behavior and various other influencing factors. Deep learning techniques were used to develop a predictive model for student interest performance and support to identify the essential relationship influencing students’ sports behavior. Identifying sports interests is complex because student interests represent different features. Existing methods cannot predict the features and the relationship between their related attributes. Therefore, previous methods had low accuracy high time, and error rate performance. To resolve this problem, a deep learning (DL) based sports interest prediction model was proposed using a deep spectral multi-perceptron neural network (DSMPNN) to identify student sports interests. Initially, the preprocessing is carried out by Z-score normalization to verify the actual margins of student interest rate to make normalization by comparing the ideal and essential margins of student interest through behavioral feature analysis using student behavioral sports interest rate (SBSIR). According to the feature dimensionality reduction, the non-relational features are reduced using the spider foraging feature selection model (SFFM) to select the essential features. Then, a deep spectral multilayer perceptron neural network (DSMPNN) is applied to predict student interest by class sports interest. The classifier proves the prediction accuracy, precision, and recall rate of up to 96% high performance to analyze the interests of the sport. The suggested system also produces higher performance than the other system.