A Multilayer Prediction Approach for the Student Cognitive Skills Measurement

Every year, a large volume of information about students' performance is processed in schools, colleges, and higher studies institutes. This information statistically associates students' performance with their study schedule and family-related characteristics. Recent methods have significantly contributed to student's cognitive skills (CSs) prediction area of research, but they are insufficient to address the challenges created by Study-Related Characteristics (SRC) of a student. Therefore, in the current attempt, we present a multilayer CS measurement method that uses SRC for student's skills prediction. The contributions of the proposed method are threefold. First, during quantization, a multilayer model is initiated by splitting SRC into five factors, and a specific range is assigned to each factor (timing schedules of studying, outing, traveling to school, and free timing as well as parent's relationships). Second, the range of CS (0-20) is divided into 21 periodic intervals (with a period of 1). The component-wise division of SRC and CS is to ensure prediction accuracy that makes the method more testable and maintainable. Third, it simulates the nonlinear relationship between CS intervals and SRC layers using Gauss-Newton method. Finally, we achieved six mathematical models for the SRC. During the experiment, the proposed method is tested on the students' performance data sets. The results reveal that the current approach outperformed the existing CS measurement techniques because we achieved a significant precision (0.979), recall (0.912), F1 score (0.9249), and accuracy measure (0.937) values. In the end, this paper is concluded by comparing the proposed method with competitive student's skills prediction approaches.