Ridge Regressive Quadratic Multivalued Feature Matching Pursuit for Skill-based Employability Identification in Higher Education

Abstract

Skill-based employability identification involves evaluating student’s skills and determining their suitability for a specific job or industry. Data mining techniques have been developed for predicting student employability based on certain skills. Skills identification is a crucial step for students in understanding employability. However, accurate and time-efficient prediction of student employability has become a pivotal focus for educational institutions. This paper introduces a novel approach using data mining techniques called Ridge Regressive Quadratic Multivalued Projection Matching Pursuit (RRQMPMP) to identify skill-based employability for students in higher education with better accuracy and minimum time consumption. The proposed RRQMPMP technique includes two major processes namely preprocessing and feature selection. First, the number of features and student data are collected from the dataset. Then the preprocessing steps are executed, including three processes namely missing data handling, duplicate data removal, and normalization to clean the input dataset. The Ridge Regressive Imputation Method is employed to handle missing data in the dataset. Subsequently, duplicate and non-duplicate data points are distinguished from the dataset using a Simple Matching Distance Measure. Finally, Quadratic Mean Feature Scaling is developed for the normalization process. With the preprocessed dataset, the feature selection step is performed by applying a Russell-Rao Index Multivalued Projection Matching Pursuit. Based on the Russell-Rao Similarity Index value, pertinent and impertinent features are identified. Finally, pertinent features are selected for skill-based student employability prediction to achieve higher accuracy and minimize time consumption as well as space complexity. An experimental evaluation is carried out with respect to accuracy, error rate, time complexity, and space complexity for different numbers of student data. The quantitatively analyzed results indicate that the performance of the proposed RRQMPMP technique increases the accuracy of skill-based student employability prediction with minimum time and space complexity compared to conventional methods.