Model for career pathway prediction for Kenya’s competency-based education curriculum

Abstract

Kenya’s Competency-Based Education (CBE) curriculum puts an emphasis on learning based on project work, co-curricular activities involvement and continuous evaluation of the learner to influence learners into the most suitable career pathway. However, the choice of career pathway of learners in junior secondary schools in Kenya greatly relies on the manual judgment of the teacher and parent, which may be subjective, lack consistency and inconsiderate of the learning context. Therefore, there needs a consistent data-driven tool that combines both academic and non-academic pointers, based on actual school data, to advise on career pathway selection and that utilizes Machine Learning to inform the choice of the learner, teacher and parent. The main objective of the paper was to develop a Machine Learning model that can help select the most suitable career pathway for the learner under the Competency-Based Education curriculum. In specific the study aimed to: review existing literature to recognize the factors considered in the determination of the career pathway choice of the student; to analyze data to determine the most consequential factors that influence the career pathway choice of the student; develop a model to predict the career pathway choice using Random Forest algorithm; and to assess and validate the model’s ability to predict the individual’s career pathway.  The study used a quantitative experimental research design, with data of 1,083 students gathered from databases in junior secondary schools in the county of Trans-Nzoia, involving both public and private schools. The data included the academic scores, project marks, extra-curricular participation, leadership involvement, gender and the school type, functioning as the mediating variable The data was cleaned, standardized, split into 80:20 training-testing set. Random Forest feature-importance recognized 7 most influential features as Integrated Science (29.89%), Agriculture (16.60%), Pre-Technical Studies (10.47%), Mathematics (8.63%), Sports (7.38%), Music and Drama (5.12%) and the type of school (4.84%). The features together accounted for 82.92% of all feature importance. The dataset was trained using three Machine Learning algorithms. Random Forest was the best performing model with an accuracy of 89%, compared to Decision Trees with 81.46% and 78% for Neural Networks. However, Decision Trees and Neural Networks had a better stability on ten-fold cross-validation, where Random Forest had an accuracy score of 86%, 82% for Decision Tree and 78.62% for Neural Network. The Random Forest model recorded AUC score of 0.90, 0.97 and 0.98 for Arts, Social Science and STEM in that order, which illustrates very good discriminative ability. The study in conclusion states that academic performance, extra-curricular participation and school environment are collectively useful in forecasting career pathway choice of students. The final result is a dependable evidence-driven decision-support system – the Random Forest career pathway prediction model, which enhances objectivity and equality in selection of a clear pathway for the student under the CBE curriculum. The tool serves as a complimentary system for career pathway selection and not as a replacement of professional career guidance. The paper suggests future studies to be done in other counties and inclusion of the interests of the student, parents and other primary statistics considered.