Development Of A Web-Based Machine Learning Application For Identifying Students’ Interests And Talents In Surabaya

Abstract

This study aims to develop a web-based machine learning application that can identify students’ interests and talents in Surabaya in a more accurate and systematic manner. A mixed-method approach was employed, combining quantitative and qualitative methods to obtain comprehensive results in the system development process. Data were collected from 100 high school students, covering academic performance, cognitive abilities, personality traits, and social skills. The data were then processed through preprocessing stages, including normalization, handling missing values, and outlier removal to ensure data quality for modeling. The dataset was split into 50% training data and 50% testing data to build and evaluate the model using the Random Forest algorithm, which operates through bootstrapping, random feature selection, and majority voting. Model evaluation was conducted using accuracy, precision, recall, and F1-score metrics, and further validated through cross-validation to ensure model generalization and reduce overfitting. The research variables include academic data, non-academic data, and psychometric test results as independent variables, while the dependent variables are major recommendations and user satisfaction levels. The evaluation results show that the system performs well with an accuracy of 83%, precision of 87.5%, recall of 80%, and F1-score of 88.52%. The system was developed using Vue.js, Express.js, and Python and can be accessed via both desktop and mobile devices. Overall, the system effectively assists students in identifying their potential and supports more structured educational planning.