Analysing Influential Factors in Student Academic Achievement: Prediction Modelling and Insight

Abstract

The fascination with understanding student academic performance has drawn widespread attention from various stakeholders, including parents, policymakers, and businesses. The 'Students Performance in Exams' dataset, available on platforms like Kaggle, stands as a treasure trove. It extends beyond test scores, encompassing diverse student attributes like ethnicity, gender, parental education, test preparation, and even lunch type. In our tech-driven age, predicting academic success has become a compelling pursuit. This study aims to delve deep into this dataset, utilizing data mining methods and robust classification algorithms like Logistic Regression and Random Forest in a Jupyter Notebook environment. Rigorous model training, testing, and fine-tuning strive for the utmost predictive accuracy. Data cleaning and preprocessing play a crucial role in establishing a reliable dataset for accurate predictions. Beyond numbers, the project emphasizes data visualization's impact, transforming raw data into comprehensible insights for effective communication. The Logistic Regression Model exhibits an impressive 87.6% accuracy, highlighting its potential in predicting academic performance. Moreover, the Random Forest Model excels with a remarkable 100% accuracy in forecasting student grades, showcasing its effectiveness in this domain.