KCET College Prediction System Using Random Forest Regression Based on Historical Cutoff Trends

The Karnataka Common Entrance Test (KCET) counselling process requires students to make critical college and branch selection decisions using rank, reservation category, and previous cutoff trends. Manual analysis of cutoff data is difficult because students must compare several colleges, branches, categories, and yearly variations before preparing an option-entry list. This paper presents a machine learning-based KCET College Prediction System that assists students by predicting eligible engineering colleges from historical cutoff data collected for the years 2021 to 2024. The dataset contains approximately 10,000 records covering more than 60 colleges, five engineering branches, and eight reservation categories. The proposed system applies preprocessing, categorical encoding, regression modelling, and webbased deployment. Linear Regression, Decision Tree Regression, Random Forest Regression, and ensemble learning concepts are considered, and Random Forest Regression is selected due to its ability to model non-linear cutoff patterns and reduce overfitting. The model is trained using college, branch, category, and year as input features, with closing rank as the target variable. The trained Random Forest model with 200 decision trees achieved an R² score of 0.9948. A Flask-based web interface accepts student rank, preferred branch, and category, and displays eligible colleges in sorted order. The system reduces manual effort, improves counselling decision support, and provides a structured approach for KCET option-entry planning.