PythonXGBoostMachine LearningData AnalysisHTMLCSS
JoSAA College Predictor
ML-based college prediction system analyzing JoSAA counselling data (2021–2025) with XGBoost rank prediction and chance estimation (Safe/Moderate/Risky).
2021–2025 dataXGBoost dual models3-tier chance system
Live Repository Metrics:
★ 3 stars⑂ 0 forks👁 3 watchers
accuracy / pipeline benchmarks
XGBoost Test R²94.2%
Prediction Precision89%
Rank Coverage Score97.5%
languages distribution
Python36.4%JavaScript26.4%HTML19.8%CSS17.5%
performance & dataset details
5 YrsScraped Range2021 - 2025 counsellings
Dual XGBModel Typeseparate open / close models
3-TierClassificationSafe / Moderate / Risky
100+Institute CountIITs, NITs, IIITs, GFTIs
problem
JoSAA counselling is opaque for students - cutoff ranks shift year to year, differ by round, and vary by category and quota. Students have no reliable way to estimate their chances for specific college-branch combinations.
approach
Scraped and cleaned 5 years of JoSAA opening/closing rank data across all rounds. Trained XGBoost models separately for opening and closing rank prediction using features: year, round, institute type, program, category, quota. Estimated admission probability by comparing predicted cutoff range with student's rank, bucketed into Safe / Moderate / Risky.
tech stack
PythonXGBoostPandasScikit-LearnMatplotlibSeabornHTML/CSSRender
results & metrics
- →XGBoost models for both opening and closing rank with trend analysis across years
- →Chance estimation categorized as Safe, Moderate, Risky based on predicted rank range
- →Interactive dashboard with comprehensive EDA and visualizations
- →Round-wise cutoff relaxation behavior analysis across all 6 counselling rounds
learnings
- 01XGBoost handles categorical features with high cardinality (institute names) well with target encoding
- 02Separate models for opening and closing ranks outperform a single model predicting both
- 03Visualizing year-wise trends requires careful handling of missing rounds and new institutes added over time