OnsMhiri/tunisian-bac-perfomance-analysis

# BAC Tunisia — Data Mining Project ### Socioeconomic Determinants of BAC Success Across 26 Governorates (2018–2025) --- ## Project Overview This project applies data mining techniques to the Tunisian BAC (Baccalauréat) exam results dataset, covering all 26 governorates from 2018 to 2025. It goes beyond descriptive analysis to discover hidden patterns — non-obvious groupings, threshold effects, unexpected performers, temporal trajectories, and convergence dynamics. **Type of data:** Panel data (hybrid). Each governorate has 8 yearly observations. This is NOT classical time series (too short for ARIMA/LSTM), but it IS rich enough for trend analysis, trajectory clustering, change point detection, and panel regression with fixed effects. **Project Goal:** Understand which socioeconomic factors (poverty, illiteracy, school infrastructure) best predict BAC success rates, identify regional clusters with distinct performance patterns, and detect temporal trends that reveal whether educational inequality is increasing or decreasing. --- ## Dataset | File | Description | |------|-------------| | `data/raw/bac_2018_2025.csv` | Raw BAC results — wide format, one row per governorate (Arabic names) | | `data/raw/ecoles.csv` | School infrastructure data (6109 schools across all governorates) | | `data/raw/socio_economic.csv` | Socioeconomic indicators: poverty rate, illiteracy rate, population | | `data/clean/bac_panel.csv` | Generated by NB02 — long format, 208 rows (26 × 8 years) | | `data/clean/bac_crosssectional.csv` | Generated by NB02 — 26 rows with engineered features | | `data/enriched/enriched_bac_data.csv` | Generated by NB01 — enriched dataset with all features merged | **Raw features per governorate:** BAC rank, number of candidates, number who passed, success rate (per year 2018–2025), plus static: number of lycées (high schools), poverty rate, illiteracy rate, population. **Note on Sfax and Tunis:** Sfax and Tunis are each split into two exam centers (1 an …