Lindiwe-22/South-African-Crime-Intelligence-Report

South Africa crime statistics analysis — GBV, rape, femicide, SAPS Q3 2025/26 | Python | SAMRC | Statsmodels | ARIMA # 🔍 South African Crime Intelligence Report ## Gender-Based Violence, Sexual Offences & Femicide: A Critical Data Analysis™ > *"Crime remains unacceptably high and continues to affect many families > and communities."* > — Acting Police Minister Firoz Cachalia, 20 February 2026 --- ## 🎯 Project Overview This notebook interrogates the claim by SAPS Crime Registrar Major-General Thulare Sekhukhune that reported rape cases decreased by 3.2% in Q3 FY2025/26 — and places that claim within two decades of femicide surveillance data published by the South African Medical Research Council (SAMRC). Produced from the perspective of a **private security analytical division** conducting occasional work with SAPS, this notebook applies rigorous data science methodology to one of South Africa's most critical public health and safety challenges. --- ## ❓ Research Questions 1. Is there truly a decrease in reported rape as stated by Maj-Gen Sekhukhune? 2. What does the SAMRC's 20-year femicide surveillance data reveal that SAPS statistics cannot? 3. Which provinces and demographics are at highest risk? 4. What does statistical modelling predict for FY2025/26 and FY2026/27? --- ## 📓 Notebook Structure | Phase | Cells | Content | |-------|-------|---------| | **Phase 1** | 1–2 | Environment setup, CSV data pipeline | | **Phase 2** | 3–10 | Exploratory data analysis — 8 visualisation panels | | **Phase 3** | 11–15 | Statistical modelling — correlation, anomaly detection, regression, forecasting | | **Phase 4** | 16–19 | Intelligence report, dashboard, recommendations | --- ## 🔬 Methods Applied - Pearson & Spearman correlation analysis - Z-score anomaly detection (annual + provincial) - Isolation Forest anomaly detection - OLS regression — simple & multiple with full diagnostics - Linear extrapolation with 95% prediction intervals - Holt's Exponential Smoothing - ARIMA(1,1,0) with confidence intervals - 3-model ensemble forecasting --- ## 📊 Data Sources | Source …