Haima-uxc/STLE-Africa-Analysis

# STLE Analysis - Explainable AI for Sustainable Transport Efficiency in Africa Supporting materials for the research paper submitted to *Research in Transportation Business & Management*. ## Quick Start 1. Install R packages: `tidyverse`, `randomForest`, `xgboost`, `treeshap`, `ggplot2`, `corrplot`, `gridExtra` 2. Run: `source("main_analysis.R")` 3. Results will be generated in `figures/` and `tables/` folders ## Dataset - **18 African countries** (2021-2023 data) - **5 indicators**: LPI Overall, LPI Infrastructure, LPI Customs, CO₂ Emissions, Road Deaths - **Sources**: World Bank LPI 2023, IEA 2022, WHO 2021 ## Key Results - **Infrastructure capacity** most important (Mean Absolute SHAP = 2.57) - **Botswana leads** (STLE = 76.8), **Madagascar lowest** (STLE = 33.4) - **Strong model performance**: R² = 0.757 - **12 of 18 countries** constrained by infrastructure deficits ## Files Structure ``` ├── main_analysis.R # Complete analysis script ├── data/ │ ├── raw_data.csv # Original dataset │ ├── processed_data.csv # STLE scores & normalized values │ └── shap_results.csv # SHAP analysis results ├── tables/ # All analysis results └── figures/ # All visualizations (25+ figures) ``` ## Requirements - R 4.3.0 or higher - Required packages: tidyverse, randomForest, xgboost, treeshap, ggplot2, corrplot, gridExtra, caret, VIM, RColorBrewer, viridis, plotly, knitr - 8GB RAM recommended - Execution time: 10-15 minutes ## Citation Paper submitted to *Research in Transportation Business & Management* (Under Review)