Governance-Aligned ESG Indicators Dataset for Interpretable Sustainability Analytics (Astra Agro Lestari, 2022–2024)

Description

This dataset contains structured Environmental, Social, and Governance (ESG) indicators derived from publicly available sustainability and annual reports of PT Astra Agro Lestari Tbk for the period 2022–2024. The data were manually extracted, standardized, and organized into a governance-aligned analytical framework to support interpretable sustainability analytics. The dataset includes environmental indicators (including carbon intensity measures), financial performance variables, and governance disclosure metrics. It was prepared to demonstrate a proof-of-concept framework integrating lightweight machine learning and explainable artificial intelligence (SHAP) within an energy-efficient and governance-driven analytical architecture. This dataset is intended for research and educational purposes in ESG analytics, sustainable computing, and interpretable machine learning applications.

Steps to reproduce

1. Obtain the publicly available sustainability reports and annual reports of PT Astra Agro Lestari Tbk for the years 2022–2024. 2. Extract greenhouse gas emissions data (Scope 1, Scope 2, and Scope 3), financial variables (revenue and related indicators), and governance disclosure information. 3. Structure the data according to the governance-aligned ESG taxonomy used in this dataset: - Environmental indicators → emissions and carbon intensity - Governance indicators → standardized disclosure scores 4. Convert financial values from IDR to USD using the exchange rates provided in idr_usd_rates.csv (if currency normalization is required). 5. Compute carbon intensity as: - Carbon Intensity = Total Emissions / Revenue 6. Normalize or benchmark carbon intensity using industry reference values provided in industry_ci_minmax.csv (if comparative analysis is required). 7. Use the structured dataset for interpretive machine learning analysis: - Apply a lightweight regression model (e.g., XGBoost or ELM) - Use emissions, financial variables, and governance scores as input features - Use carbon intensity or composite ESG score as response variable 8 Apply SHAP (SHapley Additive exPlanations) to interpret feature contributions and examine variable influence. 9. Ensure that model design prioritizes computational efficiency (minimal iterations, lightweight configuration) in alignment with green algorithm principles.

Institutions

• Binus University

  Jakarta, Jakarta

Categories

Licence