Machine Learning Model to Evaluate the Influence of Environmental, Geographic, and Sociodemographic Variables on the Prediction of Socio-Environmental Conflicts in the Mining Sector of Peru

Socio-environmental conflicts (CSA) are a significant challenge due to the negative impacts on the environment and local communities; mining, although important for the economy, has generated tensions due to environmental degradation and competition for natural resources. The objective of this study is to develop a machine learning (ML) model for CSA prediction in the Peruvian mining sector, using historical information of environmental and geographic variables provided by the Ombudsman's Office (DP), Ministry of Energy and Mines (MINEM) and the National Institute of Statistics and Informatics (INEI). The methodology used a dataset built based on historical information from the aforementioned entities, applying oversampling techniques such as Random Oversampling (ROS) and SMOTE for class balancing of the variable to be predicted (state of the CSA) and then using supervised ML algorithms such as Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), for the prediction of SWCs in the departments of Amazonas and Ancash during the period January 2008 - May 2024. The results obtained, once the above ML algorithms were compared, showed an accuracy of 98.66% with the RF-based model after applying hyperparameters, surpassing performance studies related to conflict prediction. The evaluation of these models showed their potential in environmental management but recognizes the need for further studies and continuous adjustments to improve their accuracy and applicability in the field of CSA prediction to support informed decision making.