This repository contains the code, models, and analysis for a machine learning-based study on team performance in serious gaming environments. The project combines exploratory data analysis (EDA) with supervised learning models—including Logistic Regression, Random Forest, Support Vector Classifier, and Multi-Layer Perceptron—to classify team success based on behavioural indicators. Key features include:
- Preprocessing workflows (scaling, feature selection, stratified splitting)
- Implementation of 12 supervised classification models using scikit-learn
- Hyperparameter optimization using GridSearchCV with F1-scoring
- Model interpretability using SHAP (Shapley values) and coefficient-based feature importance
- Focus on communication, collaboration, and leadership constructs in team behaviour
This work forms part of a broader PhD research agenda on AI-supported behavioural analysis in serious games and citizen science. This analysis was done in collaboration with Freark de Lange at Campus Fryslan - University of Groningen.