Improving ERP Adoption Through Predictive Modeling: A Data-Driven Recommendation System

Enterprise Resource Planning (ERP) adoption remains a critical challenge for organizations due to cost, employee resistance, and operational inefficiencies. This research presents a data-driven predictive modeling framework that leverages feature engineering, dimensionality reduction, and machine learning algorithms to enhance ERP satisfaction and adoption. The study applies Recursive Feature Elimination (RFE) for feature selection, Principal Component Analysis (PCA) for dimensionality reduction, and interaction terms to improve interpretability. Seven machine learning models, including Random Forest, Gradient Boosting, Support Vector Machine, Logistic Regression, K-Nearest Neighbors, XGBoost, and LightGBM, are trained and evaluated using cross-validation with Stratified K-Fold. Model performance is assessed through accuracy, precision, recall, and F1-score, with SHAP and Permutation Importance ensuring interpretability. The best-performing model is used to predict future ERP adoption trends, and recommendations are derived based on key influencing factors. Visualizations such as feature importance plots, confusion matrices, and impact assessments are generated to provide actionable insights. The proposed system aids in optimizing cost, enhancing employee training, and streamlining organizational processes, ensuring higher ERP adoption rates and long-term operational efficiency.