A Random-Forest-Based Framework for Predicting Schedule Achievement in Cloud Enterprise Resource Planning Projects Using the Critical Path Method and the Programme Evaluation and Review Technique

• J. Haekal Department of Industrial Engineering, Universitas Esa Unggul, Jakarta, Indonesia https://orcid.org/0000-0001-8543-1683
• R. Mu’min Department of Management, Universitas Negeri Jakarta, Jakarta, Indonesia https://orcid.org/0009-0007-2789-8016
• S.A. Lesmana Faculty of Engineering, Universitas Mercubuana, Jakarta, Indonesia https://orcid.org/0000-0002-7546-7778
• N.E. Triana Faculty of Engineering, Universitas Mercubuana, Jakarta, Indonesia https://orcid.org/0009-0004-2732-626X
• Y.D. Suwandari Faculty of Engineering, Universitas Mercubuana, Jakarta, Indonesia https://orcid.org/0000-0001-6072-915X
• D. Rukmayadi Department of Industrial Engineering, Universitas Bhayangkara Jakarta Raya, Indonesia https://orcid.org/0000-0002-2745-5123
• I.G. Prihanto Research Center for Testing Technology and Standards, National Research and Innovation Agency (BRIN), Tangerang Selatan, Indonesia https://orcid.org/0000-0001-8959-0972

DOI:

https://doi.org/10.7166/37-2-3151

Abstract

This paper introduces a robust framework for improving project timeline predictions in cloud-based enterprise resource planning (ERP) implementations by combining traditional project management techniques (the critical path method [CPM] and the programme evaluation and review technique [PERT]) with advanced machine learning methods. The primary problem is that traditional project management techniques such as CPM fail to handle uncertainty in project timelines effectively. The features for the dataset were derived from the CPM and PERT framework and were matched with 4M (man, method, machine, material) to identify which kinds of uncertainties happened in the project. Hypothetically, the random forest algorithm was used as the primary predictive tool, but the research also used a neural network model to do the benchmark in metrics such as accuracy, precision, recall, and F1-score. In the results, random forest outperformed as predicted. By integrating CPM and PERT-derived features, the model identified critical factors, including “method” and “machine”, that strongly influence project efficiency. The framework achieved a near-perfect discrimination (AUC: 0.997), demonstrating its effectiveness in navigating uncertainties and enabling data-driven decision-making for enhanced project outcomes.