Intelligent Cost Estimation in ERP Systems: A Comparative Analysis of Machine Learning Models

Dataset and Code for Intelligent Cost Estimation in ERP Systems This repository contains the dataset, Jupyter Notebook modeling scripts, and the Enterprise Resource Planning (ERP) system integration source code for the paper: "Intelligent Cost Estimation in ERP Systems: A Comparative Analysis of Machine Learning Models". This research proposes an Intelligent Cost Estimation module integrated into an ERP for Project Management (ERP-PM) system. By training segmented machine learning models for specific job categories, the system predicts the actual Owner Estimate Cost (OEC) based on bidding prices and regional locations to prevent cost overruns and procurement failures. 📂 Repository Structure & Files DataSkripsi_Bersih.csv Description: The primary historical dataset containing 110 real corporate project transaction logs. Columns: Bidding Value, Region, Job Category, Actual Expenditure. DataSkripsi_Tambahan.csv Description: The augmented dataset containing 10,000 synthetic transaction records generated to facilitate stable machine learning convergence and prevent overfitting. skripsi_perbandingan_model.ipynb Description: A Jupyter Notebook containing the benchmarking pipeline of five models (Linear Regression, Moving Average, Random Forest, XGBoost, and Artificial Neural Network) evaluated using 10-fold cross-validation. It also implements the Borda Count rank aggregation. skripsi_terstruktur.ipynb Description: A structured Jupyter Notebook focused on training, evaluating, and serializing the selected champion model (Linear Regression) into a .pkl format for production deployment. erp-pm-source.zip Description: A clean, lightweight ZIP file containing the source code of the Laravel web-based ERP-PM application integrated with the machine learning inference engine. Dependencies like vendor/ and node_modules/ are excluded for size optimization. 🚀 Getting Started A. Running the Machine Learning Models (Python) Install requirements: pip install pandas numpy scikit-learn xgboost tensorflow joblib matplotlib seaborn Open Jupyter Notebook or JupyterLab. Run skripsi_perbandingan_model.ipynb to retrain the models and generate .pkl/.keras files. B. Running the Laravel Web Application Extract erp-pm-source.zip into a directory. Install dependencies: composer install npm install Configure .env file: copy .env.example to .env and set up database parameters. Run migrations: php artisan migrate Start local server: php artisan serve npm run dev The machine learning model is integrated in the backend controller using Inter-Process Communication (IPC) via predict_all_models.py located in the backend folders.