Project Memo

This repository contains the full workflow, data, and code used to develop a machine-learning framework for accurate construction cost prediction by integrating micro-level project estimates with macro-level U.S. construction spending indicators. The project includes two source datasets—(1) construction_estimates.csv, containing project-specific material cost, labor cost, profit rate, discount/markup, and total cost, and (2) construction_spending.csv, containing national monthly spending across public, private, residential, and non-residential sectors. A third file, merged_construction_dataset.csv, combines both sources to create a unified feature set used for feature selection (RFE, SelectKBest, correlation filtering) and supervised modeling (Linear Regression and Random Forest). All scripts for preprocessing, ARFF conversion, model training, evaluation, and figure generation are included, providing a reproducible pipeline for researchers and practitioners interested in improving construction cost forecasting.