Modelling Construction Sector Output Using Steel and Cement Indicators: A Machine Learning Time Series Approach

Abstract

Cement and steel play an important role in the construction industry. In this paper, we study the relationship between key steel and cement indicators and construction sector output in India, using the Index of Industrial Production (IIP) for Infrastructure/Construction Goods as a proxy for sector performance. The monthly data such as the steel and cement production, their respective indices and growth rates, the monthly IIP For Infrastructure was collected data from April 2011 to March 2025. The data was cleaned and pre-processed and missing values were handled. Feature engineering was done and since the data was a time series, lag features were introduced to capture temporal dependencies. Bagging and boosting machine learning models were applied to the dataset to model and predict construction sector output. The top five and top ten most important features were identified and used for retraining and hyperparameter tuning. Among these, steel-related features—particularly Steel Index, Steel Growth, and Steel Production—along with lagged IIP values emerged as the strongest predictors of construction sector output. Cement-related variables had marginal influence by comparison. This machine learning approach demonstrates its potential in economic modelling and can assist policymakers and industry stakeholders in making data-driven decisions.