AI-powered delay risk prediction in construction projects: leveraging machine learning, deep learning and hybrid models

Construction work delays are a long-term challenge resulting in schedule changes, budget conflicts and dissatisfaction. Traditional time management techniques struggle to address complex project relationships and uncertainties in contemporary construction projects. AI-based methods including machine learning (ML), deep learning (DL) and hybrid models are being used for foreseeing delays proactively. The study aims to estimate the abilities of advanced AI models to forecast construction delay percentages by using real-time data from 68 different construction projects in India. A dataset was prepared by collecting data on delay factors involving varous stakeholders and project attributes through a structured survey. Data quality was enhanced through pre-processing methods including interpolation, outlier detection and removal, normalization and feature selection. The research developed a predictive model that integrates a convolutional neural network (CNN) with a bidirectional long short-term memory (BiLSTM) network, joining spatial and temporal feature extraction for improved accuracy. This CNN–BiLSTM hybrid model outperformed multiple AI frameworks, achieving the highest accuracy (95.56%) and F1-score (95.58%), indicating its effectiveness in capturing spatial dependencies and sequential patterns in delay-related data. Research shows that combined DL algorithms surpass single-model ML and DL applications in delay predicting, allowing construction professionals to make informed project plans and minimize risks.