Business Analytics for Construction Labour Productivity in Australia: An Explainable Predictive Decision-Support Framework Integrating Workforce, Weather, Equipment and Site Conditions

This research paper presents an Integrated Business Analytics decision-support framework for improving construction labour productivity in Australia. The study examines how workforce characteristics, working-time conditions, equipment availability, weather conditions, project context, and site conditions can be integrated and analysed to generate meaningful productivity insights and support construction-management decisions. Rather than focusing solely on the development of a machine-learning model, the research positions machine learning as one component within a broader Business Analytics lifecycle. The proposed framework combines descriptive analytics to identify productivity patterns, diagnostic analytics to investigate factors associated with productivity variation, predictive analytics to estimate near-term productivity and productivity risk, explainable analytics to communicate model-attributed factors, and scenario analytics to compare alternative workforce, scheduling, equipment, and site-management options. The research defines construction labour productivity using output relative to labour hours and considers activity-specific productivity measurement to account for differences between construction activities. It also establishes a structured data architecture integrating daily construction reports, workforce and labour-hour records, equipment information, project and site data, and Australian weather information. Particular attention is given to prediction-time data availability, data quality, project-level validation, temporal leakage, and the distinction between predictive association and causal inference. The framework incorporates statistical and machine-learning approaches, including baseline models and tree-based algorithms, together with model evaluation metrics such as MAE, RMSE, and R². Explainable Artificial Intelligence techniques, particularly SHAP-based analysis, are proposed to help construction stakeholders understand the factors contributing to individual predictions without interpreting model-attributed contributions as causal effects. The study further introduces scenario and decision analytics to connect analytical outputs with practical construction-management actions. These include workforce allocation, skill-mix planning, equipment allocation, work scheduling, workfront management, material coordination, and responses to adverse weather and site conditions, while maintaining safety, quality, environmental, workforce, and operational constraints. The research contributes a structured Business Analytics approach that connects construction data with measurable productivity indicators, diagnostic insights, near-term predictions, productivity-risk assessment, explainability, scenario comparison, and evidence-informed decision support. The framework is intended to provide a foundation for future empirical validation using real Australian construction project data and to support more systematic, data-driven productivity management within the Australian construction industry.