Construction risk prediction is software that uses data from past projects and live data from the site to predict delays and cost overruns, so that the contractor can act before one late task holds up everything after it. A cost overrun means that a project spends more than its budget.
On a building site, tasks depend on each other. The plumbing on a floor cannot start until the slab is cast. The tiling cannot start until the plumbing is tested. So one late task can delay many others. The chain of tasks that decides the finish date is called the critical path. A delay on the critical path delays the whole project.
The software learns from history. A machine learning model is trained on the schedules and the actual finish dates of thousands of past projects. From these, it learns which patterns usually come before a delay. Examples are a supplier who often delivers late, a crew that is falling behind, or work planned during heavy rain. Then it reads the current project's schedule and its live progress, often from construction progress monitoring. Some products also read outside data, such as weather forecasts and the prices of steel and cement.
Take, for example, a developer building a 32-storey residential tower in Kolkata. The schedule allows 8 days to cast each floor's slab. On the last five floors, the site's records show an average of 10 days. So far the tower is only about 2 weeks behind. But the model also sees that the monsoon starts in six weeks, and on similar past projects, slow slab work before the monsoon became slower during it. So it predicts that the structure will finish 11 weeks late. That would push the handover of the flats past the date the developer registered with the state's RERA authority. Under the Real Estate (Regulation and Development) Act, or RERA, a developer must pay buyers interest for each month of late handover.
The software also suggests what to change. Here it shows that a second set of formwork, the moulds into which the concrete is poured, would bring each slab back to 8 days. The project manager compares the cost of the extra formwork with the cost of a late handover, and approves it. The warning and the decision are both recorded in construction management software.
The prediction is only as good as the data. If a site records its progress late or wrongly, the model learns the wrong patterns. The method combines forecasting from time series data, numbers recorded one after another over time, with anomaly detection, which spots a value that does not fit the usual pattern. The same idea, applied to machines instead of projects, is predictive maintenance.