Byggkollen AI - An analysis tool for Construction projects

Purpose and goal

In this project machine learning was used (more specifically decision tree and regression) in order to forecast delays which can occur in construction projects. The machine learning models have been trained and evaluated on data from Byggkollen (a software tool developer for Swedish construction companies), where AI forecasts of construction projects also will be available.

Expected results and effects

Comprehensive and detailed historical reports regarding specific construction companies and construction projects resulted in data that was largely individual in its kind with regard to the construction company / construction project. This made it possible to train models that were tailored to specific customer profiles / construction projects and thus very relevant to the construction companies involved.

Planned approach and implementation

In order to understand the structure of the available data and the problem at hand, the data was visualised to spot features and correlations. The construction projects were divided into two categories, the ones which were finished on schedule and the ones that were not, features were compared between the two groups to spot possible patterns. Finally, the models were trained using relevant features which resulted in an AI that forecasts wether or not construction projects will be delayed.