Search results

Construction projects generate large volumes of data which can be used for better management of projects. In this paper, key project data is manually extracted from project site meeting minutes. Knowledge discovery technologies are then used to predict the final project duration of active projects.

Project planning and effective leadership/governance were identified from literature as the most significant factors that impact the duration of projects. These factors were hence considered as the main features for a data mining process. Items supporting these factors were extracted from site meeting minutes to create a database of 27 civil engineering projects executed over the last ten years. Data mining algorithms were used to predict from this data whether or not an active project will be completed on time.

The research showed that information from project site meetings can be used to predict final project duration of active projects with accuracy of above 80% when using random forest algorithms from Orange and RapidMiner data mining applications. The value of data to predict project duration from project site meeting minutes is demonstrated but it only becomes practically useable if the format of minutes is suitably standardised.

Some of the data mining algorithms provided accuracies of above 80% in predicting final project duration and proved the value of project data from site meeting minutes. The random forest algorithms are particularly suited to this type of data. The factors with the highest impact on the prediction of the project duration are those related to the progress of the project.

This study for the first time shows that data from site meeting minutes of past and current projects can be used to make accurate predictions of final project duration of active projects and serve as a project management tool to activate remedial measures.