Improving the accuracy of project duration with artificial intelligence
927 Downloads
Last Updated: 2-2019
This research focusses on the prediction of utility construction projects. These predictions
are important for contractors because of growing costs. To improve the project duration of
the utility construction projects, this research tests two different techniques through a case
study. The case study is conducted at the utility department of a contractor; this department
specialises in utility projects. This research then analyses and improves the accuracy of the
estimated project durations by using data from the utility construction projects.
Based upon related work in the field of project durations, artificial neural networks (ANN)
and regression are used to determine the project durations of construction projects. Before
the techniques are tested in a case study, the current situation of the contractor is analysed
to evaluate the performance of the techniques. Prediction models and scripts are thereby
developed based on the two techniques. Different parameters are selected and tested in the
model to improve the prediction of project durations of utility projects.
Based on the mean absolute error (MAE), it appears that the ANN and regression techniques
do not outperform the current situation with the contractor. The current MAE of the utility
construction projects of the contractor is 8.4 days. The MAE of the predicted project
duration of the utility construction projects with regression is 9.2 days. The MAE of the
predicted project duration with regression increased by 9.5% on average relative to the MAE
of the current project duration of the utility construction projects. Furthermore, the MAE of
the utility construction projects planned with ANN is 8.52 days. The MAE of the predicted
project duration with ANN increased by 1.4% on average relative to the MAE of the current
project duration of the utility construction projects. Based on the dataset of the utility
construction projects used in this research, the current MAE of the contractor is superior to
the MAE corresponding with the ANN technique and the regression technique.