Estimation of Delay in Prefabricated Projects Using Modern Machine Learning Approaches (Case Study: Baghdad)

We analyzed the prior literature on estimating delays in project time in the presentation section, andwe found that data uncertainty could be minimized using generative adversarial network (GAN) toaugment our dataset, which uses data to produce findings that mimicked actual world circumstances. We organized the findings accordingly. In the initial finding, we utilized four (4) algorithms on a dataset of twenty-one features containing 284, 807 transactions i.e. multilayer perceptron (MLP) neural network, support vector machine (SVM), decision tree, and k-nearest neighbor (KNN). The findings established that MLP neural network produced the largest accuracy value of (90.72%), followed with SVM (78.43%), Decision Tree (77.64%), and KNN (74.5%).Next, the GAN was used to augment the dataset to a total of 400,00 transactions, allowing the augmented dataset to result in a number of delay samples of 609. The four (4) algorithms were subsequently re-evaluated with the expanded dataset to classify and identify project delays in the dataset. The results indicated that augmentation using GAN enhanced the accuracy of the models overall. From the first process, using the MLP neural network reached an accuracy of 98.76% andSVM was 82.03%, decision tree was 80.31% and KNN was 79.95%.