Forecasting the scheduling issues in engineering project management: Applications of deep learning models

摘要

Since project monitoring aims to make decisions, which have future impacts on a project's success, accurate forecasting of project characteristics is of greater importance. This paper proposes the detection of problems in scheduling projects by introducing the forecasting of successors and renewable resource features. For this purpose, this paper presents the novel applications of forecasting models "long short term memory" (LSTM) and "gated recurrent unit" (GRU) in this study. Subsequently, forecasting results of successors and renewable resource characteristics of projects are determined by covering historical records' observed values from real-life engineering projects' datasets. Results show that LSTM and GRU models' proposed applications can reduce errors from one up to seven steps-ahead forecastings. Moreover, forecasting the increased number of successors may require more renewable resources to effectively complete the jobs in projects. Generally, the proposed models are reliable and robust to application in forecasting the project scheduling task.