•  
  •  
 

Abstract

Managing projects in the modern world seems to be growing in dimness as it incorporates risks from multiple advancements which cut across technology, interrelationships between regions and developing societal expectations. Typical project management systems do not have the necessary accuracy or flexibility to include the possible impact of change in key components, such as scope change, resource changes, and external factors changes on the end values of performance metrics such as time, cost and quality. Such constraints are overcome in this work by implementing Deep Learning models such as FNN, LSTM networks, Attention Mechanisms or Ensemble models as the predictive analytics in project management. These models leverage the Construction Industry Institute Project Dataset in establishing a robust framework for modelling nonlinear associations and the multiple relationships of project data. The study presents predictive models focused on the relationships that assist in predicting the impact of the construction project on the three major variables, timeline, cost and quality. Also, the objective of leveraging such models includes assessing their robustness across various industries, comparing the performance of different Deep Learning architectures, and developing a data-driven framework to make decisions proactively. For predictions, the proposed Ensemble model combines LSTM networks and Attention Mechanisms and predicts best at 89% for timeline delays, 92% for cost overruns, and 85% for quality impacts. These models also allow project managers to address risks in an anticipatory fashion, accelerating resource utilisation, and enhancing decision-making capability. This study is significant as it advances the formulation of an effective predictive framework, it also contributed towards the evaluation of a number of Deep Learning architectures, and developing pragmatic tools for project management practice. Finally, the study contributed towards the growing literature frontier of Ensemble modelling in the project management domain. It demonstrates complete predictive performance assessment including hyperparameter tuning for advanced predictive metrics. This study integrates conventional approaches and the contemporary requirements of the project setting thus offering a new way of addressing complexities and dynamisms in project management. By allowing for better foresight and active response, it creates possibilities for improved project performances in different sectors.

First Page

140

Last Page

150

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Download

Share

COinS