Intelligence Driven Project Risk Prediction and Resource Optimization for Complex Engineering and Manufacturing Projects

Authors

  • Asheke Mostofa M.S. in Engineering Management, Lamar University, Beaumont, Texas, USA Author

DOI:

https://doi.org/10.63125/hv05m379

Keywords:

Prediction, Optimization, Risk, Engineering, Manufacturing, Contingency Decision

Abstract

Artificial intelligence–driven project risk prediction and resource optimization address uncertainty and competing resource requirements in complex engineering and manufacturing projects. This study developed a quantitative framework linking early risk estimates with feasible labor, equipment, material, and contingency decisions. A synthetic numerical example represented 240 projects across 12 organizations, with 192 projects assigned to development and 48 to chronological testing. The framework specified retrospective longitudinal analysis, statistical and machine-learning comparisons, probability assessment, and matched resource-policy evaluation. Regression, classification, and scheduling outcomes were distinguished, while project-level separation and organizational clustering informed the proposed analytical structure. In the findings, gradient boosting achieved 93.75% delay-classification accuracy, 94.12% sensitivity, a 91.43% F1-score, and a receiver operating characteristic area of 0.970. Its Brier score was 0.071, and mean warning lead time was 14.8 days. Constructed adjusted associations linked complexity, technical changes, procurement interruptions, and equipment downtime with adverse outcomes, whereas workforce availability showed an inverse relationship. Integrated optimization reduced makes pan from 462 to 402 days, equivalent to 12.99%, and resource expenditure from USD 6.80 million to USD 5.90 million per project, equivalent to 13.24%. Equipment utilization increased by 9.4 percentage points, while recovery duration declined from 6.5 to 3.8 days. Nested-validation accuracy was 92.71%, illustrating the distinction between apparent development performance and validation results. Subgroup and disruption comparisons demonstrated reporting approaches for contextual variation, resource flexibility, and operational sensitivity. Independent external validation remained absent, and repeated evaluations did not establish replication. However, all numerical findings, intervals, and significance measures were constructed examples rather than estimates from observed records. They did not establish actual model superiority, causal effects, financial savings, or external generalizability. The study contributed an integrated analytical framework that separated predictive reliability, feasible resource actions, and operational outcomes, while identifying the empirical evidence required to evaluate their relationship in engineering and manufacturing practice.

References

Downloads

Published

2026-10-11

How to Cite

Asheke Mostofa. (2026). Intelligence Driven Project Risk Prediction and Resource Optimization for Complex Engineering and Manufacturing Projects. American Journal of Data Science and Analytics, 7(10), 126-164. https://doi.org/10.63125/hv05m379

Cited By: