Prediction of failures in the project management knowledge areas using optimized ensemble models in software companies
Abstract Project management involves coordinating various tasks and phases to achieve specific goals. Artificial Intelligence has the potential to enhance decision-making by providing data-driven insights. However, its application in project management, especially for predicting project failures, is...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-07-01
|
| Series: | Discover Applied Sciences |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s42452-025-07337-y |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849769090541944832 |
|---|---|
| author | Lamia Berriche Abderrazak Loulizi |
| author_facet | Lamia Berriche Abderrazak Loulizi |
| author_sort | Lamia Berriche |
| collection | DOAJ |
| description | Abstract Project management involves coordinating various tasks and phases to achieve specific goals. Artificial Intelligence has the potential to enhance decision-making by providing data-driven insights. However, its application in project management, especially for predicting project failures, is still not widespread. This study investigates how hyperparameter tuning and ensemble learning techniques can improve the accuracy of predicting project failures. We applied grid search to optimize several classical machine learning models, including K-Nearest Neighbors, Support Vector Machine, Decision Tree, and Logistic Regression, with average accuracies of 92.46, 93.61, 93.60, and 84.74%, respectively. Our optimized classical machine learning models’ accuracies outperformed state-of-the-art techniques. To further improve performance, ensemble techniques such as soft voting, hard voting, and stacking were explored. Combinatorial optimization was used to identify the most effective set of base learners for ensemble methods. The best soft voting combination—DT, KNN, LR, and CatBoost—achieved an accuracy of 93.21%. The stacking model with DT, KNN, and LR as base-learners and SVM as the meta-learner performed best, reaching 93.73%. CatBoost demonstrated the best accuracy, achieving 94.02%, and demonstrated the best generalization. These models showed strong performance in predicting failures related to scope and cost management but were less accurate when predicting failures in human resource management. Finally, the SHAP model showed that the ‘Reason for Fail’ feature demonstrated the largest impact on the prediction of all classes, while the project manager’s education is an important predictor of cost overruns. |
| format | Article |
| id | doaj-art-c10d2fe39d38438c8dfc5bf9f281c879 |
| institution | DOAJ |
| issn | 3004-9261 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | Springer |
| record_format | Article |
| series | Discover Applied Sciences |
| spelling | doaj-art-c10d2fe39d38438c8dfc5bf9f281c8792025-08-20T03:03:34ZengSpringerDiscover Applied Sciences3004-92612025-07-017712410.1007/s42452-025-07337-yPrediction of failures in the project management knowledge areas using optimized ensemble models in software companiesLamia Berriche0Abderrazak Loulizi1Computer Science Department, Prince Sultan UniversityGraduate School of Management, Management and Science UniversityAbstract Project management involves coordinating various tasks and phases to achieve specific goals. Artificial Intelligence has the potential to enhance decision-making by providing data-driven insights. However, its application in project management, especially for predicting project failures, is still not widespread. This study investigates how hyperparameter tuning and ensemble learning techniques can improve the accuracy of predicting project failures. We applied grid search to optimize several classical machine learning models, including K-Nearest Neighbors, Support Vector Machine, Decision Tree, and Logistic Regression, with average accuracies of 92.46, 93.61, 93.60, and 84.74%, respectively. Our optimized classical machine learning models’ accuracies outperformed state-of-the-art techniques. To further improve performance, ensemble techniques such as soft voting, hard voting, and stacking were explored. Combinatorial optimization was used to identify the most effective set of base learners for ensemble methods. The best soft voting combination—DT, KNN, LR, and CatBoost—achieved an accuracy of 93.21%. The stacking model with DT, KNN, and LR as base-learners and SVM as the meta-learner performed best, reaching 93.73%. CatBoost demonstrated the best accuracy, achieving 94.02%, and demonstrated the best generalization. These models showed strong performance in predicting failures related to scope and cost management but were less accurate when predicting failures in human resource management. Finally, the SHAP model showed that the ‘Reason for Fail’ feature demonstrated the largest impact on the prediction of all classes, while the project manager’s education is an important predictor of cost overruns.https://doi.org/10.1007/s42452-025-07337-yAIMachine learningEnsemble learningHyperparameter tuningProject managementCombinatorial optimization |
| spellingShingle | Lamia Berriche Abderrazak Loulizi Prediction of failures in the project management knowledge areas using optimized ensemble models in software companies Discover Applied Sciences AI Machine learning Ensemble learning Hyperparameter tuning Project management Combinatorial optimization |
| title | Prediction of failures in the project management knowledge areas using optimized ensemble models in software companies |
| title_full | Prediction of failures in the project management knowledge areas using optimized ensemble models in software companies |
| title_fullStr | Prediction of failures in the project management knowledge areas using optimized ensemble models in software companies |
| title_full_unstemmed | Prediction of failures in the project management knowledge areas using optimized ensemble models in software companies |
| title_short | Prediction of failures in the project management knowledge areas using optimized ensemble models in software companies |
| title_sort | prediction of failures in the project management knowledge areas using optimized ensemble models in software companies |
| topic | AI Machine learning Ensemble learning Hyperparameter tuning Project management Combinatorial optimization |
| url | https://doi.org/10.1007/s42452-025-07337-y |
| work_keys_str_mv | AT lamiaberriche predictionoffailuresintheprojectmanagementknowledgeareasusingoptimizedensemblemodelsinsoftwarecompanies AT abderrazakloulizi predictionoffailuresintheprojectmanagementknowledgeareasusingoptimizedensemblemodelsinsoftwarecompanies |