Hybrid BERT-XGBoost Model for High-Impact Research Prediction in Construction Management

Some of the sub-disciplines that make up CEM include scheduling, quality control, human aspects, sustainability, and cost estimation. Many researchers have turned to scientometric analysis to gauge the relative merits of various research philosophies. To find seminal studies and direct future research, high-impact research prediction is essential in the field of construction management. Feature selection was carried out using the ReliefF Method, and the study makes use of other data preparation approaches, including missing value imputation, data reduction, and data transformation. The incorporation of a BERT-based CM extraction method is a significant breakthrough in this methodology. It outperforms existing techniques when it comes to finding obfuscation, encryption, and evasion strategies. To further improve the accuracy of predictions, the BE-XGBoost classifier is also presented. With a recall of 93.30%, an F1-score of 97.43%, an accuracy of 97.38%, and a precision of 95.52%, the experimental findings show that the suggested model performs better than prior models. The results show that the model is good at predicting research with a big impact in the field of construction management. This study lays the groundwork for future advancements in research prediction procedures by integrating state-of-the-art extraction and classification techniques, which is a substantial contribution.