| The construction project is a typical complex system engineering.Due to its multiple construction links,long construction period and huge investment,plenty of studies have shown that the occurrence of construction project rework has a great impact on the construction period,cost,safety,etc.However,the current research and development lacks a complete rework risk early warning system,which has little effect on improving performance of the project,and reworks are often resolved after occurrence.Therefore,the purpose of this study is to build an effective construction engineering rework prediction model through which project managers can take measures in advance before rework occurs to curb the occurrence of rework or reduce losses.In this study,reasons of rework were obtained by reviewing round-trip engineering literature research and interviews with experts in the construction industry.Based on this,a questionnaire was designed,and the engineering managers completed the questionnaires face to face and a total of 330 sub-branch projects rework data were obtained.Then three algorithms were applied(Support Vector Machine,XGBoost,and Naive Bayes)to construct predictive models based on the acquired data,and evaluate the model results.The Support Vector Machine rework prediction model has the best prediction performance,with an accuracy rate of 95.9% and a recall rate of 90.9%.And based on the factor importance ranking of the XGBoost model,it is concluded that the project manager’s work experience has the greatest impact on whether rework occurs.The thesis sorts out the reasons for rework in the whole process of construction projects,deepens the control points of process control,and expands the theoretical system and methods of construction engineering risk management.The construction project rework prediction model obtained in this study enriches the construction process management system of the construction project,and has application value for improving project profitability and improving project quality. |
