عنوان رسالة الماجستير :

Predicting construction labour productivity using optimal artificial neural network, multiple linear regression models : comparative study

📅السنة: 2018-06-13

🎓التخصص: ماجستير إدارة المشاريع الهندسية

👨‍🏫المشرف الرئيسي: فائق محمد سرحان سرحان | Faiq M. Al-Zwainy

🧑‍🎓الطالب: حمزة محمد الأعمر | Hamzeh M. Al-A'amar

---

📖 الملخص (Abstract):

Construction productivity can be considered as an element in project management; therefore, predicting the rate of construction productivity for labor was an important task. However, the development of the technology tools will enable the planner to best understand the process of estimation and predicting in different stages of construction projects. The main aim of this research is to develop a novel mathematical model using Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) to predict the construction productivity rates because mathematical models and mathematical equations used for finishing stone activity are characterized by uncertainty and lack validity and verification, and traditional methods fail to calculate the construction productivity due to their slowness and lack of accuracy. Data was collected from three residential building projects in the Hashemite Kingdom of Jordan in the capital city of Amman from July 2017 to December 2017. The first project was Tebyeh Residential Building (TRB); the second project was Sinokrot Private Villa (SPV); and the third project was Aldada Private Villa (APV). The results demonstrated that (MLR) is a more powerful technique than (ANN) for construction productivity of finishing stone activity depending on validity through Mean Absolute Percentage Error (MAPE%) and Average Accuracy (AA%), which were equal to 18.615% and 81.3846% respectively; ANN technique (MAPE%) was equal to 27.06 % and (AA%) was equal to 72.94%. This result can be expressed when using multiple linear regression techniques instead of artificial neural networks in estimating and predicting construction productivity when the data of the variables are homogeneous; otherwise, use of artificial neural networks technique is preferable.

📚 الأبحاث المستلة: