Artificial neural network (ANN)‎ for estimating of overhead cost for school construction projects Gaza Strip

العناوين الأخرى

استخدام الشبكات العصبية الاصطناعية لتقدير التكاليف الإدارية لمشاريع المدارس-قطاع غزة

مقدم أطروحة جامعية

Awad, Rayya Muhammad Ahmad

مشرف أطروحة جامعية

al-Hallaq, Khalid Abd al-Rauf

الجامعة

الجامعة الإسلامية

الكلية

كلية الهندسة

القسم الأكاديمي

قسم الهندسة المدنية

دولة الجامعة

فلسطين (قطاع غزة)

الدرجة العلمية

ماجستير

تاريخ الدرجة العلمية

2017

الملخص الإنجليزي

Purpose: There are two main objectives of this research.

First is to determine the affect factor on OH.

Second is to establish a Neural Network Model that will provide any construction firm the ability to assess its overhead costs for any school project.

This may improve the construction industry‘s performance and the ability to overcome the national and international market difficulties.

Methodology: The research generates a questionnaire to select the top ten factors that affect the construction market in Gaza Strip and develop and test the model using the artificial neural network (ANN) technique.

Matlab R2013a Software was chosen to generate the Model for predicting the percentage of school projects overhead costs from the total projects costs.

This model consists of an input layer with eight input neurons, and one hidden layer with twenty neurons and one output neuron.

Data on 70 real-life school construction projects from Gaza were used in the training and validation processes.

To verify the generalization ability of the best model, testing with 11 projects (facts) that were still unseen by the network was performed.

Results: The top ten factors that result from questionnaire analysis, are company's experience, closure and the inability to obtain materials, intensity of competition from other contractors, number of projects, existence of documentation for implemented projects, management system for overhead cost, project size, mechanism of company financial dues (payments), firms need for work, and economic inflation.

The selected model has 20 hidden neurons, where MSE equal 0 and R = 1 for training phase, MSE equal 0.13 and R = 0.989 for Validation phase and MSE equal 0.13 and R = 0.987 for test phase.

The performed sensitivity analysis shows that the firm need for work, existence of documentation for implemented projects, No.

of similar projects in the same year and contract amount, have significant influence on the output of the network.

Recommendation: The model is a simple and very easy-to-use tool that can help contractors/firms during the consideration of the influential overhead cost variables and to improve the consistency of the percentage of overhead costs decision-making process.

In addition, it encourages all parties involved in construction industry to pay more attention for developing ANN in cost estimation by archiving all projects data, and conducting more studies and workshops to obtain maximum advantage of this new approach and join more outputs in a model

التخصصات الرئيسية

الهندسة المدنية

عدد الصفحات

91

قائمة المحتويات

• APA
• MLA
• AMA

لغة النص

الإنجليزية

نوع البيانات

رسائل جامعية

رقم السجل

BIM-904767