Mark-up estimation using data mining techniques

In this study, mark-up estimation using data mining techniques in constructionindustry was investigated.Firstly, a literature survey was performed about characteristics of mark-up estimationand techniques that have been used in mark-up estimation and data miningtechniques that were used in different areas.Then, a questionnaire was distributed to civil engineers that study in bid departmentsof public sectors in order to determine the factors that affect mark-up estimation. Thequestionnaire?s results were evaluated using content analysis method and targetmark-up factors were determined. According to target mark-up factors, data werecollected in construction bulletins in Turkey.Lastly, mark-up estimation was analysed according to rule extraction from trainedneural network using genetic algorithm and decision tree. A neural network programthat was developed in Matlab and Evolver 4.0 Professional for genetic algorithm andSee5/C5.0 for decision tree were used. In the rule extraction, after data wereclassified by back propagation, rules were extracted from trained neural networkusing genetic algorithm. In decision tree, the C5.0 algorithm has generated aclassification decision tree for the given data set by recourse portioning of data. Theknowledge represented in decision trees were extracted and represented in the formof classification IF-THEN rules.Keywords: Data mining, mark-up estimation, neural network, decision tree, geneticalgorithm