| Summary: | 碩士 === 國立高雄應用科技大學 === 土木工程與防災科技研究所 === 99 === This research collects tendering data for a total of 1,673 construction projects from Ccentral Government, Kaohsiung City and Taitung County. The final bidding price for construction is designated as the dependent variable and another 5 factors of budget amount, bidding bond, base price, number of bids and performance period are assigned as the independent variables. Pearson's Correlation Coefficient Analysis and the Whole Batch Least Squares Estimation Method are used to examine the relationships between each dependent variable and independent variables. Afterwards, the most highly associated independent variables are feed into the Whole Batch Least Squares Estimation Method as the prediction model inputs.
The Pearson Correlation Coefficient and the Whole Batch Least Squares Estimation Method analysis results show that three out of the five factors are highly associated with the final bidding price. These three factors are: Basic Price、Budget Amount and Bidding Bond. As a result, these three factors are used to predict the Final Bidding Price for this research.
The analysis results show that when the power of LSE model is higher, the variance obtained is smaller, and thus make the estimation results more accurate. However, for projects with smaller final bidding price, the prediction errors are larger. Through multiple regression analysis, it is found that the related error variance (prediction accuracy) is lower for the quadratic non-linear Whole Batch Least Squares Estimation.
The research results have shown that Pearson Correlation Coefficient analysis is efficient in selecting crucial factors influencing final bidding price for the prediction model. When projects are divided into several groups according to their total cost, the Whole Batch Least Squares Estimation model is able to produce satisfactory prediction results. This provides valuable information to government officials when assessing the legitimacy of the actual bidding price. In the mean time, the contractors are able to use the model with only two inputs, Budget Amount, and Bidding Bond to generate moderate prediction results for their reference when making the final decition before tendering.
|
|---|
