| 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 === 100 === After the introduction of the "procurement regulations" since 1999, most of the public construction projects are awarded to the lowest bidder. This form of tendering practice often provokes fierce competition among the bidders and the winning bids are sometimes so low that the quality of the construction project is in jeopardy. As a result, it is crucial to develop a final bid price prediction model to assist with the evaluation of the final bid price.
This research collects information from Central Government public construction projects during the past 5 years for analysis. It is observed that 93% of the collected samples are between 1 Million and 100 Million NTD. These samples are grouped into 3 categories for further prediction analysis. Budget amount, project bond and project base price are set as independent variables and the final bid price is set as the dependent variable for the prediction models. Artificial Neural Networks and Support Vector Machines are adopted to develop the prediction models and the prediction results are compared between those two models.
The results show that in the 1M~5M and 10M~100M categories, support vector machine models render better prediction results (10.56% and 9.15% MAEP) when compared to ANNs models (10.79% and 9.35% respectively). In the 5M~10M category, the MAEP for ANNs model (10.34%) is slightly better than that of SVMs model (10.42%). The prediction model is able to assist with the Government Agencies when evaluating if certain bid price is unreasonably low.
|
|---|
