Back to results list
Show full item record
Please use this identifier to cite or link to this item:
|Title:||Improving the long-term use of case-based reasoning model in early construction cost estimation||Authors:||Xiao, Xue||Degree:||Ph.D.||Issue Date:||2021||Abstract:||Construction cost estimation is a significant task because it provides valuable financial information for project decision making. In the early stage, the estimate is typically used to conduct initial feasibility studies. Since the flexibility to adjust the project scope, design, specification, and standards needs to be very high in this stage, construction cost estimation should be made as early as possible. However, it is irrational for construction companies to over-invest their design time and effort in the early stage with limited human resources. An estimator's effort will be more valuable in the later estimation stages. Many researchers therefore have explored ECCE using various techniques. These techniques, which turn out to be helpful in assisting ECCE, are based on the historical database containing previous similar projects and the target project. By its very nature, early stage estimation involves a large amount of subjectivity, making the estimator's experience of vital importance. This makes case-based reasoning (CBR) an obvious candidate, as it not only includes such basic components as a reasoning cycle and case-base, but also relies on experience or knowledge - making it a perfect match compared with other methods. Most importantly, it is also advantageous for long-term use. Since data growth has altered the way information is stored and processed, the continuously increasing amount of construction cost information facilitates the amount of data available. This fact creates huge opportunities and challenges for using the CBR model in ECCE. On the one hand, data growth greatly enriches the knowledge and experience in case-base, improving its overall performance. On the other hand, the historical database will continually increase over time as more data is added to it, resulting in a high requirement for updating and maintaining the ability of the case-base. In particular, the changed resource costs, construction methods, design styles and economic conditions create outdated and inconsistent data, which should be carefully handled and eliminated. Without proper handling, these stale data will impair the performance of each component: the typical issues being the unstable knowledge structure and low efficiency because of the continuously increasing size of the case-base during long-term use. This inevitably raises the problem between the benefits of having more data and the deficiencies of having inappropriate data.
This study therefore aims to improve the practice of long-term use of case-based reasoning in ECCE from several perspectives: understanding the parameter settings of CBR in the existing research and the case-base's influence on accuracy; improving the robustness of the CBR system; and enhancing the efficiency of the CBR system. It firstly identifies which parameter combinations are better and explores the influence the size of the case-base has on the performance of the CBR model. Then, a robustness weight determination method is introduced to improve the robustness of the ECCE CBR model, followed by an original case-base maintenance method based on weight coverage contribution to improve the efficiency of the reasoning process. The results indicate that the GA-CBR model is more accurate when the size of the case-base is small and there is no significant difference in accuracy between the MRA-CBR model and the GA-CBR model when the size of the case-base becomes large. However, GA does not have the advantage of producing a stable structure in the case-base during long-term use, while the MODAL-CBR model effectively improves the robustness and accuracy of the results. The CBM methods are classified into three strategies according to how the case-base and weight determination are selected. Strategy 1 and Strategy 2 can significantly compress the size of case-base for all CBR models. Strategy 1 generates better results in OLS-CBR while Strategy 2 generates better results in MODAL-CBR. More specifically, Strategy 1 can maintain the OLS-CBR's performance while reducing the size of case-base by 28.13%. There is also a slight improvement in the accuracy of OLS-CBR models when the size of the case-base is slightly reduced. The study provides valuable knowledge for improving the ECCE CBR model for long-term use, both theoretically and practically. The methods used not only help improve the performance of the CBR system by enhancing the robustness of generating a stable case-base knowledge structure, but also help maintain the efficiency of the case-base during long-term use. Valuable ideas are also provided of how future work can be conducted to improve the ECCE CBR model in the future.
|Subjects:||Building -- Estimates
Construction industry -- Management
Hong Kong Polytechnic University -- Dissertations
|Pages:||xvi, 243 pages : color illustrations|
|Appears in Collections:||Thesis|
View full-text via https://theses.lib.polyu.edu.hk/handle/200/11018
Citations as of Jul 3, 2022
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.