Fuzzy Linguistic Approach and Discrete Choice Theory for Building Choice Behavior Model in Household Purchase

• Main Authors: Ching-Yu Lein, 連經宇
• Other Authors: Yen-Jong Chen
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/53926931180667274204

博士 === 國立成功大學 === 都市計劃學系碩博士班 === 91 === 　　A household''s decision for housing choice is always a critical issue. During households'' housing choice behavior, it includes a very complex and uncertain process. It is very difficult to get certain information on households'' housing choice behavior. Understanding the household''s preference and establishing the evaluating criteria would help the government to formulate housing policies and the housing developers to provide housing products. The major topic of this study will be focused on the comparison of different methodologies. The main purpose of this study is to purpose a new micro-choice behavior model in household purchase and to depict the reality of human subjective perception behavior. We hope to improve the Logit model in dealing with many uncertain and fuzzy situations about housing choice behavior in household purchase. A new Fuzzy Linguistic Logit Model (FLLM) is proposed in this study to estate the probability of housing choice according to the Logit models of discrete choice theory combined with the well-developed Fuzzy Linguistic Scale (FLS) approach. 　　In empirical analysis, this study will give instances for the effects of housing location and type choice behavior in household purchase. The choice indexes will be measured to a household''s decision for housing location and type choice behavior by means of related literature reviews. This study will also design two questionnaires for both Likert scale and FLS in order to extract the weights of both certain and fuzzy attribute variables and to construct fuzzy utility function by using correlation analysis, factor analysis and reliability analysis. A new FLLM is established by applying the binary, multinomial and nested choice Logit models as well as FLS for location and type choice in household purchase. Then this study will verify and compare whether the empirical results of those three FLLMs are better or not. We selected data from a housing survey in Tainan city of the southern Taiwan for the empirical case study from the beginning of 1998 to the end of 1999. 　　The research results show the following two parts. First, in the comparison of different methodologies: the first finding of this study is that after comparing the different models, in addition to binary choice models, the goodness-of-fit, success rate of forecasting, expected demand elasticity and likelihood statistic test in FLLM are all better than those in Logit model, especially in the nested fuzzy Logit model (FNMNL). The second finding of this study is that the new model is proposed to be more capable of dealing with the problem of qualitative variables, which is one of the critical issues in quantitative approaches. Second, in the empirical results: the first finding of this study is that in the different comparison of FLS, triangular fuzzy numbers of some linguistic types have several overlapping situations. Although above situations are significantly improved after raising the numerical value of Identity Degree Measure Function (IDMF) to process the defuzzification, this result cannot reflect the real human behavior due to the samples decreased substantially. The implication of this result verifies that the human subjective perception has the fuzzy characteristics. In addition, the definitions of various types of fuzzy number are different and unsymmetrical. For the households of single and multiple housing, the opinions of ordinary people for relatively negative linguistic terms show greater differences, whereas those for more positive linguistic terms are usually more similar. The second finding of this study is that the household disposable income and housing space are very significant. Except for five alternatives of the households of multiple housing, the degree of satisfaction for neighborhood zone is very significant. Finally, in the estimation results of expected demand elasticity, the change scope of elasticity value for FLLM is less than that of Logit model. It indicates that the degree of concentration has been increased.