| Summary: | 碩士 === 中原大學 === 應用數學研究所 === 102 === Abstract
In this thesis, we propose a bias-correction fuzzy regression algorithm. The proposed algorithm can improve the most-used fuzzy c-regression (FCR) method. In FCR, it has the same drawbacks as fuzzy c-means (FCM), where the FCR and FCM algorithms are always affected by initializations. This is because the FCR is an embedding of FCM into switching regressions, so that it has still the same drawbacks as FCM. In 2008, Yang et al. proposed the so-called alpha-cut fuzzy regression to improve FCR. Recently, Yang and Tian proposed an improving method of FCM, called bias-correction FCM (BFCM). In this paper, we propose the bias-correction fuzzy regression algorithms (BFCR) by embedding the BFCM into switching regressions. Several examples are used to compare the proposed BFCR algorithm with FCR and alpha-cut fuzzy regression. The comparison results demonstrate the superiority and usefulness of the proposed BFCR.
|
|---|
