With the trend of financial deregulation, globalization, and the change of financial environment after Taiwan's entrance into WTO, the first task of the domestic banks is the improvement of service quality. This research takes customers of domestic banks (totally 30 Banks, including government owned banks, old privately operated banks, and new privately operated banks) in Taiwan area as the target of this study, and uses 23 items of service attribute measures of PZB model to ask the customers about their recognition of perceived and expected service quality of the banks they deal with. Our conclusions are as follows: Segmenting all Bank customers by two methods: cluster analysis and type of bank, we found that the former method shows significant difference for all of the four dimensions extracted by factor analysis in both discriminating validity and ANOVA analysis Comparatively, in ANOVA analysis which segmented the banks by their operation type, we found that the differences between all banks only exist in the factor of convenience level, but there is no significant difference between old and new privately operated banks. So it suggests that Taiwan resident's consumption behavior of bank service is not the same as those in different area. We name the customer groups discriminated by cluster analysis as high, middle, and low-service-quality recognition groups respectively and then compare the individual bank's average score of four factors, finally categorize all banks into high, middle, and low-service-quality groups. Banks belonging to low-service-quality group are all new ones. Five out of eight old banks are middle-service-quality group. The remaining three are of high service quality. Chi-Square test is employed to exam the goodness of fit of customers' gender, marriage, age, education, income, job, interactive time and frequency with banks, and we found that both two methods of segmentation are not significantly different in sex, but are significantly different in marriage and age. Also, we found that the groups categorized by operation type show significant difference in education, income and job, but no significant difference for the groups segmented by cluster analysis.