The main goals to establish mathematical models for map generalization lie to describe rules which maps are generalized, to provide selection index for quantity of land feature, and then to supervise processes for mapping workflow. This research adopts the mathematical statistic methods to indicate objective regular patterns of land features by means of statistical observations, induction and analysis. This article takes the map geographic name as the research object, (north, south, central, east) four parts of Taiwan as study areas, and respectively chooses a county as the resources of sample data. They are: Taoyuan county, Taichung county, Tainan county and Hualien County. According to the economic construction version of topographic maps published by Ministry of Interior, we measured on the different scale topographic maps (scaled from 1/25,000~1/50,000) to calculate the quantity f geographic names. We also used the Excel software and carried on tests to the curve fitting processes in order to decide the suitable regression model. The findings show that geographic name retention rate (Y) and map scale factor (X) have regressive relationship functioned as (Y=aX[8f7a]) and that map scale to geographic name retention rate has high explanation strength. The results offer practical reference values to the selection of geographic names regarding mapping technique.