The marketing battle among Korean telecom companies to secure new subscribers has been so fierce that their marketing expense now exceeds operating profit by more than double. The cost of this battle, however, is excessive since it is nothing but a zero-sum game, which each participant’s gain or loss of utility is exactly balanced by the losses or gains of the utility of the other participants. The general marketing strategy of telecom companies is not effective since the companies mainly focus on taking other companies’ customers and do not consider maintaining own subscribers. By increasing the accuracy of measuring termination rate of existing broadband internet users, telecom companies can be benefited saving tremendous amount of money. The purpose of this study is to identify the possibility of forecasting the Broadband internet termination rate by using time series analysis and to set the optimal region units that have condition in which the termination rate can be accurately predicted. The 63,289 termination data of A company were collected from 47 units of administrative dongs, which is similar concept to town U.S., in city of Anyang, Kunpo and Uiwang. These are located in kyonggi-do, Korea. The unit for prediction and verification of termination rate were divided by administrative districts and spatial clustered dongs, and measuring accurate termination rate was better in order of cluster, district, and total unit. Overall, the results revealed that extracting data from spatial cluster is significant in predicting the termination rate. This shows that it may be effective to predict and implement the termination rate in unit of cluster when telecom companies use marketing strategy in different regions.