Data mining is a process of discovering useful information from a data set. in data mining, there is a classification tehcnique that depends on sampling accuracy to acquire a more accurate result in data classification or prediction. therefore, a necessity in getting a good-quality sampling is required. the primary purpose of this research paper is to obtain the optimum sampling representing the original data set. through sampling, we could minimize the total data that need to be processed. because large amount of data requires longer processing time, reducing the amount of data with sampling will speed up the process of computing. in this study we introduced a new sampling algorithm with clustering approach applied to a network security data set. preliminary results showed that proposed method offer fine result fo large data set sampling.