UNB 2012 intrusion detection evaluation dataset was provided from CIC (Canadian Institude for Cybersecurity) of new brunswick university in Canada. The dataset is created by simulation considering real network intrusion. Many researches related to intrusion detection are conducted, but it is difficult to apply most of the results of the researches to real intrusion detection system. The CIC made the UNB 2012 intrusion detection evaluation dataset considering the difficulty. In this paper, we use a normal class and four attack classes. The dataset has several sub dataset separated by date. Each sub dataset has their own characteristics. We select or modify 16 features among common features of all sub dataset. We conduct comparative experiments after data partitioning, undersampling, and modeling using WEKA(Waikato Environment for Knowledge Analysis), a typical data mining tool. In the experimental results, the k-NN algorithm showed the best performance.