Cheng-Church (CC) biclustering algorithm is the popular algorithm for the gene expression data mining at present. Only find one biclustering can be found at one time and the biclustering that overlap each other can hardly be found when using this algorithm. This article puts forward a modified algorithm for the gene expression data mining that uses the middle biclustering result to conduct the randomization process, digging up more eligible biclustering data. It also raised a parallel computing method that uses the multi-core processor or cluster environment to improve efficiency. It is proved by experimental verification that the modified algorithm enhances the precision and efficiency of the gene expression data mining to a certain degree.