The Research Of Conditional Discriminative Sequential Pattern Mining Algorithm

Discriminative sequential pattern mining is one of the important topics in patterns mining,and it has a very wide range of applications.Discriminative sequential pattern mining is intended to excavate sequential patterns with significant differences from sequence data with class labels.In recent years,a variety of algorithms on discriminaive sequential pattern mining have been proposed,but the research on the redundancy of the reported pattern set is lacking.Discriminative sequence pattern mining is divided into two categories,namely,thresholdbased discriminaive sequential patterns mining and top-k condtional discriminative sequential patterns mining.There are always some redundant sequential patterns in the mining result excavated by the existing methods.Redundancy is an urgent problem to be solved in discriminative sequential pattern mining.There are many reasons that can lead to the redundancy of reported patterns,among which the subset-induced redundancy is the most critical one,i,e.,super-patterns of some significant sub-patterns can be significant discriminative sequential patterns as well.In order to solve the subset-induced redundancy issue,we proposes the concept of conditional contrast,and proposes a new data mining problem based on this concept,that is,conditional discriminaive sequential pattern mining.In this paper,the conditional discriminative sequential pattern mining is divided into two categories:(1)threshold-based conditional discriminaive sequential patterns mining;(2)top-k conditional discriminaive sequential patterns mining.We propose the CDSPM algorithm and TKCDS algorithm for these two types of data mining problems.The CDSPM algorithm is for thresholdbased conditional discriminaive sequential pattern mining.The TKCDS algorithm is used for solving the top-k conditional discriminaive sequential pattern mining problem.Experiments show that conditional contrast can well eliminate the influence of the subpatterns and remove redundant patterns.Using CDSPM algorithm and TKCDS algorithm can efficiently excavate the conditional discriminaive sequential patterns,and can filter out a large number of redundant sequential patterns.Thus,in practical applications CDSPM and TKCDS are of considerable value.