Comparative Study On Mining Sequential Patterns With Periodic Wildcard Gaps

Data Mining is the most popular research field at present.It mines implicit,previously unknown relationships,patterns and trends that have potential value for making decision from a large amount of data.Currently,it have produced a large number of sequence data in many application fields.How to mine potential and valuable patterns from these sequence data becomes a major task now days.Given a sequence S,a certain threshold and gap constraints,the problem of mining sequence patterns with one-off condition is that all the number of frequent patterns that are extracted from the given sequence S is not less than the given threshold and each adjacent character in the frequent patterns can be used at most once,namely one-off.Under the condition of mining sequence patterns with one-off,any support of super patterns are not greater than the sub patterns,therefore this problem is a mining method of meeting the nature of Apriori,and the calculation of a support of pattern problem is a pattern matching problem.Moreover,under the condition of mining sequence patterns with one-off,calculating the maximum number of occurrences of patterns in a given sequence is an NP-Hard problem,this paper adopts the Nettree structure and greedy search strategy to calculate the support of pattern,and then put forward the SBOM algorithm to achieve a sequence pattern mining with one-off condition.Experiments are carried out on real biological data.we compared with the One-off Ming algorithm and the results show that the SBOM algorithm can mine more frequent patterns,but the runtime is long.In addition,a large number of experiments are carried out on compareing this paper with mining sequence patterns with periodic wildcard gaps,and the runtime when the result of sequence mining is close are compared and analyzed.Through comparison and analysis of the mining results,we find some transformation relations of the two thresholds on mining sequence patterns with one-off condition and periodic wildcard gap,and conclude under which circumstances is suitable for the mining sequence patterns with one-off,while under which circumstances is suitable for another one.