Incremental Updated Algorithm For Weighted Negative Sequential Patterns

In recent years, with the rapid development of Internet and Database, Database storage generates amounts of data in many fields, data mining is replacing traditional traditional data processing tools to help us obtain some useful rules from a large number of data, predict and analyze potential rules. The sequential pattern mining is an essential branch of data mining, its theoretical value is being applied to practical problems, its prospects must be very extensive.Sequential pattern mining can be divided into positive and negative sequential pattern mining. At present, the positive sequential pattern mining is relatively mature, related scholars have presented weighted positive sequential pattern mining algorithms and incremental update algorithms according to the actual situation. But the negative sequential pattern started late and high difficulty in mining negative sequences. Therefore, there are less relevant algorithms and it is remain in the theoretical stage.Traditional negative sequential pattern mining algorithms don’t distinguish items important in negative sequence and ignore efficiency of mining negative sequences after sequence database frequently updates. In order to solve the two questions, this paper gives an incremental update algorithm for weighted negative sequential pattern. It allows users to set different weights for each item according to actual situations and makes full use of the previous mining results and discusses the combination conditions of weighted frequent and non-frequent item sets when databases update. It improves efficiency of mining weighted negative sequential patterns when databases update and solves the above two questions. If set item weights of sequence database as 1, this algorithm will become an efficient incremental update algorithm for negative sequential patterns. We design and compile the algorithm according to relevant mining theories. Use official UCI data sets to record numbers of mining negative sequences and running time. From the experimental results we show that the algorithm is feasible.