Clustering analysis is the basic of data mining, and K-means algorithm is the simplest clustering algorithm. However, traditional K-means algorithm has many defects-instable K value determinations, non-universal applicable SSE etc. Consequently, we introduced an improved K-means algorithm basing on the clustering reliability analysis. The algorithm effectively solves the problem on uneven density and large differences in the amount of data clustering.

No abstract

.For the purpose of showing the importance of certain items in the negative sequences and make negative sequences have more practicality, this paper presents a method for mining weighted negative sequence pattern. Set the weight value to calculate the weighted support and prune the weighted negative sequences which don’t meet conditions. We use IRIS data set which belongs to UCI data sets to verify the new algorithm. Comparing with Neg-GSP algorithm, showing the benefit of the weighted concept.

With the deepening of the negative association rules mining technology research, many key problems have been solved, but the solution of these problems are all on a single predicate in the transaction database. However, the data in the database often involves multiple predicates. This paper focuses on solving multi-dimensional support and confidence, negative association rules mining algorithm design problems. The experiment proves that the algorithm is correct and efficiency.