Enhancing the Frequency of Temporal Patterns Through Special Constraints

Abstract

In the field of temporal data mining, analysing the temporal behavior of frequent patterns, obtaining the accurate clusters, and reducing the size of identified patterns are three key issues. In this regard, several methods are available among them, the constraint based pattern mining approach has made a significant contribution. However, existing approaches for investigating patterns based on time granularities use the traditional Apriori or Interleaved algorithms, which require a long time to generate candidates. We address these issues and propose the Frequent Pattern Growth method Incorporated with Special Constraints. The framework is comprised of a frequent pattern generation approach for which complete cyclic and partial cyclic constraints are enforced for transactional datasets. To locate all eventual periodic patterns from the Spatio temporal dataset, we have formulated the problem as periodic pattern mining from the Spatio temporal database in the current thesis. The proposed method employs a periodic pattern tree miner. Initially, to produce the optimal dense clusters, a novel global pollination artificial fish swarm strategy is used in the clustering process, which combines the global flower pollination algorithm with the artificial fish swarm algorithm. Further, partial cyclic, complete cyclic, and probability constraints employed in the proposed model limit the uninteresting patterns. To improve the execution time and memory usage, further, the concept of Artificial Neural Network, with a football game based optimization method is proposed in the thesis. Initially, this method uses the teaching learning based optimization algorithm for clustering, and then Artificial Neural Network based on football game based optimization method is applied to obtain periodic patterns. Nine constraints, namely item, value, cyclic, aggregate, length, sequence, similarity, duration, and gap are employed to enhance the frequency of patterns by limiting the unwanted patterns. newline newline newline