The research into an improved algorithm of telecommunication inter-transactional association rules based on time series of all confidence

Abstract

The telecommunication network has a large scale and an intense complexity. Agents distributed over diverse network elements have collected an immense number of KPI data, the key indicators of network performance. These time series data can have mutual impact. This paper puts forward an improved algorithm named AFP-Growth to mine association rules of inter-transaction time series in the telecommunication field. Based on improvements of the conventional FP-Growth algorithm without Conditional sub-tree Generation, this algorithm has introduced a new correlation measure, that is, all confidence, thus resolving the problems of null-transaction and negative correlation in mining telecommunication data. In addition, by utilizing the features of all confidence, this algorithm has improved the pruning rule of FP-Tree, and enhanced the effectiveness of FP-Tree search, thus increasing the time and space efficiency.