A Novel Time Series Pattern Matching Model Combined with Ant Colony Optimization and Optimal Binary Search Trees Based Segmentation Approach

摘要

Time series data are a sequence of values or events obtained over repeated measurements of time and these are hefty in size, dynamic in nature. Applications comprising time series data include economic and sales forecasting, utility studies, the observations of natural phenomena suchas atmosphere, temperature and wind, coal mine surveillance, multimedia data retrieval etc. This paper accords a new pattern matching model for time series data combined with Ant Colony Optimization (ACO). Segmentation is the pre processing step of the pattern matching process. In this paperwe propose a segmentation method based on mean values of the time series data. The known patterns are exemplified as Optimal Binary Search Trees (OBST) based on their mean values. Since searching is efficient in OBST, this technique is efficient for pattern matching in time series data. TheACO is used for comparing a particular pattern and the current window of the time series data. Computational results show that our approach outpaces other existing models and harvests a very accurate matching at minimum cost.