Ant Colony Optimization For Time Series Segmentation And Its Application

Time series segmentation is a vital task in time series analysis, which is the process of segmenting the original series data into some subseries without overlap, and has a wide range of applications space and important research value. Time series segmentation has two major uses. It may be performed in order to determine when the underlying model that created the time series has changed, or to create a high level representation of the time series that supports indexing, clustering and classification. Scholars have proposed many kinds of segmentation algorithms and modified versions, which have been successfully applied to knowledge acquisition of practical problems. But there are few studies about using artificial intelligence algorithms to guide time series segmentation and applying to analyze vehicle traveling data and audio data. Based on the inner characteristic of time series data, this paper uses ant colony optimization algorithm to guide time series segmentation, and applies the segmentation approach to analyze vehicle traveling data and audio data.First of all, according to the inner characteristic of vehicle traveling data, we use ant colony optimization algorithm to guide time series segmentation, and propose a modified segmentation method based on ant colony optimization (ACO) algorithm, which named ant colony optimization based on window updating for time series segmentation (WACO). The strategy not only considers the pheromone of the visited roads, but also regards the pheromone of the potential roads according to the local continuity of time series data.Secondly, considering that the long run time of ACO algorithm limited its application to large scale problems, we present tribe ant colony optimization for time series segmentation (TACO). The approach divides series data into different sub-series at first, according to the idea of divide and conquer. Then each tribe segment the sub-series, finally merge the solution of each tribe to obtain the whole solution. To further enhance the efficiency of ACO segmentation algorithm, we design and realize parallel versions of TACO algorithm for solving large scale problems effectively.Finally, we use ACO segmentation algorithm to resolve the syllable series detection in digital audio data. Using ACO segmentation algorithm to detect the start and end positions of all syllable series, according to the short energy and entropy of audio series data. The promising segmentation results lay the foundation for speech recognition in the future.