[信息与通信]introduction to machine learning research on time sereis 机器学习在时间序列中的介绍.pdf

Introduction to Machine Learning Research on Time Series Umaa Rebbapragada Tufts University Advisor: Carla Brodley 1/29/07 Machine Learning (ML) • Originally a subfield of AI • Extraction of rules and patterns from data sets • Focused on: • Computational complexity • Memory Machine Learning Tasks for Time Series • Classification • Clustering • Semi-supervised learning • Anomaly Detection Assumptions • Univariate time series • Time series databases Single Time Series • A single long time series can be converted into a set of smaller time series by sliding a window incrementally across the time series : • Window length is usually a user-specified parameter. Challenges of Times Series Data • High dimensional • Voluminous • Requires fast technique Brute Force Similarity Search • Given query time series Q, the best match by sequential scanning is found by: • O(nd) • Finding the nearest neighbor for each time series in the database is prohibitive. Similarity Search • Clustering and classification methods perform many similarity calculations • Some require storage of the k nearest neighbors of each data instance • Critical that these calculations be fast Speeding up Similarity Search • Alternate time series representations • Search databases faster • New similarity metrics Data Mining Time Series Toolbox • Indexing • Dimensionality Reduction • Segmentation • Discretization • Similarity metric Indexing • Faster than a sequential scan • Insertions and deletions do not require rebuilding the entire index • Partition the dat