Time Series Search Based on Locality Sensitive Hashing
Mengru Zhang · 2021 3rd International Conference on Applied Machine Learning (ICAML) · 2021
The application of similarity search in large-scale data of time series is very common. It is also the main subroutine of the time series data mining algorithm. The representation of time series and complex similarity measures are the basis of time series similarity research and play a vital role in completing the task of time series similarity search. Therefore, the efficiency of time series similarity research is a serious obstacle to the development of time series mining algorithms. The traditional branch and bound method can prune the candidate set before the similarity search to achieve acceleration, but this method is only effective for low-dimensional time series. When the dimensionality of the data increases, the complexity of the algorithm increases exponentially. The performance of the algorithm will drop sharply, which is similar to a brute force algorithm in the end. In this work, we introduce a novel algorithm LSH (Locality Sensitive Hashing) for similarity search of time series subsequences. Our results show that it is effective and concise to quickly approximate longer time series, and improves the search efficiency.