Abstract

Time series classification is an important task in time series analysis. Thus, many methods have been developed for the task. However, the quality of features is difficult to measure and there is no distance measurement method for most areas. And these methods cannot extract the long-term dependency feature from time series. In order to solve these problems, we propose a new time series classification model, Long short-term memory networks and Convolution Neural Networks (LCNN). First, the model can automatically extract features from the time series. Second, LCNN solves the long-term dependence problem by introducing Long short-term memory networks (LSTM) into time series classification tasks. Third, LCNN adopts multi-branch structure to down-sampling and Gaussian noise to process the original time series, which improves the classification performance. In addition, we use the Hurst exponent to measure the long-term dependency in time series. All experiments show that LCNN improves the classification performance and is well suited for small datasets.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call