SAM-BiLSTM-based time sequence classification method

Abstract

The invention discloses a time sequence classification method based on SAM-BiLSTM. The method comprises the following steps of obtaining a one-dimensional time sequence data set and performing normalization processing on the one-dimensional time sequence data set; constructing a classification network model based on the SAM-BiLSTM time sequence; and through the trained SAM-BiLSTM-based time sequence classification model, carrying out feature extraction on the to-be-detected one-dimensional time sequence data, and outputting a time sequence classification result. According to the method, compared with an existing one-dimensional time sequence classification model, the accuracy is higher.

Description

technical field [0001] The invention relates to the field of data mining technology and deep learning of time series, in particular to a time series classification method based on SAM-BiLSTM. Background technique [0002] Time series widely exist in fields such as healthcare, electricity, and finance. The classification of time series is a very important basic task, which can be achieved by calculating the similarity between time series. Dynamic Time Warping (DTW) distance is widely regarded as the best similarity measure for time series. It uses a dynamic programming algorithm to determine the best alignment, taking into account temporal offsets, scaling, and deformation. But DTW distance has two limitations. First, the time complexity of dynamic programming algorithms is quadratic of the length of the time series, making the scale of classification and retrieval unsuitable for large datasets. Second, the DTW distance is a pseudo-metric because it does not satisfy the t...

AI Technical Summary

Problems solved by technology

First, the time complexity of the dynamic programming algorithm is the quadratic of the length of the time series, making the scale of classification and retrieval unable to adapt to large data sets

Second, the DTW distance is a pseudo-metric because it does not satisfy the triangle inequality

When combined with kernel-based classifiers or traditional index optimization techniques, the optimal solution cannot be obtained

Images