1, sequence segmentation is to better handle the sequence. It is inconvenient to extract key information (trends, averages, etc.). ) When a sequence that is too long is processed accordingly, and the information description of the sequence is inaccurate.

2. There are many algorithms in this field. What I remember are average segmentation, segmentation based on maximum and minimum, segmentation based on key points, segmentation based on sliding window and so on. The specific realization of the algorithm, you just find some information on the Internet.

In learning materials, you should first look at some basic things, such as some common methods of time series, the difference between univariate series and multivariate series, and so on. Of course, you should also know some basic characteristics of time series, such as obeying normal distribution (in fact, these are all mentioned in the review papers related to time series, so just read more papers). Then, learn from monism to pluralism, such as DTW correlation algorithm, hierarchical algorithm, time-granularity correlation algorithm, singular value decomposition, principal component analysis and so on. , and read more papers on the corresponding algorithm first, and then find some corresponding programs on the Internet.