Identifying the Head-and-Shoulders Pattern Using Financial Key Points and Its Application in Consumer Electronic Stocks

This paper constructs a new time series segmentation method using the financial implications of the head-and-shoulders pattern (HS), effectively overcoming the limitations of traditional time series segmentation methods that ignore the economic meaning of the data itself in time series mining. On this basis, this paper proposes an extremum point sliding window method to extract subsequences from longer time series. Unlike traditional subsequence extraction methods that can only obtain subsequences of fixed length, the new method obtains subsequences of different lengths, which allows us to identify more HS patterns and greatly increases the application prospects of the new method in financial chart pattern recognition. Experiments show that the new method outperforms segmentation methods such as KP, PIP, PAA, PLA, as well as traditional matching methods such as DTW and ED on simulated datasets. On the Shanghai Stock Exchange Composite Index (SSEC) and Dow Jones Industrial Average (DJI) datasets, the new method identifies more HS patterns and effectively reduces the probability of HS being misidentified. The results on the stock price trend of consumer electronics show that correctly identifying HS patterns will bring great application value in practice.