Pattern recognition in stock data based on a new segmentation algorithm

In trying to find the features and patterns within the stock time series, time series segmentation is often required as one of the fundamental components in stock data mining. In this paper, a new stock time series segmentation algorithm is proposed. This proposed segmentation method contributes to containing both the important data points and the primitive trends like uptrend and downtrend, while most of the current algorithms only contain one aspect of that. The proposed segmentation algorithm is more efficient and effective in reserving the trends and less complexity than those combined split-and-merge segmentation algorithm. The research result shows that patterns found by using the algorithm and prior to the transaction time impact the stock transaction price. Encouraging experiment is reported from the tests that certain patterns appear most frequently before the low transaction price occurrence.