with the advent of information society in the late 20th century, massive data streams, time sequence, and graphics, social networks, spatio-temporal data, text data and Web data pop up. Extracting useful information from massive data becomes the hot problem of data mining. Much time series data, consisting of values or matters measured in different times, exist in such fields as finance, engineering experiments, meteorology, medicine, transportation, etc. In the field of data mining, time series mainly focuses on trend analysis, similarity search, and association rule mining and so on. This paper, taking the share data as an example, after giving an overview of data mining research and development, makes some research on the expression, rule mining and similarity search concerning temporal sequence, and obtains some achievement in such aspects as subsection symbol expression, similarity search algorithm as well as association rule detection of time sequence. The main research contents and results are as follows:1) Introduce basic theories of data mining, including data mining analysis process, traditional analysis and mining analysis of time sequence, frequent pattern detection, strong association rules and so on, and further make an in-depth, systematic study and analysis of them.2) subsection linear expression and symbol expression of time sequenceAfter research an analysis of usual subsection linear expression of time-sequence, adopt the subsection algorithm featured by extracting special points according to slope change to compress data, which solves the problem of high dimension and massive data of time sequence in a good way. Put forward the eight element model symbol expression way based on relative slope, which effectively indicates the relation between share price change and time. What's more, stock data validation algorithm is adopted to obtain better results.3) similarity search studyAs for the time series similarity search method, raise the fast similarity search algorithm based on subsection symbolization and the smallest cycle chain code, which compared to the former, has higher efficiency, and has effectively solved the influence to similarity judgement due to the timeline translation, expansion and rotation. Stock data verification helps a lot in the analysis of share time-series data.4) study on association rule detection Study on correlation algorithm about time sequence symbolized, Proposed a new algorithm including Apriori and FP-tree, finding out the frequent pattern of single stock and multi-stock data, and generating strong association rules, which can guide the stock investors. |