First, the paper summarizes the study
- Research object
of this paper is a review of research on a similarity measure in the time series data mining represents. - Research motivation
time series data mining problem
analysis time series feature representation and similarity measure role in data mining, the significance of the main methods currently available are reviewed, their advantages and disadvantages exist; moreover discusses some concern The problem. - Literature review
discusses the current situation and characteristics of research methods; research status and methods of similarity measure.
Wherein represents
1.1 piecewise linear representation
1.2 polymerizable segment representation
1.3 symbolic representation
1.4 transform domain representation
1.5 Singular value decomposition representation
1.6 representation model based
similarity measure
1.1 Euclidean distance
1.2 Dynamic Time Warping
1.3 symbolic from
1.4 and a compression model based on variable distance
1.5 from the other measurement method - Study Design
- Using data sets
- Analysis conclusion
- Learning experience
because of the time series data is typically a time varying dimensions and varying high generation process vulnerable to environmental factors, and the presence of some noise. For such complex data to study how to derive effective information and knowledge ,, it has a very important theoretical research value and importance of social production practice and scientific research.
High dimensional nature of time series, we need to extract a time series feature local or global decomposition, reducing the dimension of the original time series, and the combined similarity measure time series of effective and reasonable time-series data mining.
Characterized in that the method represents a high-dimensional space time sequence is mapped to a low-dimensional feature space, to achieve data reduction, the reaction may be time series basic form and important information, time series data to improve mining efficiency.
Similarity measure similarity comparison method, the relationship between the effective response time series.