Loading...
Please wait, while we are loading the content...
Busca de padrões em séries temporais
| Content Provider | Semantic Scholar |
|---|---|
| Author | Batista, Rodrigo De Abreu |
| Copyright Year | 2010 |
| Abstract | This work discusses and implements two approaches used in time series pattern discovery: clustering and motifs. Initially the concepts and formalisms related to the time series are introduced, through the techniques used in the search for patterns and concluding with the presentation of implementations, and a comparison of these. At first the approach of clustering with sliding window is presented, and the problems that arise from that. In an attempt to solve these problems, Self-Organizing Maps are used as an alternative to clustering, and finally the technique of identifying perceptually important points is presented, as solution to prevent trivial matches and to reduce the dimensionality of the input series. Secondly, we describe the technique of pattern discovery based on motifs, and the motivations that led to this approach. A second implementation is presented and implemented using this new approach and, finally, the achieved results are compared by the implementation of methods. Thus, based on experiments, it was concluded that the SOM approach with sliding window generates patterns very smooth that does not resemble the original series, while the Motifs approach finds patterns with better resolutions. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://lume.ufrgs.br/bitstream/handle/10183/28327/000767833.pdf;jsessionid=3A70BBAC8F3A8142E2DA555234B1BDBA?sequence=1 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
