I need use SVM to clasify time series data.
I knew that scikit-learn is a libarary python for SVM, but i only use to clasify point data, not for sequential data as time series.
Could you tell my a library python for SVM time series data?
Thank you so much!
If you really want to classify time series with SVM (or any other classification algorithm), you need to describe the time series by derived features derived. tsfresh https://github.com/blue-yonder/tsfresh is a powerful python library for this purpose. Unfortunately, it derives some thousand of features from a time series, so a good idea is to reduce the dimensionality of the derived data with PCA or ICA
@ Yu-Da Lin: Thank you Sir. Infact my question is for classification time series.
@ Manoj Kumar: Thank you Sir. I used scikit learn for SVM in the case of point data, not series data.
@ Thomas Hoppe: Thank you Sir for your great explanation. I now understand well the approach you suggested. Each time series is map to a vector of feature fixed dimension and we can use SVM library. An other approach i defined a kernel and solve quadratic equation programming with CVXOPT to obtain parameters.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python