We are dealing with a time series regression problem requiring the prediction of a real-valued output. Let’s say the min temperature during the next 24h.
We want to transform the problem into a classification one and then predict if there is a change of a certain value in the temperature. In this case how should we present our features to the predictive model?
Thanks
You can tell me more clearly or summarize your issue. I will see and comment with you soon.
Vũ Xuân Hạnh , Thank you very much for your interest, I tried to explain my question in the attached file. Let me know if it still not clear
Ameni Mezni : I have worked on a similar problem before, and one of the ways to convert the time-series of input features for classification is to divide the input features (A1, B1, C1) into frames and average it over the chosen time frame. For example A2 could be an average of A1 over last 24 hours (or any other frame) and then you can use these averages (A2,B2,C2) for clustering and classification like K-NN. Hope this helps.
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification