I am working on hourly multivariate time series, due to the huge dimensionality of the data, I convert the data into functional data, using the Fourier basis function, the model use in this problem is the functional autoregressive model of order more than one, as per my knowledge their is no package available in R to deal with such model. So I use the alternative, I decompose the functional curve through spectral decomposition into functional principle components function and there associated scores. Then using the multivairiate time series method to forecast the principal components scores, then multiple these forecast scores with functional principle components functions to retransform to functional form from multivarite, In this a way I obtain one year a head forecast, where each day is a single curve. Know my problem start from here that I want to check the accuracy of the model, I think that I evaluate each curve into 24 points, but how to due this is in R. when I obtain these forecasted values in which each day of 24 hours, compare it with origional hourly data, and calculate the mean absolute percentage error, please help me.
Dear Faheem,
Please look at the paper, if the applied technique is useful for your research, I would be happy to share the code with you.
Article A deep learning-based evolutionary model for short-term wind...
Dear Mehdi Neshat
Thanks for your response, please share your paper code, its help me in my work
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking