Randomisation of matrix for load forecasting

2 views (last 30 days)
Is it advicable to do randomisation while doing laod forecasting using neural network tool in matlab?
while i use the below command to a 87648*6 double matrix, i get p matrix with dimension 1*87648 double.
p = randperm(n)
Since i have different values in diffent columns of the xlsx file, p returns a single column with values.Is that correct to proceed with load forecasting?Does it mix all the values and during later stage in training the system, does it takes the values correctly?

Accepted Answer

Walter Roberson
Walter Roberson on 21 Oct 2020
Edited: Walter Roberson on 21 Oct 2020
No, not in itself. You could use p as column indices, but when you submit all of the data together then it does not care what order the columns are so it would be a waste of time.
However it would sometimes make sense to take the first part of p, such as the first half, and use that as column indices to get the data to train a model with, and then use the second part of p to get column indices for data to test how well the model works on data it was not trained on.
This kind of random division into test and training sets is the default for the "shallow" neural network routines such as feedforward networks.

More Answers (0)

Categories

Find more on Deep Learning Toolbox in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!