Time Series Forecasting Using Deep Learning with NARX dataset

Hi

I am studying and trying to solve a problem, but my school isnt great and I have no I do not know where to find the information to solve the problem. I am just looking for WHERE i can find the solution, no the solution itself.

I have used Time Series Forecasting Using Deep Learning (https://au.mathworks.com/help/deeplearning/ug/time-series-forecasting-using-deep-learning.html#) which has been quite helpful but when i get to training the network (step 8) it is failing, with error:

"Error using trainNetwork. Invalid training data. Predictors must be a N-by-1 cell array of sequences, where N is the number of sequences. All sequences must have the same feature dimension and at least one time step."

The question in whole is:

Use the nonlinear autoregressive network with exogenous inputs (NARX) dataset with the following command in MATLAB. data = simplenarx_dataset;

Perform the following tasks using MATLAB.

1- Plot the sample data set and label the axis.

2- Split the sample dataset into parts, 90% for training and 10% for testing.

3- Standardize the training and testing data to have zero mean and unit variance.

4- Specify the responses to be the training sequences with the values shifted by a one-time step.

5- Create a long short-term memory (LSTM) network

6- Create a long short-term memory (LSTM) network with 200 hidden units.

7- Set the solver to 'adam' and train for 250 epochs.

8- Train the LSTM network.

9- Forecast the values of multiple time steps in the future.

10- Plot the training time series with the forecasted values (Training Data + Forecasted Data).

11- Compare the forecasted values with the test data by plotting the root-mean-square error (RMSE), observed and forecast data.

Thanks for any assistance

Answer 1

[Vatsal]

Hi glenn,

I understand that you're encountering an error while performing time series forecasting using deep learning. The dataset you're using, the NARX dataset, is a 1x100 cell array where each value of the cell array represents a single value. As a result, the steps mentioned in the provided link cannot be directly applied. However, you can use the following lines of code to obtain the "XTrain" and "TTrain" values:

timeSeriesData = cell2mat(dataTrain); % Convert cell array to a numeric array

XTrain = timeSeriesData(1:end-1);

TTrain = timeSeriesData(2:end);

After updating these values, you can update the next steps accordingly. This should help resolve the error.

I hope this helps!