How to predict values from machine learning model?

Hi everyone

I am new to machine learning and MATLAB. I am not sure what I am doing wrong. I have a table of stock market data called CIV. The table has symbol, date, and CIV. See screenshot from workspace. I want to forecast by symbol and date. For example, for symbol ABC I want forecast for 10 days. For symbol DEF I want 10 day forecast. The underlying data for the symbols will be different b/c the symbols are different. Hopefully this makes sense. If not, I will clarify if needed. I was able to export the trainmodel. How do I tell it that I want to forecast 10 days starting from date X? I created an array of the dates to use for the forecast but I can't figure out how to use it. Here are the dates:

% CREATE ARRAY OF DATES THAT ARE USED FOR FORECASTING
T0 = datetime("today");
T1 = datetime(busdate(T0,1));
T2 = datetime(busdate(T1,1));
T3 = datetime(busdate(T2,1));
T4 = datetime(busdate(T3,1));
T5 = datetime(busdate(T4,1));
T6 = datetime(busdate(T5,1));
T7 = datetime(busdate(T6,1));
T8 = datetime(busdate(T7,1));
T9 = datetime(busdate(T8,1));
T10 = datetime(busdate(T9,1));

ForecastDates = [T1 T2 T3 T4 T5 T6 T7 T8 T9 T10];

Here is how I started the machine learning app:

Here is a screenshot of the variables in my workspace:

Thank you

[KSSV]

What are your inputs and target? How did you train the data?

[Manny]

One more thing...

MATLAB tells me to use yfit = trainedModel.predictFcn(T) for forecasting but I cannot figure out how to give it an array of dates I created above. Any help is much appreciated.

[Manny]

Hi @KSSV. I entered what I showed in the first screenshot. I picked 'CIV' and then the app automatically picked the response and predictors. I only changed the training data field to 25. I used the learner app to train it.

Answer 1

[Drew]

The Classification Learner app is not specifically designed for working with time series. To analyze time series, there are several options depending on the requirements of your use case. Here are some of the options:

• With the Statistics and Machine Learning Toolbox, apply machine learning techniques to time series using the directforecaster function introduced in R2023b https://www.mathworks.com/help/stats/perform-time-series-direct-forecasting-with-directforecaster.html
• With the Deep Learning Toolbox, use deep learning techniques for time series forecasting https://www.mathworks.com/help/deeplearning/ug/time-series-forecasting-using-deep-learning.html .
• With the System Identification toolbox, perform time series prediction and forecasting for prognosis https://www.mathworks.com/help/ident/ug/time-series-prediction-and-forecasting-for-prognosis.html
• With the Econometrics toolbox, analyze time series using the Econometric Modeler. https://www.mathworks.com/help/econ/econometric-modeler-overview.html
• Additional options for working with financial time series data can be found at https://www.mathworks.com/campaigns/offers/financial-time-series-cheat-sheets.html .

If this answer helps you, please remember to accept the answer.