No-Code Forecasting with Time Series Modeler App

This demo shows how to build a time series forecasting model in MATLAB^® using the Time Series Modeler app without writing code. Starting from historical data, the workflow walks through importing the data set, selecting the response and predictor variables, checking the data visually, training a model, generating forecasts, and exporting the results for reuse in other workflows. In this example, the response variable is forecasted using several related predictor variables. After importing the data, the app makes it easy to define the forecasting setup and choose a validation split so model performance can be evaluated on unseen data. A quick visual check of the response and predictors helps confirm that the data looks reasonable before training begins. To capture nonlinear relationships, the workflow uses a deep learning model such as LSTM. The app allows you to configure training options and hyperparameters interactively, while also displaying diagnostics and performance metrics during training. Once the model is trained, the app generates forecasts over the validation window so you can compare predictions against held-out data. The final step is export. The trained model can be sent to the MATLAB workspace, and the app can also generate reproducible MATLAB code. This makes it easy to move from an interactive no-code workflow to a larger automated forecasting pipeline.

This demo is a practical introduction to building, evaluating, and exporting forecasting models in MATLAB with the Time Series Modeler app.