Tensorflow.js Time Series Forecasting

Import Dataset

Upload your CSV that contains the row that you want to forecast. CSV should contain atleast two columns representing the Timestamp and the univariate data that needs to be forecasted. After Uploading select the respective columns from the dropdown and then click extract. This should result in your data being displayed in a graph.

Training & Validation

Here we will be training a LSTM Neural Network model on your dataset. A deep down read on lstm as an approach to forecasting can be found here. You can use the default parameters set and just start training. Two graphs should inititate to help yo monitor model loss and mape.

Loss Graph

MAPE Graph


Your Model is Trained and Ready, simply click forecast to see your models forecasted results.
Note: Your forecast horizon will only go as far as the number of days you trained the model to forecast.