Time Series Forecasting

Temporal Fusion Transformer

Interpretable | multi-horizon | multivariate time-series prediction

To address the business problem, we have to return multi horizon predictions. The target we predict, cargo demand, was influenced by multi variables such as holiday, historical cargo demand, and cargo demand of other vessel. In technical words, it is a multi horizon and multivariate time series forecasting problem. We chose the Temporal Fusion Transformer (TFT) for its ability to provide interpretable predictions, which assists users in identifying globally important features, temporal patterns, and significant events.

https://medium.com/dataness-ai/understanding-temporal-fusion-transformer-9a7a4fcde74b