Why you should use transformers 🦾 in the time-series forecasting 📈 scenarios?
A thread ⬇️
A thread ⬇️
Most time-series techniques are based on ARIMA.
TS problems that showed to be challenging for ARIMA:
1) with long-term, short-term seasonality patterns
2) that include outlier events that deviate from historical trends
2/5
TS problems that showed to be challenging for ARIMA:
1) with long-term, short-term seasonality patterns
2) that include outlier events that deviate from historical trends
2/5
Deep learning methods were adopted in time-series forecasting.
Traditional CNN and RNN models showed limitations when dealing with large multivariate datasets because their complexity increases rapidly with the input size.
3/5
Traditional CNN and RNN models showed limitations when dealing with large multivariate datasets because their complexity increases rapidly with the input size.
3/5
Attention mechanisms seem like an ideal mechanism to weigh the importance of past events in future outcomes.
Choosing these weights is one of the main challenges for implementing transformers models in TS forecasting.
4/5
Choosing these weights is one of the main challenges for implementing transformers models in TS forecasting.
4/5
Thanks for learning ML and AI with us!
We highly recommend you our bite-sized overview of time-series forecasting available to everyone: thesequence.substack.com
The feedback is welcome!🙌
5/5
We highly recommend you our bite-sized overview of time-series forecasting available to everyone: thesequence.substack.com
The feedback is welcome!🙌
5/5
Loading suggestions...