SDS 463: Time Series Analysis — with Matt Dancho

Gluon Package

The Gluon package leverages deep learning for time series forecasting, utilizing models like LSTMs and Nbeats. Matt Dancho explains that Gluon T's integrates these models, making it easier to implement complex algorithms without extensive coding 1. This package is particularly useful for handling large datasets and producing robust predictions 2.

The cool thing about LSTMs is they're basically an ARIMA on steroids. They look backwards in time and use that sequence to decide how the path should look into the future.

--- Matt Dancho

Jon Krohn highlights that Gluon T's is unique in its focus on time series, unlike other frameworks like TensorFlow and PyTorch 1.

   

Nbeats and DeepAR

Nbeats and DeepAR are two prominent models within the Gluon T's package, designed for time series forecasting. Matt Dancho notes that these models are highly effective in competitions and large-scale projects, often requiring cloud resources for optimal performance 3. They are part of the Amazon Web Services (AWS) ecosystem, which enhances their scalability and efficiency 3.

Nbeats and DeepAR are two approaches to building time series neural network models as a part of the Amazon Web Services AWS Gluon T's package.

--- Matt Dancho

These models significantly boost productivity by automating many aspects of time series analysis, from preprocessing to feature engineering 4.