Member-only story
Time Series Forecasting with Deep Learning in PyTorch (LSTM-RNN)
An in depth tutorial on forecasting a univariate time series using deep learning with PyTorch
Introduction
Believe it or not, humans are constantly predicting things passively — even the most minuscule or seemingly trivial things. When crossing the road, we forecast where the cars will be to cross the road safely, or we try to predict exactly where a ball will be when we try to catch it. We don’t need to know the exact velocity of the car or the precise wind direction affecting the ball in order to perform these tasks — they come more or less naturally and obviously to us. These abilities are tuned by a handful of events, which over years of experience and practice allow us to navigate the unpredictable reality we live in. Where we fail in this regard, is when there are simply too many factors to take into consideration when we are actively predicting a large scale phenomenon, like the weather or how the economy will perform one year down the line.
This is where the power of computing comes into focus — to fill the gap of our inability to take even the most seemingly random of occurrences and relate them to a future event. As we all know, computers are extremely good at doing a specific task over numerous iterations…