From the course: AI-Powered Time Series Forecasting with Python

Unlock the full course today

Join today to access over 24,700 courses taught by industry experts.

Online features

Online features - Python Tutorial

From the course: AI-Powered Time Series Forecasting with Python

Start my 1-month free trial Buy for my team

Online features

“

Let's learn how we can apply our feature pipeline to our streaming dataset. We will build an online pipeline that listens to incoming messages, applies our window function, creates our features, and stores them to a new topic. But first, let's launch our docker container. Type docker- compose up -d. And once the docker container is running, go to the data producer script and also run it. Make sure that data is being produced to your stream. Now let's open the file feature_pipeline.ipynb. This code will generate our online feature store. Quick context: I changed our scenario here a little bit to adjust for the real-time scenario. Let's say in this real- time setting, we want to predict the energy demand not for the next 24 hours, but just for the next hour. I updated our features for that and I'm using a different model. But let's go through this code section by section. The first function parse_period is a utility function to make sure we can pass the date time stamps correctly, since…

Contents