Hands-On Data Analysis with Pandas

Further improvements

They are happy with our results, and now want us to provide them with predictions each minute. They have also promised to provide feedback within an hour. We won't implement this request here, but we will briefly discuss how we could go about this.

We have been using batch processing to update the model each month; however, in order to provide our stakeholders with what they want, we will need to shorten our feedback loop by performing the following actions:

• Running predict() on our model every single minute and having the predictions sent to our stakeholders. This will require setting up a process to pass the logs one minute at a time to a preprocessing function, and then to the model itself.
• Delivering the results ...