课程: MLOps with Databricks

免费学习该课程！

今天就开通帐号，24,600 门业界名师课程任您挑！

Create a FeatureSpec for Feature Serving

Create a FeatureSpec for Feature Serving - Databricks教程

课程: MLOps with Databricks

开始一个月试用 购买团队方案

Create a FeatureSpec for Feature Serving

“

- [Instructor] In the previous video I mentioned that prediction serving should be used when all the predictions are pre-computed and should only be looked up at the moment of request. Let's go back to our hotel booking cancellation use case and mimic this scenario. In this video, we are going to calculate the probability of cancellation for all bookings using the model we trained earlier. Start the predictions and probabilities in the feature table. Create an online table and the feature spec that defines the feature lookup. Let's create a feature table with all the predictions before we go into the code. Make sure you run pip install command and install all the requirements and restart Python. First, we load our train and test data sets into a pySpark data frame. Then we load the secular pipeline we trained early into memory and use it to generate predictions and probabilities that will be part of the feature table. Let's now create a feature table that takes a predictions data…

内容