Machine Learning Design Patterns by Valliappa Lakshmanan & Sara Robinson & Michael Munn

Author:Valliappa Lakshmanan & Sara Robinson & Michael Munn [Valliappa Lakshmanan]
Language: eng
Format: epub
Publisher: O'Reilly Media, Inc.
Published: 2021-02-24T16:00:00+00:00

Tabular data multiple ways

To see how we can represent tabular data in different ways for the same model, let’s return to the restaurant review example. We’ll imagine instead that rating is an input to our model and we’re trying to predict the review’s usefulness (how many people liked the review). As an input, the rating can be represented both as an integer value ranging from 1 to 5, and as a categorical feature. To represent rating categorically, we can bucket it. The way we bucket the data is up to us and dependent on our dataset and use case. To keep things simple let’s say we want to create two buckets: “good” and “bad.” The “good” bucket includes ratings of 4 and 5, and “bad” includes 3 and below. We can then create a boolean value to encode the rating buckets, and concatenate both the integer and boolean into a single array.12

Here’s what this might look like for a small dataset with 3 data points:

rating_data = [2, 3, 5] def good_or_bad(rating): if rating > 3: return 1 else: return 0 rating_processed = [] for i in rating_data: rating_processed.append([i, good_or_bad(i)])

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.