bootcamp2017

Baby name recommender website

https://namesilike.com

https://en.wikipedia.org/wiki/Recommender_system

Baby names dataset

Questions:

• Which names are seen a lot?
• Which names are clicked a lot?
• Which names are clicked a lot per view?

Recommendation engine notebook

Popularity-based recommender

Popularity dataset

Question: How often do people click on the most popular name (2016 popularity)?

Content-based recommender

Features

Task:

• Merge features and feedback
• Pick a user, and set up a cross-validation loop of the questions
• For each fold, build a random forest model of feedback given the features on the train set and evaluate correctness on the test set.

Collaborative filtering

Task:

• Construct a User-Product matrix
• Construct a User-User similarity matrix
• Produce recommendations based on top 20 similar users for a given user.

Bayesian statistics

Beta distribution

Probabilistic Programming & Bayesian Methods for Hackers

Pairwise transformation

Blog post

Ethics of recommendation engines

Beyond accuracy

Movie Dataset

MovieLens Datasets