Jill C tes November 16th 2020 Building a MovieLens Recommender System 

Customers who bought this item also bought… Amazon Similar items based on your browsing history… Based on your reading history… Because you watched Narcos… “Finding your best match” Jobs recommended for you… Net ix Linkedin Online Shopping Medium OkCupid 

limited inventory mainstream products unlimited inventory niche products Brick-and-mortar E-commerce 

Brick-and-mortar limited inventory mainstream products unlimited inventory niche products E-commerce 

“A physical store cannot be recon gured on the y to cater to each customer based on his or her particular interests.” - Chris Anderson 

The Tasting Booth Experiment 6 jam samples 24 jam samples vs. 

6 jam samples 24 jam samples vs. 40% of customers stopped at the limited-choice booth 60% of customers stopped at the extensive-choice booth Initial Interest The Tasting Booth Experiment 

6 jam samples 24 jam samples vs. 30% conversion rate 3% conversion rate Subsequent Purchase The Tasting Booth Experiment 

The Tasting Booth Experiment The Paradox of Choice Less is more Too much choice = Stressful 

What is a Recommender System? An application of machine learning Machine Learning Model Data Predictions 

An application of machine learning Recommender System User preferences Recommendations What is a Recommender System? 

What is a Recommender System? An application of machine learning predicting future behaviour explicit feedback implicit feedback Recommender System User preferences Recommendations 

What is a Recommender System? An application of machine learning predicting future behaviour explicit feedback implicit feedback Recommender System User preferences Recommendations Collaborative ltering Content-based ltering item user John Jim Anne Liz Erica 

Collaborative Filtering “Similar people like similar things” User-item (“utility”) matrix Users Movies Arnold Peter Susan Valerie Jean Walter Charlie 5 4 3 4 1 2 3 1 5 2 2 4 1 4 5 5 4 1 2 4 5 4 2 1 1 5 5 3 5 4 2 4 3 

Collaborative Filtering “Similar people like similar things” User-item (“utility”) matrix Users Movies Arnold Peter Susan Valerie Jean Walter Charlie 5 4 3 4 1 2 3 1 5 2 2 4 1 4 5 5 4 1 2 4 5 4 2 1 1 5 5 3 5 4 2 4 3 

Collaborative Filtering “Similar people like similar things” User-item (“utility”) matrix Users Movies Arnold Peter Susan Valerie Jean Walter Charlie 5 4 3 4 1 2 3 1 5 2 2 4 1 4 5 5 4 1 2 4 5 4 2 1 1 5 5 3 5 4 2 4 3 

Collaborative Filtering “Similar people like similar things” User-item (“utility”) matrix Users Movies Arnold Peter Susan Valerie Jean Walter Charlie 5 4 3 4 1 2 3 1 5 2 2 4 1 4 5 5 4 1 2 4 5 4 2 1 1 5 5 3 5 4 2 4 3 

Users Movies 5 3 4 1 2 3 1 2 2 4 1 5 5 4 1 2 4 5 4 2 1 1 5 3 5 4 2 4 3 “ A system cannot draw any inferences for users or items about which it has not yet gathered sufficient information.” Cold Start Problem ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 

Content-Based Filtering User and Item Features Users Features Arnold Peter Susan Valerie Jean Walter Charlie Movies Features 45 32 20 59 47 17 36 M M M F M F F CA US US US FR CA CA EN EN EN FR EN EN CA Y Y N Y N Y N N Y Y N Y N N Y Y N Y Y Y N age gender country language 96 horror? family? comedy? 97 07 10 19 16 03 EN EN EN EN EN EN EN N N N N N N Y N Y Y Y Y N N N N N Y Y Y N Y Y Y N N Y N N N N N N N N horror family comedy drama thriller language year of release 

Tutorial 

Environment Set-up Option 1: Run notebook locally Option 2: Run notebook in the cloud • Need to install Jupyter Notebook • Google Colab is a Jupyter notebook environment that runs in the cloud • Minimal set-up required (need a Gmail account) • Supports free GPU 

MovieLens Dataset 

MovieLens Dataset • Created by GroupLens research group at the University of Minnesota • Titanic dataset of recommenders