- Build the first iteration of a recommender system that could suggest top-recommended movies to users.
- Provide a GitHub repo (private is fine) with the working model and some description of how to reproduce training/testing steps.
- Provide written answers to these questions:
- A brief justification of the chosen model and the scoring metric. What alternatives would be worth exploring?
- What would be your next steps to improve the recommendation quality?
- You define the technical details of the solution: the model, feature set, test/train split, normalization, loss, scoring metric.
- You're not expected to use as much features as possible. Explore the data and take what works best for your approach.
- You're not expected to produce a model with perfect scores. It's enough to pick a sensible model, make a few tweak iterations, produce some results, and outline a path to improve it.
- Python 3.8+, Torch/TensorFlow/Keras.
- The model, the traing and testing code should be in Jupyter notebooks. Everything else may be in notebooks or python files.
- If trainng takes more than an hour on a laptop, provide the trained model and the code to load it.
movies.pickle (4107 rows) — basic info about movies:
| Column | Type | Example | Notes |
|---|---|---|---|
movie_id |
int |
109830 |
|
title |
str |
"Forrest Gump" |
|
genres |
set[str] |
{"Romance", "Comedy"} |
|
year |
int |
1994 |
|
synopsis |
str |
"The film begins with feather..." |
Detailed plot description |
aggs.pickle (28557 rows) — aggregated ratings, total and by demographic:
| Column | Type | Example | Notes |
|---|---|---|---|
movie_id |
int |
109830 |
|
rating_average |
float |
8.8 |
|
rating_count |
int |
304 |
Number of ratings collected for the group |
demographic |
str |
"age_18_29" |
Group name: total, by age, by gender |
teams.pickle (190547 rows) — cast & crew:
| Column | Type | Example | Notes |
|---|---|---|---|
movie_id |
int |
109830 |
|
person_role |
str |
"actor" |
Enum: actor/director/writer |
person_id |
int |
37097 |
|
person_name |
str |
"Giovanni Arpino" |
labels.pickle (42237 rows) — user ratings to use as labels:
| Column | Type | Example | Notes |
|---|---|---|---|
movie_id |
int |
109830 |
|
user_id |
int |
184 |
|
rating |
float |
8.8 |
1-10 scale |