The main goal of the project was to implement multiple variations of two of the most widely used algorithms for movie recommendation: content - based recommendation and collaborative filltering. Each variation was tested separately, and the most eficcient versions were combined into hybrid system.
- Python 3.x
- Pandas
- NumPy
- scikit-learn
In order to install all required dependencies, run:
pip install -r requirements.txt
All algorithms were implemented and tested, using MovieLens 100k dataset. The data were processed ang merged, in order to obtain the most suitable and convenient form to perform calculations.
The scripts responsible for interaction with the user are :
- add_user.py
- add_rating.py
- hybrid.py
This script enables to add new user to the database. After running it with command:
python add_user.py
it asks the user to rate 60 movies, in order to establish his/her preferences. Presented films are chosen in a very specific way:
- first, all movies are assigned a weight, based on the number of ratings (the idea is to ask about rather popular movies, because there's a higher chance, that the user has seen them)
- then, 500 movies are sampled, with the more popular ones having bigger chance to be drawn
- out of the sampled movies, the script choses 60 ones with the highest variance of ratings (we want to ask about polarizing movies, because the are more desctiptive - a user giving high rating to a commonly loved movie, such as Forrest Gump, doesn't provide much useful information).
After the user is done rating movies, he/her is being added to the database (in form of a Pandas DataFrame), and similatities with all the other users are being calculated.
This script is responsible for adding new ratings for existing users. It is run with a command:
python add_rating.py
After running the script, the user is asked to give his ID. After that he can searched the movie titles, and rate them.
The main script, responsible for provinding recommendations, as well as evaluating different versions of the algorithm. It is run with a command:
python hybrid.py (--r| --e) -a1 -a2 -a3 -a4 –k [--user] [--n]
| Argument | Description |
|---|---|
| --e | evaluate algorithms by calculating MSE, RMSE and MAE |
| --r | recommend movies using specified version of the algorithm |
-a1 <int> |
the weight given to the first version of the algorithm |
-a2 <int> |
the weight given to the second version of the algorithm |
-a3 <int> |
the weight given to the third version of the algorithm |
-a4 <int> |
the weight given to the fourth version of the algorithm |
-k <int> |
number of neighbors user in collaborative filltering |
| Argument | Description |
|---|---|
| --user | ID of a user that the movies are being recommended to |
| --n | number of movies we want to have recommended |