Content platforms such as Netflix or Hulu use AI to recommend programs that appeal to each viewer’s unique taste. However, most of these recommendation algorithms lack an ability to suggest a more diverse array of films to viewers. Our machine learning application bridges this gap by suggesting foreign films, low-budget films, and films directed by women to users.
We first obtained data from The Movie Database API and exported a final csv containing data on female directed films, foreign language films, and films ranging in budget. We then created a similarity matrix through Scikit-Learn’s CountVectorizer and cosine_similarity tools, which returned a sorted list of films based on each film’s unique similarity score. Data was then sorted for each endpoint in our final Flask application by percent_female_directed, foreign language, and budget bins from 0 to 15 million.
Our final application includes the following:
- An index page which sorts films by similarity scores only.
- A Female Focused page that displays a graph of films directed by women and their corresponding budget and revenue, plus a table of similar films.
- An International page which displays an interactive map of similar foreign language films.
- A Low Budget page that displays films with budgets less than $15 million. Plus an explore page that allows users to view a table of the most popular and least popular similar low budget films.
Users can view and interact with our final application here: https://movies-ml.herokuapp.com/
data_cleaning
- Contains our initial DataCleaning.ipynb file along with more data exploration and our final csv export.
similarity_matrix
- Contains our intial similiarty_matrix.ipynb file.
static
- css: Contains our CSS files for styling each page.
- data: Contains JSON files that dynamically update each time the user enters a new search.
- images: Contains our homepage image
- img: Contains images for markers added to our Leaflet map for the international endpoint.
- js: Contains our JavaScript files for our index page and each additional endpoint.
templates
- Contains each html file for our index page and each additional endpoint.
app.py
- Our main python Flask application that routes data to our similarity.py app and each additional endpoint.
requirements.txt
- Essential package dependencies needed for our final Heroku application.
similarity.py
- Our similarity matrix that sorts by similar movies from the user’s input.
| Owner | Description | Task |
|---|---|---|
| Julia | Data Exploation | 1. Call Movie Database API and review available data. 2. Perform basic data cleaning based on necessary independent variables. 3. Build up database (csv format) with films/data. |
| Christopher | Create Homepage | 1. Create html and css templates for index.html. 2. Add nav bar + search bar. 3. Create a default route in flask app that routes user input to all other endpoints. |
| Dana | Build ML Model in Jupyter Notebook | 1. Create a similarity matrix using sklearn’s CountVectorizer and cosine_similarity libraries. 2. Transfer ML Model to similarity.py |
| Robin | Create Flask App | 1. Build app.py and route data to each endpoint. 2. Route to similarity.py and filter results using methods=['POST', 'GET'] |
| Emory | Low Budget Endpoint | 1. Build JavaScript app. 2. Add an endpoint to the flask app. 3. Build html and css for Low Budget page. |
| Carmela | International Endpoint | 1. Build JavaScript app. 2. Add an endpoint to the flask app. 3. Build html and css for Female Focused page. |
| Robin | Female Focused Endpoint | 1. Build JavaScript app. 2. Add an endpoint to the flask app. 3. Build html and css for Female Focused page. |
| Jacob | Host application on Heroku | 1. Add dependencies in requiqments.txt file. 2. Debug and deploy app from GitHub to Heroku. |
Data collected from The Movie Database