/search-suggest

Building a movie recommendation system

Primary LanguageJupyter Notebook

Part 1: Using Collaborative Filtering to Recommend Alternative Movies

Case Study

The customer service team at Netflix has started to receive user complaints about Netflix having a limited catalog of movies. In response, the product team has decided that they want to suggest similar movies as a consolation to users that can't find their desired movie. The product team has asked their top two data scientist, Debbie and Juan Carlos, to create an algorithm that allows a user to receive suggested movies when their searched for movie is not available.

Stakeholders

  1. Product Team: This team has requested the ability to provide a user with suggested movies when the user is not able to find the specific movie they want to watch.
  2. Customer Service Team: This team expects that the new feature will reduce the number of user complaints about the movies not being available on their library.

Success Metrics

  • The number of user complaints related to a movie not being available should show statistically significant decreases after the feature is implemented.
  • The number of people who watch a recommended movie should increase while the number of user complaints decreases.

Part 2: Using TensorFlow Wide & Deep with LASSO regression to make movie recommendations

We've used TensorFlow's Wide & Deep model to predict movie ratings, and also made an extension to the model by implementing LASSO regression for improved feature selection. For full details, refer to the subdirectory "Part_2".

To run the code in any subdirectory, do the following:

  • To run the notebook found in the Part 1 or Part 2 follow these instructions:
    • Navigate to the "part_1" or "part_2" folder of the directory.
    • Run this code if you do not have conda installed in your machine.
    wget http://repo.continuum.io/miniconda/Miniconda3-3.7.0-Linux-x86_64.sh -O ~/miniconda.sh
bash ~/miniconda.sh -b -p $HOME/miniconda
export PATH="$HOME/miniconda/bin:$PATH"
    • This line will create a file with all the libraries needed for the notebook to run.
    conda env create -f environment.yml
    • Depending on your OS run the respective line below to activate the environment needed for the notebook.
    Windows:            activate rec
macOS and Linux:    source activate rec
    • For Part 1, this line will run jupyter lab on your machine making it available on port 8888.
    jupyter-lab Collaborative_Filtering_Part_1.ipynb
    • For Part 2, this line will run jupyter lab on your machine making it available on port 8888.
    jupyter-lab Wide-Deep-Lasso.ipynb