project-1-rec-basics-moviebuddy

Project 1 submission for the course IEORE 4571 Personalization: Theory and Application

data/

All original and derived data files used in the project

1. ratings.csv.zip (could not be added because GitHub does not allow upload of files above 100mb).
2. movies.csv.zip (list of movieIds with their title and genre)
3. final_sample.csv.zip (The preliminary sample used for training and validation of the ALS model with 1000 movies)
4. final_sample_2.csv.zip (The sample used for training and validation of the ALS model with 1500 movies)
5. final_sample_3.csv.zip (The sample used for training and validation of the ALS model with 2000 movies)
6. final_sample_4.csv.zip (Could not be added because GitHub does not allow upload of files above 100mb)
7. final_sr_red.csv.zip (Sample used for KNN training with 1000 movies and users with more than 50 ratings)
8. reduced_final_sr.csv.zip (Sample used for KNN recommendations with 1000 movies and top 100 most active users)
9. ratings_test.csv.zip (Held out test data used for evaluating our ALS model)

Four Jupyter Python3 notebooks

1. eda_sampling.ipynb - Contains code for EDA and sampling method, and also states our Objectives and business metrics
2. Personalization_project_ALS.ipynb - Contains code as well as results on our model-based recommendations, using ALS
3. Personalization_project_KNN.ipynb - Contains code as well as results for item based recommendations, using KNN
4. KNN_recommendations.ipynb - Contains the recommendations made by our optimal KNN model for a very small dataset to compute the coverage on this small dataset.
5. knn_sampling.ipynb - Showing the evaluation of kNN as we increase the sample size.

Showing results of KNN for different values of k

1. kNN Results.pdf - Observations on the accuracy metrics obtained with different values of the hyperparameter k in k-Nearest Neighbors (run across systems on different notebooks for parallelizartion.