Dataset Link - https://grouplens.org/datasets/movielens/latest/
Python 3+ Dependencies - pandas, nltk, sklearn, numpy, matplotlib, tensorflow, keras, jupyter-notebook
This project is a hybrid model of deep learning and bag of words technique for recommending movies. It consists of three parts:-
- Popularity Based - Just a simple popularity based recommendation system used for testing
- Bag of Words - Uses nltk stemmer and count vectorizer for prerocessing and cosine similarity for generaitng results
- Deep Learning - Uses a tensorflow and keras defined model for generating recommendations
After getting the recommendations from the two models I use a point based metric to rank all the recommendations to give the best curated recommendations
Please comment out the following since these are local saves I used for this model:-
- open search by ctrl + f and search
l = set()and comment out the whole block - next search
for i in l:and comment the whole cell - next search
seen = []and comment the whlole cell - next search
model = tf.keras.models.load_model(r'C:\Users\user\Jupyter Files\Recommender System\Saves\deep_learning_recsys_v4.h5')and comment the whole cell - next search
model.save(r'C:\Users\user\Jupyter Files\Recommender System\Saves\deep_learning_recsys_v4.h5')and change the path to wherever you want to save the trained model or if you dont want to then comment this out as well