Web app Link - https://movie-recommender-engine-dk321.herokuapp.com/

1. Run The Collaborative_Filtering(SVD and KNN) ipynb file on google collab only.
2. web.py is the Web app - To run it use New_data.pkl and matrix.pkl data
3. Necessary libraries needed for running web.py : requests==2.27.1 streamlit==1.9.1 streamlit_lottie==0.0.3

I made 3 Types of Recommendation Engine

In this model I used TMDB 5000 Movie dataset : https://www.kaggle.com/datasets/tmdb/tmdb-movie-metadata which contains movies.csv and credits.csv.

IMPORTANT LIBRARIES USED :

for data analysis: import matplotlib.pyplot as plt import seaborn as sns

for scaledown the values: from sklearn.preprocessing import MinMaxScaler from scipy import stats

In this Model I used Weighted average rating and some statistical approach to recommend movies on the basis of Weight average ratings whose formula is given by

From this technique we are going to calculate the average weighted rating whose formula is given by : W = (Rv + Cm)/ v + m R is the average rating of the movie C is the mean of votes w.r.t all the movies v is the number of votes w.r.t each movie(vote_count) m is the minimum votes required to be in a list Top movies

Using the average_weighted_technique I sorted the movies and give recommendations. Then I used both weighted_average_rating and popularity score giving 50%-50% priority to both and recommend popular high rated movies. then I pickle the

In this model I Used the same dataset by pickling dataset from 1st recommender System model as we need average weighted ratings in furthure process.

IMPORTANT LIBRARIES REQUIRED:

for vectorization and Cosine Distance or similarity

from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer

from sklearn.metrics.pairwise import linear_kernel, cosine_similarity

for stemming the words

from nltk.stem.snowball import SnowballStemmer

from nltk.stem.wordnet import WordNetLemmatizer

from nltk.corpus import wordnet

Approaches used :

              1. Preprocessing of data(duplicates, missing values, unnecessary columns, merging, modifying)

              2. Vectorization
              
              3. Cosine similarity
              
              4. Optimizing using weighted average ratings

In this model I first do preprocessing of dataset like finding duplicates, missing values and dropping the unnecessary columns, after that i cleaned the required columns for checking similarities between genres, cast, crew, keywords as it is a content based system. then I merged all this columns to make a final details columns and then convert it into the vectors of strings and then check the similarity scores between each movie and recommend movies with high similarity.

In this model I used movie_lens.csv dataset as it contains users with ratings. In this model I filtered the movies on the basis of no. of user ratings and minnimu no. of ratings required(threshold rating).

Datasets :- https://raw.githubusercontent.com/susanli2016/Machine-Learning-with-Python/master/movielens_data/ratings.csv https://raw.githubusercontent.com/susanli2016/Machine-Learning-with-Python/master/movielens_data/movies.csv

Approches used :

             1. Statistical Filtering
             
             2. Pivot Table
             
             3. Correlation
             
             4.KNN algorithm 

Datasets :- https://raw.githubusercontent.com/susanli2016/Machine-Learning-with-Python/master/movielens_data/ratings.csv https://raw.githubusercontent.com/susanli2016/Machine-Learning-with-Python/master/movielens_data/movies.csv

NOTE : Run The Collaborative_Filtering(SVD and KNN) ipynb file on google collab.

Web App Images: