Movie Recommendation System

Overview

This project involves creating a movie recommendation system using a dataset containing movie attributes. The goal is to recommend movies similar to a given movie based on content similarity.

Dataset Link - https://www.kaggle.com/datasets/tmdb/tmdb-movie-metadata

Dataset

The dataset includes the following columns: budget, genres, homepage, id, keywords, original_language, original_title, overview, popularity, production_companies, runtime, spoken_languages, status, tagline, title, vote_average, and vote_count.

Data Preprocessing

1. Handling Missing Values: Missing values were identified and handled. Specifically:

   • overview had 3 missing values, which were removed.
   • Other columns had no missing values.
   • No duplicate entries were found in the dataset.

2. Data Transformation:

   • Genres, Keywords, Cast, and Crew: Extracted and converted relevant information to a list format.
   • Tags Creation: Combined overview, genres, keywords, cast, and crew into a single tags column for each movie.
   • Text Normalization: Applied stemming to the tags column using the Porter Stemmer to standardize the text data.

Feature Extraction

• Count Vectorization: Converted the tags into numerical features using CountVectorizer with a maximum of 5000 features and excluding common stop words.

Similarity Measurement

• Cosine Similarity: Calculated pairwise cosine similarity between movies based on their vectorized tags to quantify content similarity.

Recommendation System

• Functionality:
  • The recommend function finds the most similar movies to a given movie based on cosine similarity scores.
  • It excludes the movie itself and returns the top 5 most similar movies.

Example

For the movie "Avatar", the system recommended:

• "Titan A.E."
• "Small Soldiers"
• "Independence Day"
• "Ender's Game"
• "Aliens vs Predator: Requiem"

Conclusion

This recommendation system effectively suggests movies similar in content to the input movie, enhancing user experience by providing relevant suggestions based on textual analysis.