Movie Rating Prediction Model

Goal

The purpose of this notebook is to build a predictive model for movie ratings based on various features such as keywords, genres, director, and average user ratings. The model predicts how a particular user will rate a movie on a scale of 0.5 to 5.0.

Steps

1. Data Collection

   • Fetch movie details like keywords, genres, average rating, title, and release date from the TMDB API using the movie name.

2. Data Preparation

   • Process and encode features like keywords, genres using label encoders.
   • Utilizing Embeddings: Since our data has has high cardinality features like keywords (1932) , simply using one-hot encoding will lead to probolems like Dimensionality Curse, increase in model computational time, and space. Instead , we use Embeddings for categorical features. This approach captures semantic relationships between categories and reduces dimensionality, leading to better performance and efficiency in training the model.
   • Prepare input data by padding keywords and genres to fixed lengths.

3. Modeling

   • Train a neural network with embedding layers for categorical features (keywords, genres) and fully connected layers for numerical features (like average_rating).
   • Use metrics like Mean Absolute Error (MAE) and R-squared (R²) score to evaluate model performance.

4. Evaluation

   • Predict movie ratings and compare them with actual ratings given by a user.
   • Investigate cases where the model's predicted rating deviates significantly from the given rating.

5. Interpretation

   • Display predictions along with the movie's title, release year, and detailed features.
   • Use the model to predict and print the rating for any given movie by name.

Metrics

• Mean Absolute Error (MAE): Measures the average absolute difference between predicted and actual ratings. A low value indicates a better fit.
• R² Score: Represents how well the model captures the variation in the actual ratings. An R² score close to 1 suggests the model fits well.