This project aims to predict IMDb movie ratings using Machine Learning techniques. The dataset contains information about Indian movies, including genres, directors, actors, duration, and vote counts.
The project covers the complete Data Science workflow:
✅ Data Cleaning ✅ Exploratory Data Analysis (EDA) ✅ Feature Engineering ✅ Machine Learning Model Building ✅ Model Evaluation ✅ Feature Importance Analysis
| Metric | Score |
|---|---|
| MAE | 0.983 |
| RMSE | 1.254 |
| R² Score | 0.151 |
- 👍 Votes were the most influential feature.
- 🎭 Drama was the most common genre.
- 🎬 Gulzar was the highest-rated director (minimum 10 films).
- ⭐ Dev Anand had the highest average rating among actors with 50+ movies.
- 📊 Genre and actor information significantly impacted predictions.
To build a machine learning model capable of predicting IMDb movie ratings based on:
- 🎭 Genre
- 🎬 Director
- ⭐ Lead Actors
- ⏱️ Duration
- 👍 Number of Votes
Dataset Size
| Metric | Value |
|---|---|
| Total Records | 15,509 |
| Final Cleaned Records | 7,558 |
| Features | 10 |
| Feature | Description |
|---|---|
| Name | Movie Name |
| Year | Release Year |
| Duration | Movie Duration |
| Genre | Movie Genre |
| Votes | IMDb Votes |
| Director | Director Name |
| Actor 1 | Lead Actor |
| Actor 2 | Supporting Actor |
| Actor 3 | Supporting Actor |
| Rating | IMDb Rating ⭐ |
The following preprocessing steps were performed:
- Removed missing values
- Removed duplicate records
- Converted Duration into numeric format
- Converted Votes into numeric format
- Label encoded categorical variables
- Prepared training and testing datasets
| Column | Missing Values |
|---|---|
| Year | 528 |
| Duration | 8269 |
| Genre | 1877 |
| Rating | 7590 |
| Votes | 7589 |
| Director | 525 |
| Actor 1 | 1617 |
| Actor 2 | 2384 |
| Actor 3 | 3144 |
After cleaning:
✅ No missing values remained.
| Genre | Count |
|---|---|
| Drama | 1137 |
| Drama, Romance | 443 |
| Action, Crime, Drama | 417 |
| Action | 391 |
| Drama, Family | 291 |
Drama is the dominant genre in the dataset, indicating that Indian cinema is heavily represented by drama-oriented films.
| Genre | Average Rating |
|---|---|
| History, Romance | 9.40 |
| Documentary, Family, History | 9.30 |
| Documentary, Music | 8.90 |
| Documentary, Thriller | 8.70 |
| Documentary, Sport | 8.60 |
Documentary and historical genres tend to receive the highest audience ratings.
| Director | Avg Rating |
|---|---|
| Gulzar | 7.55 |
| Anurag Kashyap | 7.40 |
| Bimal Roy | 7.29 |
| Shyam Benegal | 7.25 |
| Govind Nihalani | 7.15 |
Directors such as Gulzar and Anurag Kashyap consistently produce highly-rated films.
| Actor | Movie Count |
|---|---|
| Mithun Chakraborty | 231 |
| Dharmendra | 217 |
| Jeetendra | 179 |
| Ashok Kumar | 173 |
| Amitabh Bachchan | 162 |
Mithun Chakraborty and Dharmendra are the most frequently appearing actors in the dataset.
| Actor | Avg Rating |
|---|---|
| Dev Anand | 6.79 |
| Shammi Kapoor | 6.71 |
| Ashok Kumar | 6.44 |
| Sanjeev Kumar | 6.43 |
| Rajesh Khanna | 6.38 |
Dev Anand has the highest average rating among actors with substantial film counts.
Most movie ratings fall between 5 and 7.5, indicating a relatively balanced distribution of movie quality.
Movies with higher vote counts tend to cluster around average-to-high ratings, suggesting popularity influences rating stability.
No strong relationship exists between movie duration and ratings.
🌲 Random Forest Regressor
- Genre
- Director
- Actor 1
- Actor 2
- Actor 3
- Votes
- Duration
⭐ Rating
| Dataset | Size |
|---|---|
| Training Set | 6046 |
| Testing Set | 1512 |
| Metric | Score |
|---|---|
| MAE | 0.983 |
| RMSE | 1.254 |
| R² Score | 0.151 |
- The model predicts movie ratings with an average error of approximately 1 rating point.
- The R² score indicates that the current features explain around 15% of the variance in movie ratings.
- Additional metadata and advanced feature engineering could improve performance.
| Rank | Feature | Importance |
|---|---|---|
| 1 | Votes | 19.34% |
| 2 | Genre | 16.65% |
| 3 | Actor 1 | 13.49% |
| 4 | Director | 12.89% |
| 5 | Actor 3 | 12.88% |
| 6 | Actor 2 | 12.63% |
| 7 | Duration | 12.12% |
👍 Votes emerged as the most influential feature for predicting movie ratings, followed by Genre and Lead Actors.
- 🐍 Python
- 🐼 Pandas
- 🔢 NumPy
- 📊 Matplotlib
- 🤖 Scikit-Learn
- 📓 Jupyter Notebook
- Implement One-Hot Encoding
- Hyperparameter Tuning
- Try XGBoost Regressor
- Feature Engineering using Actor/Director historical ratings
- Include Budget, Revenue, and Review Features
Waheed
Machine Learning Internship Project
⭐ If you found this project useful, consider giving it a star!