To see the full analysis, code implementation, and final prediction results, please click on the file: vgs-data-analysis.ipynb.
This project focuses on Predicting Nintendo Game Ratings by analyzing historical data from Metacritic. The goal is to identify patterns in professional critic scores (Meta Score) and user ratings to predict how future or imaginary titles might be received.
The analysis utilizes a dataset of Nintendo games spanning multiple generations—from the Nintendo 64 to the Switch.
-Data Cleaning & Preprocessing: Handling missing values, filtering out unreleased or canceled titles, and converting temporal data for analysis.
- Feature Engineering: Utilizing one-hot encoding for categorical variables such as game platforms and ESRB ratings.
- Machine Learning Models:
- Linear Regression: Used as a baseline to explore the relationship between features and Meta Scores.
- Ensemble Methods: Implementation of
DecisionTreeClassifier,BaggingClassifier, andRandomForestClassifierto improve prediction accuracy and robustness.
- Natural Language Processing (NLP): Leveraging
Word2Vec(via Gensim) and PCA (Principal Component Analysis) to analyze and vectorize game titles, allowing the models to "understand" naming conventions in relation to scores.
The project culminates in a Game Score Predictor. This tool allows you to input any game name—whether it exists or is entirely made up—and receive a predicted Meta Score and User Score based on historical trends.
- Language: Python
- Libraries: pandas, numpy, scikit-learn, matplotlib, gensim
- Environment: Jupyter Notebook