Smartphone Models Insights
Problem Statement
Can a phone's spec sheet predict how users will rate it? This analysis examines hundreds of smartphone models to uncover which hardware specifications - RAM, camera resolution, battery, screen size - most strongly predict user satisfaction, and builds a multi-class classifier to predict rating tiers from specs alone.
Methodology
Exploratory Data Analysis
The first analysis compared price segments against features. Three price segments were defined:
- Low: Price ≤ $130
- Mid: $130 < Price ≤ $350
- High: Price > $350
Ratings Correlation
Spearman's rank correlation was used to test the significance of feature relationships with user ratings.
Results
Classification Model
A multi-class CatBoost classification model was built to predict user rating categories:
- Low: Rating ≤ 69
- Medium: 70 < Rating ≤ 79
- High: Rating > 80
Model Accuracy: 85% (vs. 51% baseline), confirming that smartphone specifications are strong predictors of user ratings.
Feature Analysis
Next Steps
- Incorporate temporal data to track how feature importance shifts across smartphone generations
- Add brand-level analysis to understand how manufacturer reputation affects ratings
- Explore deep learning approaches for more nuanced feature interaction modeling