This project applies Principal Component Analysis (PCA) to analyze consumer shopping behavior using a dataset of 3,900 customer observations and 16 behavioral/demographic attributes.
Developed as part of the MDA3083 Predictive Analysis coursework (Group 5).
- Domain: Retail & Customer Behavior Analytics
- Dataset Size: 3,900 records, 16 variables
- Primary Methodology: Principal Component Analysis (PCA)
- Tools Used: Python (
scikit-learn,pandas,matplotlib,seaborn), Power BI / Dashboard
- Data Preprocessing: Clean dataset, remove non-behavioral features, and encode categorical variables using
OrdinalEncoderand standard scaling. - Dimensionality Reduction: Reduce dataset complexity from 13 features to 9 principal components while retaining 79.52% of total variance.
- Behavioral Insights: Identify key drivers behind customer purchasing decisions, subscription responsiveness, and seasonal trends.
- PC1 (Customer Incentive Profile - 16.57%): Captures promo-responsive and subscription-oriented customers.
- PC2 (Magnitude & Sizing Trade-off - 8.46%): Links fit size and higher spending on heavy apparel (coats, boots).
- PC3 (History Contradiction - 8.18%): Identifies new, younger shoppers making first-time purchases.
- PC4 (Brand Engagement Profile - 7.99%): Connects high previous purchases with top review ratings for brand loyalists.
- Total Gross Revenue: $233K
- Total Transactions: 3,900
- Average Order Value (AOV): $59.76
- Average Review Rating: 3.75 / 5.0