Outlier Detection and Data Cleaning 

Visualized boxplots for detecting extreme values.

• Significant outliers in features such as Income and Age could distort clustering if left untreated.

• Total Spend is heavily impacted by a few high-spending customers, possibly VIP clients or bulk buyers.

Heatmap of Numerical Features 

• Income and Total spending show a strong positive correlation, confirming that higher-income customers tend to spend more.

• Product categories like Wines, Meats, and Gold also show positive correlations with Total spending, indicating preferred items among high-spenders.

• Age and Recency have little to no correlation with spending, implying that younger or older age alone is not a strong predictor of customer value.

Feature Relationship Analysis 

• A positive correlation between income and total spending indicates that income is a strong predictor of purchasing behavior.

• Weak or no correlation between Age and Recency, suggesting that recent purchases are not strongly age-dependent.

• Parental status does not distinctly separate customer behaviors, which hints that other factors (e.g., income, lifestyle) are more influential.

Elbow Method 

The distortion score (within-cluster sum of squares) was plotted against the number of clusters (k) to determine the optimal k-value.

• k = 5 is the optimal choice as the curve starts to flatten, indicating diminishing returns beyond this point.

• Adding more clusters after k = 5 provides minimal improvement and could lead to overfitting.