The following questions are based on the Turkish E-commerce…
The following questions are based on the Turkish E-commerce retail data that we have already used in class. We used K-Means clustering to see whether the 5,000 orders in the e-commerce dataset fall into natural groups. Seven numeric inputs are used. Unit price and quantity were left out because total amount already reflects them. They are: customer age, discount amount, total order amount, session duration, pages viewed, delivery time, customer rating. Categorical variables, such as product category, city and payment method, and the yes/no returning-customer flag, were not used to form the clusters. They were only used afterward to describe the groups. Before clustering, all seven variables were standardized, which converts each to a z-score with a mean of 0 and a standard deviation of 1. The model was run separately with K = 2, 3 and 5 clusters. For each solution, the performance metrics are provided below. Number of Clusters( K) inertia (within cluster sum of squares) silhouette 2 30767.7 0.486 3 27247.967 0.161 4 24664.69 0.149 5 22539.371 0.15 6 20770.867 0.157 7 19035.324 0.154 8 17677.817 0.153
Read Details