ADS Capstone Chronicles Revised

20

Confusion Matrix for Multiclass Classifier for Decision Tree

personalized recommendations. Similarly, the KNN model’s performance metrics revealed the

6 Discussion The results of the study highlight the effectiveness of different recommendation models - SVD, NMF, and KNN - in enhancing the accuracy and relevance of product suggestions within the company’s e-commerce platform. Notably, the SVD model showed significant improvements in both RMSE and precision@5 as the number of components increased, suggesting that a deeper understanding of user-item interactions is crucial for generating

optimal number of neighbors (k=2) provided the best balance between prediction accuracy and recommendation relevance. These findings underscore the importance of selecting and tuning the appropriate model parameters to align with specific business goals, such as improving customer satisfaction and increasing sales. When comparing our findings to existing studies, our results are consistent with the literature that validates the effectiveness of SVD and NMF in collaborative filtering tasks. Previous research also highlights the sensitivity of KNN-based

64