Page 172 - Emerging Trends and Innovations in Web-Based Applications and Technologies

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
Figure 6 presents the loss graph of the proposed model. As expected, the training loss is initially high, representing the
learning phase. The validation loss begins to minimize as the model progresses, demonstrating that the system is adjusting and
improving with each epoch. The model successfully adapts to variations in item categories within Quick Mart’s marketplace,
reducing the loss for real-time classification and boosting overall marketplace functionality.

Figure 6: Confusion Matrix
The confusion matrix provides crucial insights into the true and predicted labels for the items categorized within Quick Mart’s
platform. As shown in Figure 6, the classifier has effectively categorized items into 11 classes, including regular, faulty,
refurbished, and other product categories. While the model performs well across most categories, a small number of items from
specific categories are misclassified. This is typical in complex marketplaces, where product variations are numerous and
require continuous optimization of the algorithm. To assess the system's performance, key metrics such as accuracy, precision,
and recall were evaluated for each product category, ensuring reliable item identification.

Figure 7: Experimental Results
Figure 7 demonstrates that as the number of epochs increases, the accuracy of Quick Mart’s smart solutions improves, leading
to a noticeable reduction in testing set loss. This trend suggests that the platform’s algorithm learns and adapts over time,
leading to more reliable second-hand product identification and transaction processes. The continuous reduction in loss and
improvement in accuracy highlights the system's capability to handle real-world variability, offering a seamless experience for
users engaging in the second-hand marketplace.

IJTSRD | Special Issue on Emerging Trends and Innovations in Web-Based Applications and Technologies Page 162