Page 179 - 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
The performance of the custom-designed model is illustrated in Figure 4. The blue and orange curves represent the validation
and training accuracy, respectively. The x-axis corresponds to the number of epochs, while the y-axis depicts the percentage
accuracy. As the epochs progress, the training accuracy remains consistently high, while the validation accuracy is slightly
lower. This slight gap is a common phenomenon in machine learning, often indicative of overfitting. Nevertheless, the model’s
robust performance underscores its ability to accurately categorize a diverse range of items within Quick Mart’s marketplace.

Figure 5: Model Training and Validation Loss
Figure 5 presents the loss graph for the proposed model. Initially, the training loss is high, reflecting the model’s learning phase.
However, as training advances, the validation loss steadily decreases, signifying the model’s effective adaptation to the dataset.
This reduction in loss across epochs highlights the system’s ability to handle the varied product categories within Quick Mart’s
marketplace, optimizing real-time classification to enhance the user experience and boost operational efficiency.
Classification Insights

Figure 6: Confusion Matrix

The confusion matrix in Figure 6 provides detailed insights into the classifier’s performance across 11 product categories,
including regular, faulty, refurbished, and other segments. While the system demonstrates high accuracy in categorizing the
majority of items, a few products from specific categories are misclassified. Such misclassifications are expected in complex
marketplaces with highly diverse product offerings. To address this, continuous optimization of the algorithm is essential. Key
performance metrics such as accuracy, precision, and recall were analyzed for each category to ensure the classification
process remains robust and reliable.

IJTSRD | Special Issue on Emerging Trends and Innovations in Web-Based Applications and Technologies Page 169