Classification of Image Quality of Tuban Legen Beverages Using Convolutional Neural Network (Cnn) Method Based on Random Forest

This study focuses on developing a classification method for the quality of Tuban legen beverage using a combination of Convolutional Neural Network (CNN) and Random Forest. The siwalan plant (Borassus flabellifer L.), commonly found in Tuban Regency, East Java, produces sap that is susceptible to fermentation and microbial damage. Therefore, it is essential to process legen properly to maintain its taste and quality. The dataset used consists of 1280 images of legen beverage categorized into two classes: Steril and Campuran. The research process involves data collection, preprocessing, and splitting the dataset into training, validation, and testing sets with various ratios. CNN, specifically the VGG16 architecture, is used for feature extraction from images, while Random Forest serves as the classification algorithm. The model is optimized using GridSearchCV and evaluated with metricssuch as accuracy, precision, recall, F1-score, and confusion matrix. The results indicate that the combination of CNN and Random Forest can achieve optimal accuracy of up to 100% for certain data configurations, both for images with and without background, with a data ratio of 80% training, 10% validation, and 10% testing. Data augmentation is applied to enhance model variation and generalization, contributing to improved classification accuracy. Although these results demonstrate the model's ability to recognize patterns, the study suggests potential for further improvement. It is recommended to explore other classification methods, conduct deeper hyperparameter optimization, and apply transfer learning to enhance model performance. Additionally, further research is needed to understand the impact of background on classification and how to address it. This study makes a significant contribution to the development of a quality classification system for legen beverages by utilizing more robust and accurate deep learning technology. These findings can serve as an important reference for future research in image classification and food product quality, providing insights for developing more efficient and effective methods.