Deep Learning for Prediction Contents High Calories on Fast Food Menus

Abstract

The rising consumption of fast food, often high in calories, has contributed to increasing rates of obesity and non-communicable diseases. This study applies a Deep Learning approach to predict high-calorie content in fast food menus using numerical nutritional data. The dataset comprises 515 menu items with 17 attributes, including calories, fat, cholesterol, protein, sugar, sodium, vitamins, and food categories. Data preprocessing involved handling missing values, normalization, categorical encoding, and splitting the dataset into training and testing sets with an 80:20 ratio. A Multi-Layer Perceptron (MLP) model with three hidden layers was implemented, employing ReLU activation and a sigmoid output for binary classification. Model performance was evaluated using accuracy, precision, recall, F1-score, and ROC-AUC, and compared with Logistic Regression, SVM, and Random Forest. Results show that the MLP achieved an accuracy of 0.89, precision of 0.87, recall of 0.91, F1-score of 0.89, and ROC-AUC of 0.94, outperforming classical methods. The high recall underscores the model’s capability to reliably detect high-calorie items. These findings indicate that deep learning effectively captures complex relationships among nutritional attributes compared to traditional algorithms. This research contributes to the advancement of intelligent nutrition-based systems that can be integrated into mobile or web applications to support healthier lifestyle decisions. Future work may expand to larger and more diverse datasets, explore multimodal data such as images, and assess the broader health impacts of high-calorie fast food consumption.