Multilabel Machine Learning-Based Detection of Allergens in Food Recipes

Food allergens are substances that can trigger allergic reactions or intolerances in some individuals. According to recent data, the prevalence of food allergies worldwide ranges from 10% to 40%. In Indonesia, around 20% of children in their first-year experience reactions to the foods given to them. This research focuses on developing a machine learning model to detect allergens in food recipes, utilizing K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP) methods with a multilabel classification approach. The primary challenge is the difficulty of identifying hidden allergens in the diverse ingredients of recipes, which can be harmful to individuals with food allergies. This study utilizes 15,823 data points from a food recipe dataset, labeled both manually and automatically with five main types of allergens. After data Preprocessing and feature extraction using TF-IDF, the models were trained and tested with an 80:20 ratio. Results indicate that the SVM with hyperparameter tuning on the manually labeled dataset performed the best across all allergen types, achieving average F1-Scores of 0,9776.