Intelligent Smart Cooking: Predictive Cooking Time Model Using Machine Learning and IoT

Intelligent Smart Cooking: Model Prediksi Cooking Time Menggunakan Machine Learning dan IoT

Abstract

Smart cooking systems have gained significant attention as part of the growing smart home ecosystem, yet most existing solutions rely on static, rule-based thresholds that lack adaptability to variations in food type, weight, and cooking conditions. This study proposes an Intelligent Smart Cooking system that integrates Internet of Things (IoT) sensing with a machine learning–based predictive model to estimate cooking time in real time. Temperature data were collected from 180 cooking sessions using a DS18B20 sensor, while MQ-series gas sensors supported safety monitoring. A dataset containing temperature curves, heating rates, food mass, and water volume was constructed and used to train three regression models: Multiple Linear Regression, Support Vector Regression, and Random Forest Regression. Experimental results show that Random Forest achieved the best performance with an MAE of 18.93 seconds and an R² of 0.954, demonstrating strong capability in capturing nonlinear cooking behavior patterns. The trained model was deployed into the IoT system to enable predictive cooking automation, real-time flame control through a servo motor, and hazard prevention via gas detection. User evaluations also indicated high usability and reliability of the system. The findings highlight the potential of combining IoT and machine learning to improve accuracy, safety, and efficiency in next-generation smart kitchen technologies.