Machine Learning-Based Compost Maturity Prediction Using Multisensor IoT Data

Abstract

Conventional composting process monitoring still faces various limitations, particularly in determining the level of compost maturity, which generally relies on visual observation and operator experience. The development of Internet of Things (IoT) technology enables real-time multisensor data acquisition during the decomposition process, but most research still focuses on developing monitoring systems without utilizing the resulting data to build predictive models. This study aims to develop a Machine Learning-based compost maturity prediction model using a multisensor dataset obtained from an IoT-based Smart Composting Bin system. The data used includes environmental parameters that affect the composting process, such as temperature, humidity, media moisture, and gas concentrations that are recorded periodically. The research stages include data preprocessing, handling missing data and outliers, normalization, feature selection, Random Forest model training, and performance evaluation using metrics appropriate to the prediction target type. To increase model transparency, this study also applies Explainable Artificial Intelligence through SHAP analysis to identify the contribution of each sensor parameter to the prediction results. The expected results show that the Machine Learning approach is able to predict the level of compost maturity with high accuracy while identifying the most influential environmental factors during the decomposition process. This research contributes to the utilization of IoT data not only as a monitoring medium, but also as a basis for intelligent decision-making in data-based organic waste management systems, thereby supporting the implementation of smart waste management and the concept of a circular economy.