This study presents a novel method for monitoring the fermentation process of tempeh using a low-cost electronic nose (E-nose) and multispectral sensors, combined with machine learning models. By integrating support vector machine (SVM), random forest (RF), and k-nearest neighbors (k-NN) algorithms, the data fusion from these sensors significantly improved the classification accuracy of fermentation stages, reaching up to 98.26%. With the fusion dataset strategy, the support vector regression (SVR) prediction model achieved the highest \(\:{\text{R}}_{\text{T}}^{2}\) and \(\:{\text{R}}_{\text{p}}^{2}\) values of 0.99 and 0.96, with the lowest \(\:{\text{R}\text{M}\text{S}\text{E}}_{\text{T}}\:\) and \(\:{\text{R}\text{M}\text{S}\text{E}}_{\text{p}}\) scores of 2.34 and 4.15 \(\:\text{c}\text{f}\text{u}/\text{g}\) , outperforming the individual dataset model. Gas chromatography-mass spectrometry (GC-MS) identified 75 volatile compounds that contributed to distinct odor profiles at each fermentation stage, highlighting the system’s ability to provide detailed insights into the process. This integrated approach offers enhanced monitoring of tempeh fermentation, improving quality control and consistency compared to conventional techniques. The findings demonstrate the potential of combining E-nose and multispectral sensors with machine learning to achieve high accuracy in food fermentation classification. However, further development is needed to create user-friendly, cost-effective systems suitable for large-scale industrial applications. This study contributes to advancing sensor technology in food production, paving the way for more efficient and scalable solutions.
