A comparative analysis of machine learning approaches for predicting maturity in watermelon using acoustic and quality features

This study investigated the integration of acoustic and destructive quality features for machine learning-based classification of watermelon maturity into three categories: immature, mature, and over-mature. Nine machine learning algorithms were evaluated using various performance metrics. Among these, Support Vector Machine (SVM) achieved the highest classification accuracy of 97.72 % with the feature combinations frequency-amplitude-TSS-firmness-moisture and frequency-amplitude-TSS-firmness. This was closely followed by Extreme Learning Machine with accuracies of 97.68 % and 94.78 %, respectively. SVM also demonstrated superior Matthews Correlation Coefficient values of 95.29 %. Models utilizing only the acoustic frequency achieved an accuracy of 92.47 %, while the combination of frequency and amplitude yielded an improved accuracy of 96.98 %. These findings provide a strong foundation for developing a real-time, in-situ maturity prediction systems for watermelon, enhancing quality assessment in agricultural practices.