<![CDATA[Fruit quality is an important factor in selecting fruit for consumption because it affects consumer health and satisfaction. Identification of fruit quality has become the focus of research, and one of the approaches used is a non-destructive approach through measuring the gases produced by the fruit. Machine learning can be used to process this gas data and build system models that can classify fruit quality. This research discusses the application of the DCS-OLA and Stacking dynamic ensemble learning algorithms to build a fruit quality detection system model. The basic methods used to build models are Logistic Regression, Decision Tree, Gaussian Naïve Bayes, and Mul-ti-Layer Perceptron. The fruit used is mango with a shelf life of 7 days and Srikaya (sugar apple) with a shelf life of 4 days. The condition of the initial dataset is unbalanced. The research results show that trimming the mango dataset to only 4 days according to the shelf life of sugar apple helps reduce the difference in shelf life between the two. Then jittering and balancing techniques are used to increase and balance the number of datasets between the two types of fruit. High accuracy is achieved by the DCS-OLA ensemble and stacking ensemble by combining the basic methods of Logistic Regression and Decision Tree, especially in balanced dataset conditions. In conclusion, the use of ensemble learning in detecting fruit quality has great potential for real-world applications. However, further validation is needed with larger datasets and a wider variety of conditions.]]>
Copyrights © 2024
