Abstract Discrimination of wild and cultivated mushrooms is important for their quality assessment. In this study, data fusion of infrared and ultraviolet spectroscopies combined with support vector machine or random forest were applied for the discrimination of wild and cultivated W. cocos and P. portentosusis. For the low- and mid-level data fusion classification models, the accuracies of validation set are 90.91% and 98.70%, respectively. However, the parameter c is too large which means the models are unreliable. For the high-level data fusion, the results showed that the accuracy of validation set is 100% and the sensitivity and specificity increased significantly compared to the single technique. Therefore, high-level data fusion combined with random forest is an effective method for the discrimination of wild and cultivated grown W. cocos and P. portentosus, and may be applied for protecting the steady development of the edible mushroom market.

中文翻译：

---

使用支持向量机和随机森林模型的紫外和红外光谱融合，用于鉴别野生和栽培蘑菇

摘要 野生和栽培蘑菇的鉴别对于它们的质量评估很重要。本研究采用红外和紫外光谱数据融合结合支持向量机或随机森林的方法对野生和栽培的椰枣和红斑海棠进行鉴别。对于低级和中级数据融合分类模型，验证集的准确率分别为 90.91% 和 98.70%。但是，参数 c 太大，这意味着模型不可靠。对于高层数据融合，结果表明验证集的准确率为100%，灵敏度和特异性较单一技术显着提高。因此，结合随机森林的高级数据融合是区分野生和栽培生长W的有效方法。