基于化学成分的烤烟香型分类研究

本研究以全国清香型、浓香型、中间香型烤烟为研究对象，以6个区域24个C3F中部叶为分类样本，采用随机森林分类算法进行香型分类，结果表明：随机森林算法对烤烟香型起到了很好的区分作用，分类正确率达到了87．5％；基于随机森林分类的同时得到了分类变量在香型分类中的重要性，烟叶常规化学成份在香型分类中的重要性从高到低依次为：总钾→总氮→总糖→还原糖→总氯→烟碱．因此，随机森林分类可以用于烤烟的香型分类，同时也可以用于判断分类变量对分类目标的重要性程度．

Study on the classification of flue-cured tobacco based on chemical components

The observation was sampled from six areas including 24 cutter leaves in China. The targets were studied based on random forest algorithm. The results shows that the random forest algorithm classify the type of flue-cured tobacco smartly and fast, meanwhile the classification accuracy could measure up 87.5%. what＇s more important, the importance of each variable could be lay out under random forest algorithm. In this case the importance of each variables indicated total K＋ 〉 total nitrogen 〉 total sugar 〉 reducing sugar 〉 total chlorine 〉 nicotine. So it proved a good algorithm to classify the type of flue-cured tobacco, besides the by-products of each variables＇ importance could be got easily.