【摘要】 以"竹藤种质资源数据库"中126个竹种为基础数据,应用K均值聚类算法(K-means算法)、小批量K均值聚类算法(Mini Batch K means算法)、凝聚层次聚类算法,依据竹种酸不溶木质素、苯醇提取物、综纤维素、基本密度、纤维长度、纤维双壁厚、纤维壁腔比7个特征对竹种筛选分级,遴选优良造纸竹种,分析聚类算法对竹藤种质资源的聚类有效性。结果表明:K均值聚类算法具有最高的卡林斯其-哈瑞本斯(Calinski_Harabaz(CH))指标值(63.568),聚类效果最好,126个竹种分为3级。其中:Ⅰ级造纸竹种,包含糯竹、空竹等51个竹种,适合制作高级纸张;Ⅱ级造纸竹种,包括毛竹、思劳竹等62个竹种,适合在排污能力强的造纸厂做纸浆造纸原料。K均值聚类算法是竹种造纸适应性分析的高效可靠的方法,具有较好的应用前景。更多还原

【Abstract】 With 126 bamboo species in the Bamboo and Rattan Germplasm Resources Database, K-means clustering algorithm, Mini Batch K means clustering algorithm and Agglomerative Hierarchical clustering algorithm were applied, according to the seven characteristics of acid insoluble lignin, benzene alcohol extract, holocellulose, basic density, fiber length, fiber double wall thickness and fiber wall cavity ratio, bamboo species were screened and graded, excellent papermaking bamboo species were selected, and the clustering effectiveness of clustering algorithm on bamboo and rattan germplasm resources was analyzed. K-means clustering algorithm has the highest Calinski Harabaz index value(63.568), the clustering effect is the best, and 126 bamboo species are divided into three levels. Among them: Grade I papermaking bamboo species, including 51 bamboo species such as Cephalostachyum pergracile Munro and C. latifolium Munro, which are suitable for making high-grade paper; Grade II papermaking bamboo species, including 62 bamboo species such as Phyllostachys edulis and Schizostachyum pseudolima McClure, are suitable for making pulp and papermaking raw materials in paper mills with strong sewage discharge capacity. K-means clustering algorithm is an efficient and reliable method for adaptability analysis of bamboo papermaking with a good application prospect.更多还原