泛函连接网络计算软件及其在生物多样性研究中的应用

Functional Link Artificial Neural Networks and agri-biodiversity analysis

针对农田生物多样性分析的需要 ,研制出泛函连接网络 (FLANN)计算软件。该软件由 7个Java类和 1个HTML文件组成 ,是一种Internet在线计算工具 ,可运行于多种操作系统和Web浏览器上 ,并在各种类型的PC及工作站上使用 ,可读取多种类型的数据库文件。对水稻田昆虫生物多样性的两组取样调查数据Zmar18和Zapr15 ,用生物多样性工具软件LUMP和非监督分类 -离差平方和聚类法进行统计归纳及分类 ,分别划分为 2 1个和 2 0个功能群 ,各包含 6 0个样本。以FLANN计算软件对昆虫生物多样性进行了模式分类分析。结果表明 ,泛函连接网络的模式分类及预测与实际测查结果吻合良好。泛函连接网络Internet在线计算软件的应用可促进生物多样性数据采集和分析的规范化 ,有利于数据和信息共享 。

A computational software using Functional Link Artificial Neural Network (FLANN) was developed. This software includes seven Java classes and one HTML file. It is an Internet online computational software which can be run on various operational systems and web browsers, and can read various databases. Two data sets of rice insect biodiversity were lumped and clustered employing the biodiversity software LUMP and hierarchical clustering, which resulted in two sets of training samples each with 60 samples and, 21 and 20 functional groups, respectively. Ten samples to be predicted were prepared for each set of data. FLANN was used to make pattern classifications. The simulated and predicted results with FLANN software gave good fitness to the investigated results. The development of FLANN software also provides a new tool for the development of Internet intelligent analytical system for biodiversity studies.