基于决策树的香蕉气候适宜性区划

DIVISION OF BANANA FOR CLIMATIC SUITABILITY BASED ON A DECISION TREE

香蕉是华南重要出口水果,开展香蕉气候适宜性区划可为合理利用资源和优化香蕉布局提供科学依据。针对传统气候区划方法过于依赖知识经验、不具学习能力的不足,尝试使用ID3决策树方法进行广西香蕉气候适宜性区划。根据香蕉生长气候需求,选择年平均气温、年降雨量、≥10℃活动积温、年极端最低气温、日平均气温≤8℃连续天数和年平均日照时数等6个气候因子,利用广西90个气象站1971—2000年观测资料,结合台站空间信息建立栅格推算模型,把各个因子推算为l km×l km栅格数据,并根据香蕉区划指标对各个因子进行离散化运算。通过实地考察并结合离散化后的因子栅格数据,获取香蕉气候适宜性样本点,并使用ID3算法建立决策树,结果总体精度达到92.5%。区划结果表明,ID3决策树算法适用于香蕉气候区划。

Banana is an important export-oriented fruit in South China. Division of banana for climate suitability provides scientific basis for rational use of natural resources and optimization of banana distribution. To overcome the shortcomings of traditional methods of climate division that over-rely on knowledge and experience and have no ability to learn, a approach based on a so-called ID3 decision tree method is provided. Annual mean temperature, annual precipitation, accumulated temperature of ≥ 10 ℃, annual extreme minimum temperature, number of consecutive days with average temperature of ≤ 8 ℃, and mean annual sunshine duration are taken as climatic factors to build Grids Reckoning Models. Climate data from 1971 to 2000 and geographical information data of 90 weather stations in Guangxi are used to build models, with each of the climate factors projected as 1 km^l km raster data and discretization operation carried out on each of the climate factors based on an index of climatic suitability division. Combined with discreted raster data with fieldwork data, samples of banana for climate suitability division are used to build a decision tree based on the ID3 algorithm and the total classification accuracy is 92.5%. Results shows that the ID3 decision tree method is suitable for banana climate division.