多变量时序回归树的黄瓜产量预测模型  

Yield prediction model of cucumber based on multivariate time seriesreg ression tree

作　　者：陈湘芳[1] 陈明[1] 冯国富[1] 池涛[1] 

机构地区：[1]上海海洋大学信息学院,上海201306

出　　处：《计算机工程与设计》2012年第1期407-411,共5页Computer Engineering and Design

基　　金：国家863高技术研究发展计划基金项目(2007AA10Z238)

摘　　要：为提高黄瓜产量预测能力,将黄瓜的全生育期作为研究对象,提出了一种基于多变量时间序列的回归树预测模型。将黄瓜全生育期采集到的多环境变量作为一个时间序列,根据黄瓜的生长期将整个时间序列划分成多个子序列,最后利用回归树算法对时间序列样本进行建模。该研究以戴多星这一黄瓜品种为实验对模型进行验证,实验结果表明,该模型能准确地预测黄瓜戴多星的产量,平均单株产量误差不大于0.05kg。To improve the predictive ability of cucumber production, taking the whole period of it as the research object, a yield prediction model based on multivariable time series regression tree is presented. Multi-environmental variations collected to do as a time series during the growth period of cucumber, and according to the growth periods of cucumber, the entire time series are divided into multiple sub-sequences, and are modeled by time-series regression tree algorithm. The Deltestar that is a kind of cucumber, taken as model to text. The results show that this model can predict the cucumber yield accurately and the average inaccuracy of yield per plant is less than 0. 05 Kg.

关 键 词：黄瓜产量 预测模型 时间序列 决策树 产量预测 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]