论文部分内容阅读
针对目前预测农作物产量只利用年产量或其变形,而没有考虑气象因素对产量预测的影响导致误差偏大的问题,在基于商空间粒度理论框架下的农作物产量预测中,考虑气象因素如光照、平均气温、降水量对农作物产量的影响,提出利用支持向量机方法构造模型对气象时间序列进行数据挖掘(产量预测)。粒度分析和实验结果表明:混合粒度预测模型不仅降低了问题求解的复杂性,而且误差较低,其预测值平均绝对百分误差为0.884 9,均方根误差37.3,希尔不等系数为0.004 4,与其他预测模型相比误差最小。基于商空间理论的支持向量机产量预测模型可较好地应用于产量预测中。
In view of the fact that crop yield is currently forecast using only annual yield or its deformation without considering the impact of meteorological factors on yield forecasting, the error is large. In the forecast of crop yield based on the quotient space granularity theory, the meteorological factors such as light, Average temperature and precipitation on the crop yield, this paper proposes to use the SVM method to construct the model to mine the meteorological time series (yield forecast). Particle size analysis and experimental results show that the mixed particle size prediction model not only reduces the complexity of problem solving, but also has a lower error. The average absolute percentage error of prediction is 0.884 9, root mean square error is 37.3 and Hill inequality coefficient is 0.004 4, compared with other prediction model, the minimum error. SVM output forecasting model based on quotient space theory can be applied to production forecasting well.