【摘要】 【目的】研究信息扩散近似推理方法在年降水量预测中的应用,并对比其他方法分析其推广前景。【方法】通过分析年降水量时间序列的特性,提出了基于当前趋势以及相邻年份降水量的年降水预测规则。利用信息扩散近似推理描述年降水量间的复杂非线性关系,并以某灌区长系列降水资料为样本进行实例计算。【结果】信息扩散近似推理方法预测效果较好,该方法误差绝对值和为1.673,小于人工神经网络和线性自回归方法的统计结果。【结论】信息扩散近似推理可将样本点转换成模糊集,部分弥补了由于数据的不完备性所造成的信息空白,并可将矛盾模式转换成兼容模式。通过与传统预测方法相比较发现,该模型能够很好地光滑样本数据以及较好地发掘知识,有较高的预测精度和推广应用价值。更多还原

【Abstract】 【Objective】The method of applying information-diffusion approximate reasoning to predict annual precipitation was studied and compared with other methods in order to analyze the application prospects.【Method】According to features of precipitation time-series,prediction rules were suggested based on current tendency and the neighbor year precipitation.This may help the information-diffusion approximate reasoning describe the complex nonlinear relation in precipitation data.Applying the rules a precipitation time-series in an irrigation area as an example,the forecasting result was obtained.【Result】The information-diffusion approximate reasoning method had fewer errors and a better effect on precipitation prediction than artificial neural network and linear autoregressive method.【Conclusion】The method can transfer sample points to fuzzy sets and take advantage of more information,and even may switch conflict mode to compatible mode.Results indicate that annual precipitation prediction with information-diffusion approximate reasoning model is good at the mining of uncertain knowledge,and can find out more information and make the data series more smoothly than traditional methods.In fact,this is a new data mining model for time-series.Any similar problem can be dissolved better by the method.It is significant to spread this method.更多还原