基于ANFIS的香梨始花期预测研究

Prediction of the Beginning Date of Fragrant Pear Flowering

应用自适应模糊神经推理系统(Adaptiv network-based fuzzy inference system,ANFIS)对库尔勒地区香梨始花期进行了预测,3月平均气温、3月下旬平均气温、5 cm地温、10 cm地温、2月下旬日照、3月上旬日照、3月中旬日照、3月下旬日照8个指标为输入变量,香梨始花期单项指标作为输出,结果表明:预测的平均相对误差为2.51%,经t检验和回归分析表明预测值和实测值相差不大,相关系数为0.928 6,具有较好的一致性,模型具有较高的精度和稳定性,说明ANFIS模型可以很好地描述香梨始花期和气象因子的关系,为香梨始花期的进一步研究提供了参考。

This paper used ANFIS （Adaptiv network-based fuzzy inference system） to predict the beginning of fragrant pear flowering in Korla, Xinjiang, China. Eight meteorological factors in March were selected as input variables, and the beginning date of pear flowering as output variable. Results showed that the mean relative error of the prediction was 2.51%. t-test and regression analysis indicated that the predicted value differed just slightly from the observed value and their correlation coefficient was 0.9286. The model showed high accuracy and stability. The ANF[S model could illustrated the relation between the beginning date of fragrant pear flowering and the meteorological factors, and provided further reference in the study of beginning date of fragrant pear flowering.