Abstract: At present, there are time consuming and poor effect of forecasting in the method of Pr}decison on fruit diseries. A new forecast method SVR-D1. 1 based on regression was proposed in this paper, features reduction was conducted by using SVM. The method can select keen correlative features repeatedly and construct its dynamic optimizing parameters. Relativity statistical analysis was conducted between the real data and the forecasting data of Dangshansu pear scab, which show the method is more super and more valid than the current method in efficiency and precision of forecasting the occurrence tendency of Dangshansu pear scab. I}he experiment showed that the approach has obvious advantage on fitting degree, the reasoning efficiency and accuracy.

Key words: Support vector rcgression, Fcature selection, Disease forecasting, Pear scab