Abstract: To establish an accurate wind direction prediction model,a data analytics-based hybrid algorithm（DAHA）was proposed.First,a classification and regression tree algorithm were employed to obtain the model inputs and reduce the inputs dimension. Second,the original data was decomposed by variational mode decomposition（VMD）into a series of subsequences for further study. And the dimensions of the subsequences were reduced based on information entropy theory to optimize the model inputs. Then,the deep belief network（DBN）was employed to model the relationship between wind direction and inputs. Finally,a least square support vector machine（LSSVM）was utilized to correct the prediction result to improve the prediction accuracy. The experimental results based on practical data illustrated the proposed DAHA algorithm can predict the ultra-short time wind direction with errors under 1%,which has significant meaning to yaw scheduling.