To find an accurate and effective frequency prediction method is the key to improve HF communication. With analysis on ionosphere propagation features, and on the basis of ionospheric data in Sanya, Hainan, the Deep Learning method is applied to the ionospheric parameter prediction, then the method is compared with the Elman neural network prediction method. The prediction of the Deep Learning method is more close to the real value with less training cycles, and more accurate scientific basis is provided for the HF frequency prediction.