With the vigorous development of network technology, optical networks are developing in the direction of ultra-high-speed and large-capacity. Since the structure of the optical network is becoming more complex. the optical signal will inevitably suffer various impairments in the optical network. In optical communication equipment, the most important parameter of the physical layer is the Optical Signal-to-Noise Ratio (OSNR). Its value directly determines whether the service can operate normally. Once the requirements are not met, it will cause transmission errors or failures, reducing service quality, and transmission consumption. This paper proposes an accurate and efficient OSNR prediction method for optical communication nodes. By combining analytic prior knowledge methods and deep learning-based posterior knowledge methods, an optical communication node OSNR based on hybrid machine learning algorithms are proposed. The prediction method uses prior knowledge to reduce the training cost of neural networks and provides high-accuracy OSNR predictions. The result shows that the method proposed in this paper can provide high-accuracy machine learning models under more demanding conditions.