With the rapid development of computer network technology, the network scale is becoming larger and larger, the structure is becoming more and more complex, and communication failures occur frequently, which seriously affects the stability and reliability of the network. Traditional fault prediction methods are difficult to meet the needs. This paper focuses on the application of deep learning algorithms in computer network communication fault prediction. By constructing a fault prediction model based on deep learning, using data preprocessing, model training and optimization steps, and combining it with actual network data for experiments. The research results show that the traffic rate of the proposed method fluctuates between 0.65 and 0.92 Mbps, with no obvious monotonic trend, and the packet loss rate fluctuates significantly, with a minimum of 0.02% and a maximum of 0.45%. This provides strong support for network operation and maintenance personnel to take measures in advance and reduce the impact of failures, and helps to ensure the stable operation of computer networks.