Abstract: A conjugate gradient based on continuous network peak prediction algorithm and model of functional analysis were presented. The main factors affecting the network peak forecast were analyzed. The SVM model including  network traffic, network peak range and network signal strength was formulated. The principal component analysis method was adopted to realize the network peak PCA to estimate the system design. Through the largest continuous functional conjugate gradient, the peak of the internet  was predicted under the singular positive semidefinite double periodic complex analysis. The degree of contribution of the network traffic and network signal peak to the network was investigated, the network characteristic of peak was considered by using state information fusion processing to reduce the prediction error. The experimental results showed that the algorithm of network peak prediction accuracy was higher, the prediction error was controlled within 1.5%,  and the performance of the algorithm was superior