Total organic carbon (TOC) is a very important parameter in logging evaluation for source rocks and is important to determine the hydrocarbon generation potential of source rocks and the prediction of adsorption gas content. Taking the source rocks of block H in Canada as example, the correlation relationship between TOC and the conventional logging is analyzed on the basis of the conventional logging, the laboratory analysis TOC are collected. Based on the results of correlation analysis results, GR, RHOB, NPHI, PEFZ, URAN and ratio of deep resistivity to shallow resistivity are selected as the basic parameters, and BP neural network prediction model for TOC content is established. The TOC content predicted by the BP neural network model is also compared with the results from laboratory. The application of BP neural network prediction model of TOC in block H of Canada shows that the neural network prediction model of TOC based on logging has very high accuracy. It also shows a good application prospect. It is of great significance for analyzing the hydrocarbon generation potential and the gas content prediction of source rocks in block H.