Abstract: Taking the shale layer of the third member of Hetaoyuan formation in Biyang sag of Nanxiang basin as an example, the organic carbon contents from 5 wells were predicted by means of improved methods of ΔlogR, CARBOLOG and multivariate regression analysis,and then the predicted results were compared with the limited core analysis data. The results show that the multivariate regression analysis is the best among these three methods in predicting accuracy and applicability, by which the well logs or related categories that well reflect the variation of organic carbon content can be selected, such as neutron log, etc., also, some log data that are influenced by special minerals can be avoided for the reason that high content of pyrite in shale may lead to lower resistivity and higher density, which may weaken the responses to the organic matter for these logs, hence increasing the objectivity and reliability of the predicted results