Abstract: In this paper, we briefly discuss two types of logging prediction methods, namely multi-parameter regression and multi-mineral model. We determined the mineral composition of shales from the upper section of the third member of Hetaoyuan Formation in the deep sag area of the Biyang Depression using these methods and compared this with measured data. The results showed that the major components are quartz and clay minerals, with an average of 24.2% and 23.4%, respectively. The contents of plagioclase, calcite, dolomite, potassium feldspar and pyrite reduce in a descending order. The content of pyrite is about 3.1% on average. Zeolite, gypsum and other minerals are visible only in a few samples. Measured data validation proved that the predictions by logging using these two models are accurate, especially the multi-mineral model, which has an accordance rate of 0.8. The mineral composition section in well BYHF1 and the brittle mineral content in shales from the upper section of the third member of Hetaoyuan Formation showed that the multi-mineral model can predict the distribution rules and trend of minerals better. It is significant for shale oil sweet spot prediction, reservoir mineral evaluation and shale oil development.