The quantitative prediction of the mineral composition, porosity, and kerogen content of shales is significant for the evaluation of shale oil and gas potential and the hydraulic fracturing process. We have developed a new method for the shale’s components prediction (SCP-) by combining the back-propagation (BP) neural network and an improved method based on conventional logs. First, we constructed and calibrated the shale fraction model according to the volume of the minerals, kerogen, and porosity determined through laboratory analyses. Subsequently, we calculated the kerogen volume by the combination of the improved technique and the conversion equation between the kerogen volume and the organic carbon content. Finally, the BP neural network was trained with the input parameters of the kerogen volume and the sensitive logs, and the output parameters of the mineral volume (clay, silicate, carbonate, and heavy minerals) and porosity. We used the cross validation method to optimize the structural parameters of the BP neural network. The SCP- method, which is a nonlinear technique, takes into consideration the influence of the organic carbon of the residual oil on the calculation of the kerogen volume. We successfully implemented the SCP- method to evaluate the shale components of well Shen 352 in the Damintun Sag, China. The evaluation results of the SCP- method are in good agreement with the measured core sample properties and mineral composition derived from Schlumberger elemental-capture spectroscopy logs, confirming the accuracy and reliability of the SCP- method in predicting the mineral composition, porosity, and kerogen content in shale.