<p>Shale content is an important parameter to interpret the lithology and physical properties of reservoir. It is of great significance for exploration and development of oil and gas reservoir to accurately calculate shale content. At present, we usually use the empirical formula method or the cross plot method to predict the shale content, which can not make full use of other logging information. Traditional machine learning methods, while capable of establishing nonlinear relationships between well-log parameters and shale content, exhibit limited predictive accuracy and face challenges in achieving further improvements. Deep learning technology can automatically extract complex high-dimensional nonlinear features from the data, and make full use of the shale content features reflected by various logging curves. Based on the natural gamma, neutron, natural potential, resistivity and acoustic loggings as the characteristics of deep learning, this paper constructs a deep neural network model for shale content prediction, which uses ReLu activation function, Adam optimization algorithm and dropout regularization method. The validation using actual shale content prediction data demonstrates the feasibility and potential application of deep learning technology in reservoir parameter prediction.</p>