X oilfield in the South China Sea is characterized by sparse well pattern and typical delta facies deposits with intercutting sand bodies. The use of seismic attribute fusion technology to predict the planar distribution of sand bodies in dense well network areas has shown good application results in many oil fields, but it is not effective in the condition of large well spacing and sparse well pattern in offshore areas. When thick sand bodies are developed above and below the target layer, the influence of surrounding rock on the target layer should be considered. Therefore, the machine learning attribute intelligent fusion sand body prediction method based on reducing the interference of surrounding rock is adopted for research. Firstly, a variety of seismic attributes of the original seismic data of the target horizon are extracted, from which the sensitive attributes that can characterize the thickness of the sand body are selected. Secondly, the optimal properties of the surrounding rock in the wavelength range of 1/4 above and 1/4 below the target layer are extracted. Finally, a machine learning algorithm based on support vector regression is used to establish a complex mapping relationship between the optimal attributes of the target layer and surrounding rock and the thickness of the target layer sand body, in order to improve the accuracy of sand body prediction. The research results indicate that the correlation between the fusion result and the thickness of the sand body is significantly improved after intelligent attributes fusion of the target layer by reducing the interference of surrounding rock and the distribution pattern of sand body is more consistent with geological understanding. This method can be used for reference in improving the accuracy of seismic prediction for similar oil fields.