矿产资源勘查走向深部已成为国际大趋势，深部找矿预测面临着信息难获取、难提取以及难融合等关键问题，研究立体成矿信息提取-三维反演建模-找矿智能预测方法体系是创新深部三维矿产资源预测与评价的关键之一。本项目结合多尺度分析思想、尺度不变分形理论以及机器学习算法，研究基于小波-分形-神经网络的深部找矿信息挖掘方法，包括多尺度成矿信息提取、三维成矿信息反演以及多源找矿信息融合，目标在于借助并有机融合不同学科新方法与算法，有效提取与反演深部立体找矿信息、约束三维地质建模以及深度挖掘多源找矿信息。本项目研究不仅能够进一步丰富和发展前人建立的三维成矿预测理论与方法，而且有助于从方法与算法角度解决深部立体找矿信息提取、反演以及融合等关键技术难题，选择内蒙古曹四夭矿床为例有望为该区深部找矿提供技术支撑和找矿依据。

Mineral exploration in deep space is becoming a new tendency in the worldwide due to the decreasingly discovered ore deposits in outcropping areas. However, the data-driven prediction of deep mineral resources is challenged by several difficulties including data acquisition, extraction and fusion in a 3D space. One potential scheme for addressing above issues is to mining 3D information from an integrated perspective of fractal, wavelet and machine learning. In this proposal, we focus on developing a wavelet-fractal-neural network (WFN) model for multiscale anomalies extraction, 3D geophysical inversion and multivariate information integration in support of mineral prediction in deep space. This idea comes from the fact that wavelet transform provide a powerful tool for multiscale analyzing the geo-anomalies, while fractal theory provides natural architect for building deep neural network for mineral prediction. Accordingly, the wavelet-based multiscale decomposition, fractal-based DEXP imaging and fractal-based wavelet neural network will be proposed respectively to address the issues of 3D geo-anomalies extraction, inversion and integration, and the presented results will further assist in the prediction of deep mineral resources in Caosiyao porphyry Mo ore deposits.