Abstract：Crisis mines and old mines possess great mineral resource potentials in their deep and marginal parts, but the search for resources in the depth is frequently confronted with enormous difficulties and risks. Old mines have usually experienced high degrees of exploration and exploitation, resource prognosis and prospecting goals have to be transferred to the three-dimensional space in the depth, and geophysical and geochemical anomaly information is strongly interfered by mining activities. Therefore, the adoption of new prognostic and appraisal methods for crisis mine prospecting seems to be an important means for making breakthrough in the appraisal and prospecting work in the deep and marginal parts of old mines. Aimed at solving the problems of evaluating and prospecting for replaceable resources in crisis mines, this paper established a core workflow for the stereoscopic quantitative prediction of concealed ore bodies, i.e., the flow chart of "geological data integration→quantitative mineralization information extraction→three-dimensional quantitative prediction", by using a data-driven and knowledge-driven combination method. From the viewpoint of the field analysis, the ore-controlling field models of magmatites, faults, strata, lithology and folds and the mineralization distribution filed model were constructed, with their solutions reached by transferring continuous models into discrete models. By means of statistical analysis and nonlinear programming, the correlation between geological ore-controlling fields and mineralization indexes was analyzed, and then the quantitative location parameters which describe the control of geological factors over mineralization were extracted. Thus the quantitative index set of geological ore-controlling factors was obtained. The stereoscopic quantitative prediction models composed of mineralization mathematical models and ore potential estimation models were built on the basis of the relationship between the mineralization indexes and the geological ore-controlling factor indexes. With the prediction models, the three-dimensional localization and quantitative prediction of concealed ore bodies were carried out in the deep and marginal parts of the Dachang mine, and the prediction database was created, based on which the unit prediction maps were drawn according to elevation levels. A three-dimensional visualization software system for displaying and querying the prediction results was developed. Deep prospecting drill holes were designed and put in practice following the prediction results and new ore bodied were thus found in the depth of the Changpo section and the Tongkeng section within the Dachang mine, which shows that the prediction models and results are effective.