Abstract: The mineral resources prediction and assessment in the 21st century are conducted with the support of the geography information system (GIS) and by building various spatial information databases such as geological, geophysical, geochemical and remote sensing spatial databases, performing spatial analysis and information conversion and extraction for these databases and finally outlining integrated anomalies. The premise of this work is to build high-quality spatial databases and the key to this is raising the quality of the spatial databases and ensuring the correctness and integrity of the data in the databases. So it is an inevitable problem to check data. The basic mineral resources prediction and evaluation database is characterized by a huge data capacity, complex structure and multiple types of data. The authors suggest that: the metadata check can effectively control the integrity and time limitation of the data, the data dictionary technique can ensure the attribute accuracy; and the quantity rules can control the logic consistency and position accuracies of the data. At last, the data check is a good tactic, which has been proved by checking the basic spatial databases for nationwide important metallogenic provinces and belts.