HyperCube is helping the mining industry to extract new insights from unprecedented volumes of data.
Learnings from large datasets allowed us to consider the entire range of variables for the first time and discover new and counterintuitive association rules which represents new knowledge in the field of mineral exploration.
Réjean Girard, Président of IOS Services géoscientifiques Inc
Mineral exploration is a complex endeavor. Prospecting is becoming more and more expensive as new mineral deposits become harder to find. While new tools and techniques allow for the collection of a wealth of data across large territories, the challenge remains how best to use this new intelligence.
Scientists are struggling to find efficient methods of analysis and it is becoming clear that traditional approaches are no longer suitable to handle the volume of data generated.
In order to predict and discover new mineral deposits, therefore, the industry needs to develop solutions capable of handling data at scale so that it can extract insights with greater precision and in less time.
Using publicly-available historical data, stretching back over several decades, we created a flat file database to accommodate all relevant characteristics. The data, which includes heteroclite, asymmetrical, fragmentary and grouped information, formed a training model that contained more than 1.4 million observations and 100 variables.
These variables included geophysical, geochemical and sediment information. Applying an algorithm, we were able to explore all possible data combinations and apply rules most likely to predict mineralization. The training dataset which contained a random selection of 50% gold deposits, helped us create a final model composed of 15 association rules containing a total of 13 variables.
The HyperCube model outperformed the current model in a number of ways. It has resulted in: