An Integrated Neural Network and Machine Learning Model for Multi-Dimensional Mineral Resource Assessment

The potential of the developed multi-dimensional analysis system to assist interpretation from large data sets through derived digital mapping models is presented accompanied by preliminary results. The concept learning model is original and its application general. Applications exist in general science, engineering, earth resource management and economics. The developed system incorporates advances in machine learning, neural networks, object-oriented software engineering, knowledge representation and fuzzy logic. ' The specific application presented involves automatic earth resource modelling and assessment involving geological theories of mineral potential. The developed system functions as a new and powerful mineral exploration tool, To date the model has been trialled on an earth resource database and several published data sets. A more general potential application is knowledge discovery within databases. The specific theory that is tested in the experimental results is the nature of the cause-effect relationship between the hypothesised geological causes and the measured effects within a previously published test survey. Also integrated here are a neurophysiological cerebellar model of human learning, a new concept similarity measure for pattern recognition, modelled components to explain human concept formation through knowledge discovery and a means of permanent representation.The implementation is as a new artificial neural network architecture functioning as a hybrid expert system. Understanding of programmed concept learning is extended through the model. Also established is a basis for a new approach to multivariate data analysis. Spatial analysis for decision support is integrated over many dimensions, relationships, data types and data modes. Application of the developed methodology and system adds value to data bases. Discovered knowledge is represented as new generalisations, interpolations, classification hierarchies, aggregations and associations within a dynamic, maintainable and reusable integrated knowledge framework. New digital knowledge products and intelligent access follow. Permanently stored, maintainable, easily accessible and discovered knowledge improves decisions through extended hypothesis testing, spatial data analysis and inferencing, prediction, logical deductions and explanations.