SiroSOM identifies relationships within large and diverse datasets typical in the geosciences and is also an effective predictive tool.
Background
Extracting useful knowledge from data is crucial to continued innovative development of industrial and environmental techniques and systems.
As technology and engineering evolve, the ability to measure variables increases and the volume of data generated can be overwhelming.
SiroSOM is a data analysis tool that uses the self-organising map (SOM) technique and although it has been specifically designed for mining and exploration data, more recently it has been adopted by other fields. SiroSOM works on a vector quantization methodology that allows unsupervised analysis to identify linear and non-linear relationships in datasets. The method allows industry and research to integrate data that otherwise may remain disparate such as:
-
geophysical
-
geochemical
-
visual attributes.
Used as a mining and exploration analysis tool, the method allows industry to quickly identify promising targets, areas of concern and a multitude of other knowledge that assists with better exploration, mining and processing strategies.
SiroSOM was initially developed for mining and exploration and is becoming more popular within that industry. However, its use is also becoming more favoured in other sectors such as social sciences, business and forestry.
What is SOM?
SOM is a method of analysis that was first described by Kohonen in the 1980s and since then, it has been further developed and is becoming a more popular form of data analysis.
SiroSOM was initially developed for mining and exploration, its use is also becoming more favoured in other sectors such as social sciences, business and forestry.
Although most people think of data analysis as ‘statistics’, unlike many statistical methods, SOM allows data of different types to be used in conjunction with each other, without supervision. This means that naturally occurring and often non-linear relationships may be divulged from the dataset and such relationships can be crucial to better understanding systems and processes.
Data exists in various formats – for example it may be:
-
descriptive, such as a soil or rock texture
-
subjective, such as a person’s judgment on how dense a bacterial bloom may be
-
a continual measurement such as a seismic wave form or it may be point data such as drill-hole data.
Where more than one type of data exists in one domain it may be difficult to assess all data in an integrated fashion. SOM however, has the ability to consider all data irrespective of data type. Similarly, SOM can also manage data that may be spatially related airborne geophysics and geochemistry.
SOM can also analyse datasets that are incomplete such as geochemical assay data where some samples were not tested for all elements or results of questionnaires where some candidates chose not to answer all questions. SOM has the important ability to ‘predict’ missing values in a dataset.
SOM works by grouping the data into naturally occurring ‘clusters’ or classes that exhibit common characteristics.
SiroSOM has the ability to allow spatial datasets to be clustered and viewed spatially in two or three dimensions. The relationships found between the components are also valuable SOM outputs. SOM’s ability to provide evidence of outliers, anomalies and inconsistencies within the dataset is proving to be equally valuable.
Find out more about the Minerals Down Under Flagship.
Site Map | Legal Notice and Disclaimer | Privacy | Copyright | Last updated: 29 October 2008
