Search

Using the Metropolitan Police Service Central Drug Trafficking Database (London, UK), the Auto-Contractive Map neural network technology is applied to identify unknown associations among individuals contained within it. Whilst a database may contain associations of known individuals, it may be mined to identify associations that are altogether unknown and unexpected. It is shown that individuals can be identified who belong to the same gang or drug trafficking circle. The results produce a profile that is mathematically justified and devoid of any political involvement. Key to this analysis is the organization of the datasets such that a meaningful analysis and interpretation can be made. This organization is described and illustrated using the drug trafficking data.

This chapter shows a useful application of self-organizing mapping artificial neural networks (SOM ANN) to actual data. The dataset comes from investigations in London about drug trafficking conducted by Scotland Yard Metropolitan Police. The SOM system used in this application is empowered by filtering the codebook through the minimum spanning tree and by the graph analysis of the codebooks themselves. This application represents a new way to use SOM ANN in an actual situation: MST and graph analysis permits the discovering of new hidden features and relationships in the dataset, usually not visible with the classic SOM approach. A massive drug trafficking database that has been collected over a period of time by the Scotland Yard Metropolitan Police is analyzed, and the results are explained that yield a series of useful profiles.

The self-organizing map artificial neural network takes high-dimensional data and produces a diagram (map) that displays it in one or two dimensions. In short, humans can visualize interactions when displayed in one, two, or three dimensions, but not four or more dimensions. Data composed of only one variable can “see” a point on an x-axis diagram; data composed of two variables can be displayed on an x–y axis diagram; data composed of three variables can be displayed on an x–y–z axis diagram, and the visualization stops here. We simply cannot visualize diagrams in four or more dimensions, and that is where the self-organizing map comes into play. It has the ability of analyzing data in an unsupervised way (without any preconceived indication of the number of patters present in the data) and placing the resulting analysis in a one- or two-dimensional diagram. While some information might be lost in the translation is more than made up with the insights that one can glean from the resulting diagram.