General Description of the Process of Behavioural Profiling

The study of patterns of behaviour and the grouping of users according to exhibited behaviour is called behavioural profiling. Behavioural profiling uses detailed records of the relationship between the organisation and the user, such as records on product usage, account balance or transaction history. Beyond just knowing that someone did something, behavioural profiling involves capturing records of events and actions over time and using these stored records of interactions to model typical behaviour and deviations from that behaviour. Sometimes, this is augmented with data from outside databases, such as census data.

Behavioural profiling is performed through data mining, a process that ranges from data selection and preparation to post processing and includes the interpretation of the emerging results. This chapter provides an overview of the process of data mining, including a discussion of the main models and algorithms used and a reflection on the relationship between these and the objectives of the data mining exercise – e.g., an inductive process that aims to uncover patterns or relationships previously unknown versus a deductive process that looks for confirmation, or indeed departures, from accepted patterns or models of behaviour.

Data mining has its origins in quantitative disciplines, including the artificial intelligence community (e.g., machine learning and pattern recognition) and the mathematical community (e.g., statistics and uncertainty processing). However, human cognitive factors can deeply affect the results of a data mining effort. This chapter concludes with a discussion of how the data mining user can influence the outcomes of data mining.