What are expert systems and how complex is SELLEA?

Expert systems are much more than expert models, they are machines that think and reason as an expert would in particular domain
SELLEA system’s domain of expertise is selection of most likely buyers from bank’s client data for a particular offering
the system, therefore, emulates all the steps that an expert would perform when faced with such a task
These steps which are embedded into the system are initiated and directed by configuration parameters. These steps include:
Assessment of available data-fields and their values, parametrisation, selection and prioritisation of available expert models, and
Calibration of model scores to produce a single target-list per sales objective

Twenty seven years ago, Stevens defined Expert Systems with a following statement:

Expert Systems are machines that think and reason as an expert would in a particular domain. For example, a medical-diagnoses expert system would request as input the patient’s symptoms, test results, and other relevant facts; using these as pointers, it would search its data base for information that might lead to the identification of illness.

A common banking example of an Expert System is an ATM machine – it emulates the actions that an expert in particular domain (in this instance, a bank teller) would perform when faced with a client requesting a cash withdrawal. The inputs required by an ATM system are quite simple and consist of few parameters such as client identification, account balances and utilisation of various limits... In this example, the definition of an Expert System becomes the same as that of an Expert Model (i.e. client identified + funds in account available + account within limits + funds in the ATM available + denominations available = perform transaction).

Expert System can also get much more complex than just a set of Expert Models.

SELLEA consists of many individual Expert Models for the same Sales Objective (seven, in the example of Term Deposits for Individuals). These models are created in a ‘spine form’ as their actual value parameters are set based on bank’s data. Then, there are Expert Models on top of these Expert Models which are used to prioritise between selections made by each model based on content and definition of Input Data that each model requires and receives… meaning that the approach to selection of how much to trust each data-field within a single Expert Model and how much to trust the outcome of each Expert Model becomes a new set of Expert Models…