This application was developed as an aid to HR staff to help identify suitable candidates for promotion.  The principle can be used for a wide variety of talent spotting and staff selection/deployment aids.  A neural network can be trained on data that is specific to an organisation, so that it reflects accurately the culture, policies and methods relevant to that organisation.  This example application is developed from a range of data that is not specific to any organisation but reflects promotion criteria that are applicable generally across the board

Note that the promotion criteria are applicable to a wide range of middle management appointments.  In this context, middle management is defined as being the management of a number of supervisors who are responsible for the work of a number of skilled or semi-skilled workers.  Some of the live data used in the training has been drawn from an organisation that works round the clock in shifts.  A shift can contain some 60 workers with 4 or more supervisors. The promotion post is that of a shift manager responsible for managing all staff and supervisors on the shift.  Applications specific to an organisation would perform generally in the same way but would contain the cultural differences and requirements of that organisation.

Knowledge contained within the neural network is drawn from performance levels achieved by staff promoted to the middle management level over time.  The neural network training process is described in the section on developing a neural net.

Note too that there is no element in this network to reflect the different promotion prospects of male staff versus female.  That element could only be reflected in a specific network that deals with identified promotion posts within a particular organisation.

Neural networks are ideally suited to this type of application.  They show the world as it really is, rather than one reduced to a series of formulae and statistics.  So long as the training data is sufficiently well distributed, they will show up many none-obvious relationships; personal attributes which are important for one personality type are less so for another

The Pie Chart shows the most important five factors for any given selection of attributes.  Three standard set of results are available; they can be activated by selecting the appropriate preset result set.  You can then change the attributes by clicking on the radio buttons.  Feedback is instantaneous.   Note that the colours apply to the ranking of the attributes; thus blue is always the top, red the second and so on. The colour associated with a given attribute may therefore change from one instance to the next.