Detecting Changing Financial Relationships: A Self Organising Map Approach

摘要

In this paper Kohonen's Self Organising Map is used as a tool for detecting changing financial relationships. A Self Organising Map (SOM) is a dimension reducing transform that maps an high dimension information set to a two dimensional grid that is amenable to visualisation. This dimension reduction step is a key component of all financial analysis tasks. The potential of this method to identify structural change is investigated in the context of the problem of takeover target identification. Use of SOM analysis on two samples from different time periods charts temporal instability in the information sets of sufficient magnitude to breach the stationarity assumptions of standard statistical modelling methods, the results are confirmed by probabilistic regression analysis. This finding helps to explain the poor discriminatory power of many takeover target prediction exercises.