Preserve Discovered Linguistic Patterns Valid in Volatility Data Environment

Abstract

Many data mining techniques have been developed and shown to be successful in financial domains. A further aim is to make sense of numerical data through a human-friendly way, by which general patterns are extracted in terms of linguistic concepts. Problems associated with the linguistic mining approach are the effective representation and the validity preservation of the linguistic patterns. The volatile data may vary linguistic concepts and make previously discovered patterns invalid. This paper aims to solve the problem. Based on the cloud model proposed in our previous works, linguistic patterns can be represented effectively. Outdated linguistic patterns can be valid by a GA-based validity preservation technique in line with current data set. An example of Hong Kong stock market is given to illustrate how the technique works.