Viability prediction for retail business units using data mining techniques: a practical application in the Greek pharmaceutical sector

In this paper, we explore the effectiveness of supervised learning methods in predicting the short-term viability of retail pharmaceutical businesses. We use data mining techniques such as linear discriminant analysis, k-nearest neighbour (k-NN) and the C4.5 Decision Tree to classify retail business units from the Greek pharmaceutical sector into viable and non-viable classes, while operating in an environment of strict fiscal control and many changes of regulations. The issue of viability prediction for business units, in a period that has been characterised as the most crucial economic and financial crisis of the last decades globally, is vital for all players involved in an economic system. The effectiveness, accuracy and promptness of identifying non-viable business units are important goals for every link of an economic chain, which has to cope with decisions that will minimise the costs and losses that the current crisis causes.