Classification of selected stocks in Thailand using cluster analysis

Abstract

Among the growth and development of Thai economy, the Stock Exchange of Thailand (SET) plays a significant role in collecting capital for various institutes in any industries. SET index becomes the important indicator of Thai economy. Information on stocks is neccessary for investing in SET. One of the beneficial informations is stocks ranking. However, currently there has not been any ranking done by considering the suitable criteria simultaneously. It can not be done because there are too many stocks in SET. But the correct way to achieve this is to classify stocks into groups and then rank groups of stocks later. This thesis proposes objectives, grouping and ranking, using cluster analysis algorithms and complementary techniques. This aims to determine distinct classification or grouping of stocks in SET, 100 selected stocks are used, based on fundamental performance indicators. All variables used are weighed by expert. Euclidean distance has been used as the measure of dissimilarity between stocks. This approach does not require assumptions. Group characteriatics are inferred from the means and use cluster medoids as reference points for measuring between-cluster distances. Group ranking comes from multi-criteria decision making under considerate criteria along with their objectives. The results of analysis suggest that there are 4 distinct groups of stocks, called as excellent, very good, good and fair performance stocks. The study only uses fundamental analysis, lacking consideration of psychological and any other effects; therefore, the results shown may conflict with normal sense, that is, this study reveals the illusion.