VALUE VERSUS GROWTH PORTFOLIO PERFORMANCES: A DATA MINING APPROACH ON TURKISH STOCKS Halit GONENC and Onur KOYUNCU Hacettepe University, Department of Business Administration ABSTRACT This study applies a simple data mining approach to define the stock market movements and accuracy of certain stock performances based on Book-to-Market ratio and size effects. Using the Istanbul Stock Exchange as a laboratory, we modify the Fama and French approach (1995) that forms and measures the returns of size and book-to-market portfolios. We combine the Fama and French approach with a simple data mining application consisting of prespecified association rules based on trade volume along with book-to-market ratio and size. We find the evidence of a very small value premium in the period 1991-1999, but when the period is divided to two parts at year 1995, a shift was detected for the second period where value premium did not exist. Even though our findings are generally justified by the data mining approach, we observe changes in magnitudes of the results. Keywords: Value Premium; Data Mining; Emerging Market. JEL classification: G 12; C 12 1. Introduction Data Mining, as a subprocess of Knowledge discovery, has become a useful however not yet a popular tool in the business area in the last decade. By making use of knowledge discovery, regular or irregular systematic patterns in large data sets are automatically searched or and detected. The main phases of knowledge discovery can be summarized as identification of the problem and the solution space, specification of the strategic approach, data mining and implementation and monitoring the resulting model. At this point, more emphasis is to be put on the last subprocess, data mining, which plays a crucial role in data analysis. Data mining is an iterative approach that explores and extracts the hidden and mostly complex information which, when revealed, contributes to the aim to be reached. The revealed information provides strategic help to the user in case of existence of an exact 1