Search

In this study, 42 companies operating in food, textile and cement sectors within İstanbul Stock Exchange 100 (ISE-100) have been handled. The aim is to classify these companies into three groups according to financial ratios. The average values of 10 financial ratios of these companies between the years 2006-2011 have been handled. Based on these ratios, classes are derived from cluster analysis. These ratios and the results of the cluster analysis are the data set of this article. In order to test the performance of the learning algorithm and classification leave-one-out cross-validation method is used. The classification study conducted by Support Vector Machines approach has performed 95.23% correct classification with the help of 12 support vectors. Moreover, input sensitivity analysis has been conducted and 4 most efficient ratios have been determined out of these 10. These ratios are removed from the model one by one starting from the less influential one in order to investigate by which ratios the most effective Support Vector Machine model is obtained. It is seen that the best model is obtained by using the first 3 ratios. The classification success for this model is 97.61% and the number of support vector is 12. [ABSTRACT FROM AUTHOR]

In this study, 42 companies operating in food, textile and cement sectors within İstanbul Stock Exchange 100 (ISE-100) have been handled. The aim is to classify these companies into three groups according to financial ratios. The average values of 10 financial ratios of these companies between the years 2006-2011 have been handled. Based on these ratios, classes are derived from cluster analysis. These ratios and the results of the cluster analysis are the data set of this article. In order to test the performance of the learning algorithm and classification leave-one-out cross-validation method is used. The classification study conducted by Support Vector Machines approach has performed 95.23% correct classification with the help of 12 support vectors. Moreover, input sensitivity analysis has been conducted and 4 most efficient ratios have been determined out of these 10. These ratios are removed from the model one by one starting from the less influential one in order to investigate by which ratios the most effective Support Vector Machine model is obtained. It is seen that the best model is obtained by using the first 3 ratios. The classification success for this model is 97.61% and the number of support vector is 12. [ABSTRACT FROM AUTHOR]