Multiclass Classification of Thyroid gland data employing Linear Discriminant Analysis

Abstract With the advent of internet, complex data is tremendously increased. It is essential to analyze the data. Multiclass classification, has been an important problem for a long time in supervised learning. In multiclass classification the independent variables are used to map it with k number of classes. Linear discriminant analysis is a supervised learning method to classify k number of classes. We consider applying Fisher’s Linear Discriminant Analysis(LDA) procedure to the multiclass classification. In this work, we used thyroid gland dataset to classify state of the thyroid gland. The experiments were conducted to evaluate the Fisher’s LDA method using four metrics: precision, recall , F1-score and accuracy. As the results indicated by the experiments, the Fisher’s LDA shown the best classification accuracy. Further, the Fisher’s LDA method is compared with multiclass linear regression and multiclass support vector machine.