Combining multiple diagnostic tests for classification

The data that will be analyzed in this thesis is a Liver Cancer data set which was collected at Shanghai Cang-zheng Hospital in China. The data set includes 145 subjects, of which 54 patients have hepatoma, 39 patients have hepatitis and hepatocirrhosis - chronic liver disease and 52 individuals are...

Πλήρης περιγραφή

Αποθηκεύτηκε σε:
Λεπτομέρειες βιβλιογραφικής εγγραφής
Κύριοι συγγραφείς: Baika, Alexia Artemis, Βαΐκα, Αλεξία Άρτεμις
Άλλοι συγγραφείς: Τσιμήκας, Ιωάννης
Γλώσσα:English
Δημοσίευση: 2019
Θέματα:
Διαθέσιμο Online:http://hdl.handle.net/11610/19523
Ετικέτες: Προσθήκη ετικέτας
Δεν υπάρχουν, Καταχωρήστε ετικέτα πρώτοι!
Περιγραφή
Περίληψη:The data that will be analyzed in this thesis is a Liver Cancer data set which was collected at Shanghai Cang-zheng Hospital in China. The data set includes 145 subjects, of which 54 patients have hepatoma, 39 patients have hepatitis and hepatocirrhosis - chronic liver disease and 52 individuals are healthy. To each subject corespondents 236 Markers. For simplicity, we create two types of groups. The first group consists of healthy individuals and the second group consists of diseased patients. The goal is to correctly classify a subject in one of the two classes, diseased-non-diseased. For this classification task we use two supervised classification methods: Logistic Regression and Support Vector Machines. Logistic Regression was developed by David Cox in 1958 and is one of the most traditional parametric classification methods. Support Vector Machines was created by Vladimir Vapnik (1995). It can efficiently perform a non-linear classification using what is called the kernel trick, by mapping the inputs into a high-dimensional feature spaces. In many cases the goal is to develop a model which can explain the relationship between the features and the dependent variable. A severe mathematical problem is when the dimension of the data is greater than the number of the available data points. For this purpose we will describe methods for feature selection and regularization, including subset selection and lasso.