An Alternative Method of Analysis in the Absence of Control Group
Although control groups are an important part of psychology, there are times when an appropriate control group is difficult to obtain. In the machine learning community, Support Vector Machine has often been successfully used for classification. Moreover,
- PDF / 320,809 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 93 Downloads / 132 Views
An Alternative Method of Analysis in the Absence of Control Group Felin, Joachim Diederich and Insu Song
8.1 Introduction Control groups have played an important part in experimental psychology. There are, however, situations where control groups are hard to find or finding an appropriate one would be costly. In clinical settings, such as hospitals and private practices, data available for research are usually obtained from patients with mental illnesses or defects. In order to draw reasonable conclusions, however, control groups are often seen to be necessary as comparisons. Finding ‘‘healthy’’ controls in hospitals and private practices is not easily done given practitioners’ lack of resources and time. It would, therefore, be advantageous in situations like these to have an alternative method of analysis that allows practitioners to conduct a valid study even when control groups are not available. Although such a method of analysis may appear foreign in the field of psychology, it is not so in the machine learning community.
Felin School of Business and IT, James Cook University Australia, Brisbane, Australia e-mail: [email protected] J. Diederich (&) School of Information Technology and Electrical Engineering, The University of Queensland, Brisbane Q4072, Australia e-mail: [email protected] I. Song School of Business and IT, James Cook University, Singapore Campus, Singapore 574421, Singapore e-mail: [email protected]
M. Lech et al. (eds.), Mental Health Informatics, Studies in Computational Intelligence 491, DOI: 10.1007/978-3-642-38550-6_8, Springer-Verlag Berlin Heidelberg 2014
151
152
Felin et al.
8.2 Background 8.2.1 Support Vector Machine Researchers in the machine learning community have been using a popular method, called Support Vector Machine (SVM), for data analysis and classification. SVM was developed from statistical learning theory by Vapnik and colleagues [10]. It is called a ‘‘supervised’’ machine learning method because the data entered into the programs are labelled. In other words, the machine learns to distinguish one category from the other based on labelled examples [16]. When new data are entered, the machine uses the knowledge from the previous examples it has learned in order to classify the new data into categories. Each of the data entered is represented as a dot (vector) in the input space. A maximum margin classifier, such as SVM, separates the data by finding the best boundary decision through an algorithm. The boundary decision, or the hyperplane, maximises the margin between the nearest points to it on each side or support vectors [10, 21]. Sometimes, this hyperplane is in the form of a simple straight line, although, in most cases, it is non-linear. When the hyperplane is non-linear, an appropriate algorithm function or a kernel can be chosen to map the input space (where data are entered) into a feature space [18]. The method described above is for a regular two-class SVM. It requires the input of both positive and negative examples. In other words, data fro
Data Loading...
