UCLA

MyTherapistMatch.com seeks to match visiting patients with suitable therapists after patients fill out the online questionnaire which consists of many psychological questions. However, a problem with this website is that many patients in fact do not end up scheduling a session with a therapist. The website founder believes that one of the major reasons is the length of the questionnaire. Therefore, to reduce annoyance for users, the task becomes selecting a subset of necessary questions from the questionnaire. The website provides patient selection data and patient action data which records how a patient interacts with a matched therapist. This thesis tries to implement hierarchical clustering method on both the question responses and the questions themselves, in order to find a reasonable way to pick the necessary questions. Correlation coefficients and Pearson's chi-squared test are used to define the metrics in hierarchical clustering. Satisfiable results are obtained. A linear model is also used to find the relationship between question responses and patient actions.