Predicting Psoriasis Area Severity Index (PASI) using decision tree classifier

Project description:

Input: Data points which are comprised of 13 features such as age, sex, body mass index (BMI), the result of positivity/negativity of skin test on the Head & Neck, Trunk, Upper, Lower or palmoplantar areas, and age of onset (the age that the patient is diagnosed). The dataset is comprised of 190 data points. 30% of data set is used for test data. The first and last five patient data are as follows:

Psoriasis Area Severity Index (PASI): Depending on the computed index, it is divided into four categories as mild (index < 5), moderate (5 < index < 12), severe (12 < index < 20), very severe (20 < index < 72).

Goal: To predict PASI for a new data point.

Approach: A decision tree classifier with maximum depth of 5 is used to classify data points in the training set based on their features. The tree is shown in the following.

The figure can be downloaded from the link. The score of trained classifier is 0.79 meaning that the average accuracy of model is 0.79.

The source code of this project is available in my Github page.