Performance measurements of 12 different machine learning algorithms that make personalized psoriasis treatment recommendations with a database of psoriasis patients responding to treatment

Abstract

The improvements in the performance of parallel processors have contributed greatly to the rapid increase in artificial intelligence (AI) applications. Compared to traditional processors, parallel processors can perform artificial learning operations within a minute that will last for months. AI revolution has been made in many fields today. AI is also frequently used in the field of health. In this study, psoriasis patient data collected by our doctors who contributed to the article were trained with 12 different machine learning methods. There are eight inputs and three outputs in our dataset consisting of 100 patients. The main purpose here is to obtain the highest accuracy for the right treatment method to be applied to future patients. The outputs in the dataset used for training indicate which type of patient is cured with which treatment method. Our input values in the dataset are age, disease type, family history, arthritis, pitting, smoking, stress, and gender. The results obtained in this study with less patient data lay the foundation of a system that will be created with much higher accuracy with more data in the future.