According to the Web 2.0 paradigm web contents should not be created by specific authors and organizations but the users themselves. It has been proven that after reaching a critical mass of users, the database of such websites is growing faster and better representing the personal taste and opinion of their users. A collective knowledge base like this is able to function as a basis of a decision supporting system, thus comes the idea of creating a community-based healthcare system.
Developing a diagnostic system based on community knowledge took the major part of my work. We can not assume that users own the necessary professional knowledge to solve this problem, but they can upload their case histories which can be processed with statistical and machine learning methods, thus the problem could be solved. I have developed three methods which were based on distance, distribution and machine learning. By processing the results of these methods with majority vote, the classification error of the diagnosis can be reduced to a very low rate.
Since I did not have access to real case histories I had developed two different models of diseases (a simple and a realistic one) and a method to simulate user behavior during the data upload. The diagnostic system was tested on the data set created with these models and the test results were documented.
A minor part of my work was to develop a method to suggest cures for a known disease and rate them, which can function according to the experiences of the users like the duration of the sickness, cost of the cure and possible side-effects, or simply by their rating of the possible cures.