Daily health professionals are sought out by patients, motivated by the desire to
stay healthy, making numerous diagnoses that can be wrong for various reasons.
In order to reduce diagnostic errors, an application was developed to support
health professionals, assisting them in diagnoses, assigning a second diagnostic
opinion. The application, called ProSmartHealth, is based on intelligent algorithms to identify clusters and patterns in human symptoms. ProSmartHealth
uses the classification algorithm, Support Vector Machine, to train and test diagnostic suggestions. This report aims to study the reliability of the application,
using two strategies. First, to study the influence of data pre-processing, that is,
if the accuracy improves when the data is processed before. The second strategy
intends to study whether the number of training data influences accuracy. This
study concludes that the use of a database with data pre-processing, and the
number of training data used to train the model, influence the accuracy of the
model, by improving application accuracy in eight percent.
