Case based expert system for hypertension diagnosis

Abstract

This paper describes a case based expert system for hypertension diagnosis in Santa Clara city, Cuba, carried out as part of an investigation to know the incidence of the disease in this population. The studied population’s sample was formed by 455 men and 394 women, with ages between 18 and 78 years. The cases were classified in three groups: normal (people with normal blood pressure), with prehypertension (people that has risk to suffer hypertension) and people with hypertension. A statistical process was made with this sample in which multivariate techniques as Discriminant Analysis and Logistic Regression were used. The results of these techniques and the application of the Fuller’s Triangle Method were used in the system for ranking the hypertension risk factors and for obtain their importance grade (weigh). CHAID technique allowed reduces the comparisons of the cases making more efficient this process. With the application of the analysis of variance (ANOVA) and TwoStep Cluster Analysis the features comparison function for continuous variables was obtained. With all this information the likeness function was build to compare the new case with the rest of cases of the case base. The adaptation of the solution of the most similar cases was carried out with the application of nearest k-neighbors algorithm. Finally the system was validated and was checked an effectiveness of 96%.