Methodology

Software Implementation

The information composing the knowledge base is obtained and strengthened in accordance with how prevalent the said diseases are in the Philippines by utilizing the statistics released by Department of Health. From the physician’s clinical eye and experience, a patient is given a diagnosis based on the set of features and each features’ degree of severity, thus, the group surveyed four physicians to profile these diseases to capture their medical knowledge and experience.

After acquiring information for the knowledge base, sets and rules are established for the formal representation of the fuzzy logic. This section details the representation of the inputs, weighted rules, and outputs used in the system.

1. Input Membership. The temperature, days experiencing fever, and severity of symptoms, uses these input membership functions/curves:

For the geographical location, each region is simply classified into three sets: “no”, “maybe”, or “yes”. A “no” indicates that the certain disease is not prevalent in that region (below the lower threshold); a “maybe” indicates the certain disease may or may not be prevalent in that region (between lower and upper threshold); “yes” indicates that a certain disease is highly prevalent in that region (above the upper threshold). A sample is shown below:

The expert system processes these gathered data using fuzzy logic and weighted rules in the following order: (1) Fuzzification; (2) Weighted Rules; (3) Mamdani Inference; and (4) Defuzzification.

1. Fuzzification. Fuzzification is the process of transforming crisp values into degrees of membership for linguistic terms of fuzzy sets, or simply converting a scalar value into a fuzzy value. It is only applied for inputs - temperature, days experiencing fever, and severity of symptoms, which are classified into six fuzzy sets and from their respective membership function, its fuzzy value is obtained (Refer to input membership curves above).

2. Weighted Rules. After fuzzification of the input, an inference is made based on a set of rules. These rules were established using the disease profiles obtained from four physicians, and a total of 130 rules was generated. Once all input has been fuzzified, the system continues with the next stage - making decisions by running through the knowledge base rules which determines if the severity of this symptom is a significant factor in a certain mosquito-borne disease. Moreover, a “weight” is assigned to each rule to strengthen the medical jurisdiction of the system, for the significance of presence of each the features in a mosquito-borne disease is not uniform.