UNAIR NEWS – Children under five years old are very susceptible to disease. From a number of potential diseases, there are at least four diseases that have highest possibility to affect children under five, such as flu symptoms, fever, vomiting, and inability to drink or suckle.
Indah Werdiningsih revealed her findings from her research with Rimuljo Hendradi, Purbandini, Barry Nuqoba, and Elly Ana, by reading health information system data using certain methods.
Correct health information system reading is very important as the knowledge from these data can be a material to determine diagnosis and certain policies.
“Diagnosis of early childhood diseases often takes a long time and is prone to mistakes,” she said.
Indah explained, the health information system produced big data. However, hidden knowledge in data cannot be accessed using traditional methods.
“Data mining is one solution that can be applied to explore hidden information in health information systems. Data mining allows the extraction of knowledge in large volumes of data. The data is explored and analyzed to find meaningful patterns and rules, ” she said.
Descriptive data mining, continued Indah, is one type of data mining to find patterns from a collection of data. The process involves clustering, association rule discovery, and sequential pattern discovery.
Meanwhile, association rule mining is a data mining process used to find rules that can regulate associations between items in a data set. Apriori association techniques have been proven effective in finding various trends in health information systems.
“The research data was obtained from hospitals and health centers in Surabaya. Data collection techniques used were interviews and document analysis, ” she said.
Indah said that interviews were conducted to obtain information on risk factors for the early childhood diseases. As a result, it is found that they are closely related to certain symptoms, but the patterns and factors that influence early childhood disease cannot yet be determined.
“Analysis of the document is also carried out to determine the types of diseases and risk factors that influence the diseases,” she said.
A total of 16 factors of symptoms of early childhood diseases were used in the study, coughing; pneumonia; severe pneumonia; diarrhea; mild dehydration diarrhea; severe dehydration diarrhea; persistent diarrhea; severe persistent diarrhea; dysentery; fever without general danger sign; fever with general danger signs; measles; measles with severe complications; measles with complications; suspected dengue fever (DHF); and DHF.
“As a result, 21 influential factors, weight, height, sex, flu, cough, fever, diarrhea, stridor, blood in the stool, vomiting, seizures, unconsciousness, inability to drink or suckle, hollowed eyes, fussiness, abnormal thirst, turbidity on the cornea, fever 2-7 days, turgor, diarrhea 14 days or more, and difficulty in breathing, ” she said.
Indah said, one of the rules produced with the greatest confident value was Antecedent (symptoms of flu, fever, vomiting, inability to drink or suckle), then Consequent (affected by medium class disease ) with support = 0.25 and confident = 0.95. The number of items produced by the Apriori Algorithm is 272 items, while the other methods are 956 items. The experimental results showed that a priori algorithm can produce a more complete rule and better computational performance. (*)
Author: Feri Fenoria
Editor: Khefti Al Mawalia
Reference:
https://content.sciendo.com/view/journals/cait/19/3/article-p154.xml?lang=en
Indah Werdiningsih, Rimuljo Hendradi, Purbandini, Barry Nuqoba, Elly Ana (2019). Identification of Risk Factors for Early Childhood Diseases Using Association Rules Algorithm with Feature Reduction. Cybernetics and Information Technologies. 19(3): 154 167
