PREDICTING MATERNAL MORTALITY IN AMHARA REGION USING DATA MINING TECHNIQUES

Kindie, Abinet

dc.contributor.author	Kindie, Abinet
dc.date.accessioned	2020-10-07T10:59:11Z
dc.date.available	2020-10-07T10:59:11Z
dc.date.issued	2020
dc.identifier.uri	http://hdl.handle.net/123456789/11278
dc.description.abstract	Maternal mortality is the death of pregnant women due to complications arising during the period of pregnancy and after delivery. Predicting maternal mortality and identifying the major determinants for maternal mortality are important for decision making during treatment and follow-up. Therefore, this research is aimed to apply data mining techniques to build a model that can assist in predicting maternal mortality. The data is taken from Debre Markos Referral Hospital, Felege Hiwot Specialized Referral Hospital, and Addis Alem Hospital located in Amhara regional state. The six-step hybrid knowledge discovery method is employed as a framework for the activities done in this study. Dataset pre-processing was applied for missing value handling, noise removal, and data transformation. In this study, RStudio data mining tool and classification-based data mining techniques such as decision tree, naïve Bayes, and Support vector machine were employed to build the predictive model. The performances of the models were evaluated using sensitivity, specificity, precision, recall, and Accuracy. 10-fold cross-validation and percentage split were adopted as the test option to check the performance of each classifier. Experimental results show that the most effective model to predict the status of maternal outcome and determinant factors of maternal death appears to be the Pruned J48 decision tree model with a classification accuracy of 97.56%, precision 96.83%, sensitivity 99.29%, specificity 94.78%, and recall 99.29%. The extracted rules from the selected models show that the maternal condition, Age, and obstetric complications were the major determinant factors of maternal mortality. Postpartum Haemorrhage (PPH), Eclampsia, and Antepartum Haemorrhage APH) were the major complications that have a high impact on maternal mortality. The outcome of the study can be used as an assistant tool for physicians to make a more consistent diagnosis of factors that causes maternal mortality. The possibility of integrating the results of this study with a knowledge-based system should be discovered so that domain experts can access the system in their problem solving and decision-making tasks.	en_US
dc.language.iso	en	en_US
dc.subject	Information Technology	en_US
dc.title	PREDICTING MATERNAL MORTALITY IN AMHARA REGION USING DATA MINING TECHNIQUES	en_US
dc.type	Thesis	en_US