Malaria is one of the leading causes of death in Ethiopia. Though there are many efforts to control malaria, the complexity of the problem is still very severe. So there is a need to investigate in detail the synergic effect of risk factors with temperature, altitude, type of visit and malaria type and their causes of death. Hence, in this study an attempt is made to determine the hierarchical importance of different risk factors and their patterns on malaria death occurrence and type of case identification. Knowledge discovery techniques are evaluated to support and uncover knowledge to scale up the malaria prevention and intervention program in Ethiopia. CRISP-DM methodology with classification algorithms such as J48 decision tree, JRip rule induction and Multilayer Perceptron (MLP) Neural Network are adopted to uncover knowledge from total datasets of 37, 609 records. An attempt is made to pre-process the data using business and data understanding with detail statistical summary in order to fill missing values and detect noisy once. Essential target dataset attributes have been constructed by integrating WHO malaria databases, National Metrological data and National Mapping data. Classification techniques discover important attributes/factors that determine malaria cases and occurrence of deaths. J48 Decision tree and MLP correctly classify 95.9% and 97.4%, respectively to predict occurrence of death. The findings of this research indicate that rainfall is the significant factor that determines the prevalence of malaria. When the number of malaria cases increases there is a high probability of death occurrences; the risk is relatively high with those less than 5 years of age. In most zones, malaria transmission rate is high from the month of May to January because of favourable climatic conditions for malaria reproduction.
Data Mining; Malaria; Neural Network; J48 Decision Tree; JRip Rule Induction