dc.description.abstract |
Clinical data refers to health-related information that is associated with regular patient
care. It provides information to health care professionals to improve the quality and
safety of the care they provide to their patients. Based on huge data mining research is
used to improve health care, to plan and make decision policy for satisfy medication
process. Hence, health service planning and utilization become limited. Thus, adult
mortality levels and trends in the developing countries become hampered. Therefore, in
this study, we proposed a data mining prediction model to identify determinant attribute
and consideration factors for adult mortality in case of patient dataset in Felege Hiwot
Referral Hospital. The study contains 7095 instances from Felege Hiwot referral hospital
recorded datasets that age between 15-60 years. To develop the model, we used
classification techniques and data mining algorithm such as J48 decision tree algorithm,
Support vector Machine, Random Tree, K-Nearest Neighbor and Naïve Bayes algorithm.
In this research Attribute selection for better accuracy is performed by using
GainRatioAtrributeEval with rank. We used data KDD model approach to processed
data. K-Nearest Neighbor algorithm is selected algorithm to build the model that predict
adult mortality in better accuracy and possible for correct classification with value
88.27% and K-Nearest Neighbor was processed in 0 second speed, 88.3% recall, 88.9%
precision and 88.5% F-Measure scored in this research.
Finally, National Classification of Disease, Duration of illness, Length of stay was
significantly selected attribute to predict adult mortality. From this research Urinary Tract
Infection, tuberculosis, congestive heart failure; renal disease, Road traffic accident and
Poisoning were main factors of adult mortality which needed attention to minimize adult
mortality by found the cause of establishment of such diseases and provided community
awareness.
From this study K-nearest neighbor is recommended Algorithm for build model to predict
adult mortality with 88.27% accuracy was possible. Duration of illness was very
dominant attribute to adult mortality in Felege Hiwot Referral Hospital in this study.
Keyword: National classification of disease, mortality, length of stay, KNN, clinical data |
en_US |