|
|
دومین کنگره انفورماتیک پزشکی و هفتمین همایش سلامت الکترونیک
|
|
|
| عنوان فارسی |
پیش بینی عوامل موثر بر مرگ نوزادی با استفاده از تکنیک داده کاوی |
|
| چکیده فارسی مقاله |
|
|
| کلیدواژه‌های فارسی مقاله |
|
|
| عنوان انگلیسی |
Predicting risk factors of neonatal mortality using data mining techniques |
|
| چکیده انگلیسی مقاله |
Background and Aim: Neonatal mortality is the death of newborns between 1 and 28 days of age and it causes harmful effects on families and society. Each year about 15000 infants dies in Iran. This amount is 3 times more than developed countries. Reasons that results to neonatal death are not completely known. But, study on previous patient documents and histories can help identify major risks of pregnancy and mother features that leads to the child death. Data mining algorithms is the tool to study on these records. Results from data mining can address the risks and advance our knowledge in that term. It’s a way for early prediction of diseases, alerting abnormal health conditions as a knowledge assistant for physicians. Methods: This study was done on a database of 2788 records and 134 variables. Data extracted from a questionnaire containing both demographic and clinical information of mothers and fetus. A total number of 1162 births leads to death and other 1626 records were normal births in this case control study. In the first step data has been cleaned and normalized. All variables recoded to nominal data and also new factors generated out of other values. Data has been divided into 3 groups by their clinical properties. Extra variables unrelated to our subject, excluded by the help of attribute selection algorithms and expert opinions. At last different algorithms have been tested with emphasis for the best death predictions. Results: Algorithms show their best results after attribute reduction. Total 134 questionnaire variables reduced to 46 and these attributes also studied in three groups. Three data groups were: “Mothers’ diseases”, “Pregnancy complications” and “Mother and fetus characteristics”. Each contains some principle demographic variables, related clinical data and one class variable defining the death or live situation of neonatal. Best result emerged out of Naive Bayes classifier algorithm. Its precision average was 0.806 and recall shows the number of 0.803. Related confusion matrix shows death records were better classified than live ones. Conclusion: Our study shows that data mining can help predict most adverse traits that weaken the chance of neonatal to survive. Application of clinical data mining can help discover hidden knowledge behind medical records for physicians. |
|
| کلیدواژه‌های انگلیسی مقاله |
neonatal، death، mortality، data mining، risk |
|
| نویسندگان مقاله |
amir hossein zolfaghari - student research committee, school of management and medical information, shiraz university of medical sciences, shiraz, iran
mehdi nasiri - shiraz university of medical science, health human resources research center, school of management amp; 38; information sciences, iran
elahe hosseini - student research committee, school of management and medical information, shiraz university of medical sciences, shiraz, iran
mahmoud hajipour - phd student of epidemiology, research center office, epidemiology department, school of public health, shahid beheshti university of medical sciences, tehran, iran
|
|
| نشانی اینترنتی |
http://mieh-2018.modares.ac.ir/browse.php?a_code=A-10-208-1&slc_lang=fa&sid=1 |
| فایل مقاله |
فایلی برای مقاله ذخیره نشده است |
| کد مقاله (doi) |
|
| زبان مقاله منتشر شده |
en |
| موضوعات مقاله منتشر شده |
|
| نوع مقاله منتشر شده |
|
|
|
|
برگشت به:
صفحه اول پایگاه |
دوره مرتبط |
کنفرانس مرتبط |
فهرست کنفرانس ها
|
