پديد آورنده :
هراتيان، آرزو
عنوان :
تحليل داده هاي آزمايش خون با استفاده از روش هاي يادگيري ماشين
مقطع تحصيلي :
كارشناسي ارشد
محل تحصيل :
اصفهان : دانشگاه صنعتي اصفهان
صفحه شمار :
سيزده، 91ص. : مصور، جدول، نمودار
استاد راهنما :
زينب مالكي
استاد مشاور :
فرزانه شايق
توصيفگر ها :
يادگيري ماشين , دادهكاوي , دادههاي آزمايشگاهي , پيشبيني تستهاي آزمايشگاهي
استاد داور :
محمدرضا احمدزاده، ناصر قديري مدرس
تاريخ ورود اطلاعات :
1399/05/10
رشته تحصيلي :
مهندسي كامپيوتر
دانشكده :
مهندسي برق و كامپيوتر
تاريخ ويرايش اطلاعات :
1399/05/11
چكيده انگليسي :
Blood test data analysis using machine learning methods Arezoo Haratian Arezoo Haratian@ec iut ac ir 2019 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language Farsi Supervisor Dr Zeynab Maleki zmaleki@cc iut ac ir Advisor Dr Farzaneh Shayegh f shayegh@cc iut ac ir Abstract Today there is a huge amount of clinical data which can be used as a good source to improve healthcare systems andmedical treatment However usually it is not easy to use these databases and we must use and develop advanced methodsto extract useful information from them Among all clinical database using clinical laboratory data is important because itis inexpensive and accessible and can also easily represent the general condition of the patient Therefore using machinelearning tools for analyzing laboratory data can provide a good basis for research and knowledge discovery During last decades many researchers investigate different type of laboratory data specially blood test They use variousmachine learning methods aiming reducing the laboratory costs increasing the accuracy of prognosis and diagnosis andfinding correlations among variables In this thesis we consider blood tests data from a clinical laboratory in Isfahan to findcorrelations among variables One of main challenge for analyzing such a real data is large amount of missing data In thisregard we use existing imputation methods and propose a new imputation method based on Bayesian network which filleach of variables with a specific Bayesian network defined for that variable Finally we will consider several classificationmethods to predict the value of a specific variables such as creatinine It shows that the best accuracy is obtained by usingthe new proposed imputation method which verify the strength of the method The area under the ROC curve for theclassification models developed to predict red blood cell and creatinine was 0 97 and 0 90 respectively Key Words1 Machine learning 2 Data mining 3 Laboratory data 4 Test result prediction
استاد راهنما :
زينب مالكي
استاد مشاور :
فرزانه شايق
استاد داور :
محمدرضا احمدزاده، ناصر قديري مدرس