ارائه روشي نوين در تشخيص زودهنگام بيماري ديابت با استفاده از تركيب درخت تصميم و الگوريتم ژنتيك

A New Method For Early Detection Of Diabetes By Using Decision Tree And Genetic Algorithm

مقدمه ديابت يك بيماري متابوليسمي است كه ناشي از عملكرد ترشح ناقص انسولين مي‌باشد كه با بالا رفتن قند نمايان مي‌شود. بالا بودن مداوم قند خون منجر به بدعمل كردن و خرابي ارگان‌هايي از قبيل كليه‌ها و چشمان، اعصاب، رگ‌هاي خوني و قلب مي‌شود. در دهه‌هاي اخير، چندين تكنيك براي تشخيص ديابت پياده‌سازي شده است[1]. ديابت در دودسته تقسيم‌بندي مي‌شوند: ديابت نوع يك ديابت نوع دو . اخيراً روش‌هايي خودكار سازي تشخيص بيماري ديابت توجه محققين را به خود جلب كرده است. كه از اين ميان تكنيك‌هاي داده‌كاوي نقش پررنگي را در اين تحقيقات دارند.زيرا استفاده از روش ها و ابزارهاي داده كاوي مي تواند به كشف الگوهايي منجر شود كه در تشخيص بيماري بسيار مؤثر هستند.داده‌كاوي يك فنّاوري نوظهور با پيشرفت هوش مصنوعي و تكنيك‌هاي پايگاه داده است كه در سازمان‌هاي تجاري مختلف استفاده مي‌شود تا كارآمدي و اثربخشي يك پروسه‌ي تجاري را بهبود ببخشد[2, 3].

ABSTRACT Medical diagnosis is one of the most important application areas for data mining tools that can be used with more precision and less error for the medical diagnosis. Many researchers have focused on the development of data mining methods with the purpose of improving accuracy in the mentioned area. It can be examined by using these methods that whether people have been affected from the diabetes or not. There are various techniques available in this field, that each one has its own advantages and disadvantages. In this regard, the present study has aimed to predict diabetes mellitus with the weighted averaging method and through by a combination of three algorithms in the decision tree called ID3, C4.5 and C5.0 and thus the accuracy prediction index has been improved. The proposed approaches in this study include maximum votes, simple averaging, and the weighted averaging methods. The conducted calculations on the data set of diabetes mellitus using computational approaches show that the best performance is related the weighted averaging model that is equal to 92.03%. In addition, there are also the available techniques in the literature of the subject show that the weighted averaging method has been superior among the introduced conventional models in previous researches and has obtained a higher accuracy. Keywords: Diabetes disease, Decision Tree, Genetic Algorithm, Supervised learning