Abstract
The automatic diagnosis of the diseases using the computer is a fertile field for many researchers who are trying to design systems that help to reduce the mistakes made by inexperienced doctors or because of the influence of the pressures of life .This search deals with the use of (KNN) to diagnose the heart attack and then propose to improve the performance of KNN by using the genetic algorithm to control the basic joints of this method through determining the value of K, database segmentation, in addition to the reducing of the features. The proposed system has succeeded in increasing the accuracy of diagnosis from 75% to 100 %.
Keywords: K nearest, Genetic Algorithm, heart attack, automatic diagnosis.
الخلاصة
التشخيص الأوتوماتيكي للأمراض يشكل حقلا خصبا للعديد من الباحثين الذين يحاولون تصميم نظم تساعد في تقليل الأخطاء التي قد يرتكبها أطباء لا يملكون قدرا كافيا من الخبرة او تحت تأثير ضغوطات الحياة. يتناول هذا البحث استخدام طريقة المجاور الأقرب (KNN) في تشخيص مرض النوبة القلبية و من ثم يقترح تحسين أداء هذه الطريقة عن طريق استخدام الخوارزمية الجينية للتحكم بالمفاصل الاساسية لطريقة المجاور الاقرب وهي تحديد قيمه K وتقسيم قاعدة البيانات اضافة الى عملية تقليص الخصائص . النظام المقترح نجح في زيادة الدقة إلى 100 % بعد أن كانت 75 % فقط.
الكلمات المفتاحية: المجاور الأقرب, الخوارزمية الجينية, النوبة القلبية, التشخيص الأوتوماتيكي.
1. Introduction
The heart can be affected by different types of dangerous diseases Impact on human lives [1]. According to the World Health Organization (WHO), 12 million deaths occur worldwide every year and heart diseases are the reason [2]. Data mining is applied in medical field to predict diseases [3] such as diseases of the heart, and various types of cancer using medical datasets based on information collected from realistic people.
Computerized diagnostic uses medical databases to classify them inorder to build systems that are capable of diagnosing and addressing the emerging situations. There are many techniques that are used in the classification, some of them belong to a category of suppervised learning while the unsuppurevised class includes other techniques.the suppervised methods like neural networks or Bayesian classifier use Information extracted from the training data to classify another set of data dedicated to the testing stage. K nearest neighbor (KNN) is another example of suppervised methods but it does not extract any information from training set, it just use this set to make a comparison with a case that is waiting for classification and this is the so-called lazy learning [4].
The number of featurs in the database (the symptoms of the disease) is also affects the classification process. Depending on the method used, this number can result in time consumption and low accuracy. To get rid of the negative aspects of a the larg number of symptoms, a reduction method can be used to choose the symptoms which have a substantial influence in the diagnosis and neglect unimportant symptoms[4], for this purpose genetic algorithm (GA) is the good candidate to do so. This work is dedicated to study how to enhance the diagnosis of heart attack based on a Statlog database.
The paper is organized as follows:section 2 shelds the spot on the most important papers that is work on the same dataset . K-nearest method is presented in section 3. Genetic algorithm principles are reviewed in section 4.Details of the proposed method are described in section 5. Section 6 contains experimental results. Conclusions are presented in section 7.