اجراي الگوريتم خوشه‌بندي بهبود يافته بر روي داده‌هاي ناباروري

Development of an Improved Clustering Algorithm for use on Sarem Hospital Infertility Data

داده‏كاوي تلفيقي از روش‌هاي هوش مصنوعي براي شناسايي اطلاعات يا استخراج دانش از داده‏هاست، به‌نحوي كه دانش حاصل در حوزه‏هاي تصميم‏گيري، پيش‏بيني، پيش‌گويي و تخمين مورد استفاده قرار گيرد. تحليل رفتار مشتريان، دسته‏بندي مشتريان، شناخت نيازهاي مشتريان و پيش‌بيني در مباحث پزشكي ازجمله كاربردهاي داده‌كاوي است. خوشه‏بندي يكي از روش‌هاي بدون نظارت الگوريتم‏هاي داده‏كاوي است كه به يافتن يك ساختار مشخص درون‌مجموعه‌يي از داده‏هاي بدون برچسب مي‏پردازد. يكي از الگوريتم‏هاي متداول خوشه‏بندي، الگوريتمk-means است. از معايب اين الگوريتم «انتخاب تصادفي خوشه‏هاي اوليه» در آغاز الگوريتم است كه موجب تفاوت نتيجه در هر بار اجراي الگوريتم مي‏شود. در اين پژوهش، با استفاده از الگوريتم سلسله‏مراتبي مدل جديدي ارايه شده كه مي‌كوشد مشكل الگوريتم k-means را برطرف سازد. در ادامه، نتيجه‌ي اجراي اين الگوريتم تلفيقي جديد روي داده‏هاي واقعي مربوط به «ناباروري بيمارستان صارم» ارايه شده است.

Nowadays, leading-edge advanced medical tools and new ways of communication are two important considerations in any medical discussions. Algorithms introduced in the field of data mining are able to appropriately interpret and analyze a variety of problems. Data mining is a category of methods helping extraction of information from given data in a way that the output is useful, either for decision making, prediction or estimation. Banking procedures, customer behaviour analysis, medical applications, classification of customers and their needs are some of the fields that employ data mining. In the medical field, data mining has been employed successfully in diagnosis and treatment of diseases. Along with global advances in medical science, researchers and the medical community in Iran have progressed notably in discovering new methods of dealing with infertility. Clustering is an unsupervised method used in data mining to partition a set of unlabeled objects by putting them into groups, such that elements in each group are more similar to each other, in some sense, compared to elements of other groups. One of the common clustering algorithms is known as K-means. Even though successful, the randomness embedded in the algorithm, while selecting initial clusters, causes a variation in results which is not desired. This body of work proposes a new method which uses a hierarchical algorithm to improve the initial cluster selections of K-means. This new ensemble method is applied to infertile patient data in Sarem Hospital and the results are shown. This data has been collected from patients with infertility problems being treated using the ICSI method.