تشخيص هوشمند بيماري هپاتيت با استفاده از آناليز اجزاي اصلي و هم جوشي طبقه بندي كننده ها

A new intelligent hepatitis diagnosis using principal component analysis and classifiers fusion

ﺳﺎﺑﻘﻪ و ﻫﺪف: در ﺳﺎل ﻫﺎي اﺧﯿﺮ، ﺑﯿﻤﺎري ﻫﭙﺎﺗﯿﺖ در ﺟﻬﺎن ﺑﺴﯿﺎر ﺷﯿﻮع ﭘﯿﺪا ﮐﺮده اﺳﺖ. ﺗﺸﺨﯿﺺ ﺻﺤﯿﺢ ﺑﯿﻤـﺎري ﻫﭙﺎﺗﯿﺖ ﮐﺎر ﺳﺎده اي ﻧﻤﯽﺑﺎﺷﺪ. ﻫﺪف از اﯾﻦ ﻣﻘﺎﻟﻪ، اراﺋﻪ ﯾﮏ ﺳﯿﺴﺘﻢ ﻫﻮﺷﻤﻨﺪ ﻣﺒﺘﻨﯽ ﺑﺮ ﺗﮑﻨﯿﮏ ﻫﺎي ﯾـﺎدﮔﯿﺮي ﻣﺎﺷـﯿﻦ ﺟﻬﺖ ﺗﺸﺨﯿﺺ ﺑﯿﻤﺎري ﻫﭙﺎﺗﯿﺖ ﻣﯽباﺷﺪ. ﻣﻮاد و روش ﻫﺎ: اﻟﮕﻮرﯾﺘﻢ ﭘﯿﺸﻨﻬﺎدي ﺷﺎﻣﻞ ﺳﻪ ﻣﺮﺣﻠﻪ اﺳﺎﺳﯽ ﻣﯽ ﺑﺎﺷﺪ: ﮐﺎﻫﺶ اﺑﻌـﺎد، ﻃﺒﻘـﻪﺑﻨـﺪي و ﻫـﻢ ﺟﻮﺷـﯽ ﻃﺒﻘﻪ ﺑﻨﺪي ﮐﻨﻨﺪه ﻫﺎ ﺑﺎ ﯾﮏ دﯾﮕﺮ. ﻣﺠﻤﻮﻋﻪ داده ﻫﺎ از اﻧﺒﺎره داده ﻫﺎي ﭘﺎﯾﮕﺎه داده ي UCI ﮔﺮﻓﺘﻪ ﺷﺪه اﺳﺖ. در اﺑﺘﺪا ﺗﻤـﺎم دادهﻫﺎ ﻧﺮﻣﺎل ﺷﺪه اﻧﺪ. ﺳﭙﺲ ﺑﺎ اﺳﺘﻔﺎده از آﻧﺎﻟﯿﺰ اﺟﺰاي اﺳﺎﺳﯽ ﺗﻌﺪاد وﯾﮋﮔﯽ ﻫﺎ ﺑﻪ 10 ﮐﺎﻫﺶ ﭘﯿﺪا ﮐﺮده اﺳﺖ. در ﻣﺮﺣﻠﻪ ﺑﻌﺪ از ﺳﻪ ﻃﺒﻘﻪ ﺑﻨﺪي ﮐﻨﻨﺪه ﺟﻬﺖ ﻣﺪل ﺳﺎزي داده ﻫﺎ اﺳﺘﻔﺎده ﮔﺸﺘﻪ اﺳﺖ. ﺟﻬﺖ ﺑﻬﺒﻮد ﮐﺎراﯾﯽ و اﻃﻤﯿﻨﺎن ﺑـﯿﺶ ﺗـﺮ ﺑـﻪ ﻧﺘﺎﯾﺞ ﺳﯿﺴﺘﻢ، ﻧﺘﺎﯾﺞ اﯾﻦ ﺳﻪ ﻃﺒﻘﻪ ﺑﻨﺪي ﮐﻨﻨﺪه ﺑﺎ اﺳﺘﻔﺎده از رايﮔﯿﺮي وزن دار ﺑﺎ ﻫﻢ ﺗﺮﮐﯿﺐ ﺷﺪه اﺳﺖ. ﯾﺎﻓﺘﻪ ﻫﺎ: اﻟﮕﻮرﯾﺘﻢ ﭘﯿﺸﻨﻬﺎدي ﺗﻮاﻧﺴﺖ ﺑﺎ اﺳﺘﻔﺎده از اﻋﺘﺒﺎرﺳﻨﺠﯽ ﻣﻨﻘﻄﻊ 10 ﻻﯾﻪ، دﻗـﺖ 96/32 را اراﺋـﻪ دﻫـﺪ ﮐـﻪ ﻧﺴﺒﺖ ﺑﻪ ﮐﺎرﻫﺎي ﻣﺸﺎﺑﻪ ﻧﺘﯿﺠﻪ ﺧﻮﺑﯽ ﻣﯽ ﺑﺎﺷﺪ. ﻧﺘﯿﺠﻪ ﮔﯿﺮي: ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﻧﺘﺎﯾﺞ، ﺳﯿﺴﺘﻢ ﭘﯿﺸﻨﻬﺎدي ﻣﯽ ﺗﻮاﻧﺪ ﺑﻪ ﻋﻨﻮان ﯾﮏ ﻫﻤﯿﺎر ﻫﻮﺷﻤﻨﺪ ﺟﻬـﺖ ﺗﺸـﺨﯿﺺ ﻧﻬـﺎﯾﯽ ﭘﺰﺷﮑﺎن ﻣﻮرد اﺳﺘﻔﺎده ﻗﺮار ﮔﯿﺮد.

Introduction: In recent years, hepatitis diseases have become prevalent in the world. The correct diagnosis of hepatitis disease is not a straight task. The goal of this paper is to introduce a new intelligent system for automatic hepatitis diagnosis based on machine learning approaches. Materials and Methods: the proposed approach consists of three stages, namely dimension reduction, classification, and fusion of classifiers. The hepatitis disease features were obtained from UCI machine learning repository. First, features have been normalized. Then, the number of these features is reduced to 10 from 19 by principal component analysis. In the next step, the reduced features are fed to three classifiers. Finally, a classifiers fusion to improve the efficiency and more reliable results using majority voting is presented. Results: the proposed approach obtained a classification accuracy of 96.32 via 10 fold cross validation. Conclusion: according to the results, the proposed system can be used as an intelligent partner for the final hepatitis diagnosis by physician.