تحليل داده هاي بيماران ديابتي در راستاي خوشه بندي و تجويز دارو براساس الگوريتم پيشنهادي

Analysis of Diabetic Patients' Data for Clustering and Prescription Drug Based on Proposed Algorithm

ﺧﻼﺻﻪ ﻣﻘﺪﻣﻪ دﯾﺎﺑﺖ ﯾﮏ اﺧﺘﻼل ﺳﻮﺧﺖ و ﺳﺎزي در ﺑﺪن اﺳﺖ ﮐﻪ ﺗﻮاﻧﺎﯾﯽ ﺗﻮﻟﯿﺪ ﻫﻮرﻣﻮن اﻧﺴﻮﻟﯿﻦ در ﺑﺪن از ﺑﯿﻦ ﻣﯽرود. ﻫﺪف ﮐﻠﯽ از اﻧﺠﺎم ﭘﮋوﻫﺶ ﺣﺎﺿﺮ ﮐﺸﻒ داﻧﺶ ﻧﻬﻔﺘﻪ در دادهﻫﺎي ﺑﯿﻤﺎران دﯾﺎﺑﺘﯽ اﺳﺖ، ﮐﻪ ﻣﯽﺗﻮاﻧﺪ ﺑﻪ ﭘﺰﺷﮑﺎن در ﺧﻮﺷﻪﺑﻨﺪي ﺑﯿﻤﺎران ﺟﺪﯾﺪ و ﺗﺠﻮﯾﺰ داروي ﻣﻨﺎﺳﺐ ﻣﻄﺎﺑﻖ ﻫﺮ ﺧﻮﺷﻪ ﮐﻤﮏ ﻧﻤﺎﯾﺪ. روش ﮐﺎر در اﯾﻦ ﻣﻘﺎﻟﻪ از اﻟﮕﻮرﯾﺘﻢ MR-VDBSCAN اﺳﺘﻔﺎده ﺷﺪه اﺳﺖ. ﭘﯿﺎدهﺳﺎزي اﯾﻦ اﻟﮕﻮرﯾﺘﻢ در ﺑﺴﺘﺮ ﻫﺪوپ ﻣﺒﺘﻨﯽ ﺑﺮ ﭼﺎرﭼﻮب ﻧﮕﺎﺷﺖ-ﮐﺎﻫﺶ ﻣﯽﺑﺎﺷﺪ. اﯾﺪه اﺻﻠﯽ ﺗﺤﻘﯿﻖ اﺳﺘﻔﺎده از ﭼﮕﺎﻟﯽ ﻣﺤﻠﯽ ﺑﺮاي ﯾﺎﻓﺘﻦ ﭼﮕﺎﻟﯽ ﻫﺮ ﻧﻘﻄﻪ اﺳﺖ. اﯾﻦ اﺳﺘﺮاﺗﮋي ﻣﯽ ﺗﻮاﻧﺪ ﻣﺎﻧﻊ از اﺗﺼﺎل ﺧﻮﺷﻪﻫﺎ ﺑﺎ ﭼﮕﺎﻟﯽﻫﺎي ﻣﺘﻔﺎوت ﺷﻮد. ﻧﺘﺎﯾﺞ اﻟﮕﻮرﯾﺘﻢ ﻣﻮردﻧﻈﺮ ﺑﺮ روي دﯾﺘﺎﺳﺖ اﻧﺘﺨﺎب ﺷﺪه، ﺗﺴﺖ و ارزﯾﺎﺑﯽ و ﻧﺘﺎﯾﺞ ﻧﺸﺎن از دﻗﺖ ﺑﺎﻻ و ﮐﺎراﯾﯽ و ﻣﻘﯿﺎسﭘﺬﯾﺮي آن دارد. ﻧﺘﺎﯾﺞ ﺑﺪﺳﺖ آﻣﺪه ﺑﺎ ﻧﺘﺎﯾﺞ اﺟﺮاي ﺧﻮﺷﻪﺑﻨﺪي k-Means ﻣﻘﺎﯾﺴﻪ ﺷﺪ، اﻟﮕﻮرﯾﺘﻢ MR-VDBSCAN در ﻣﻘﺎﯾﺴﻪ ﺑﺎ آن از ﺳﺮﻋﺖ اﺟﺮا ﺑﺎﻻﺗﺮ ودﻗﺖ ﺗﺸﺨﯿﺺ ﺑﻬﺘﺮي ﺑﺮﺧﻮردار ﻫﺴﺖ و ﻫﻤﭽﻨﯿﻦ ﺗﻮاﻧﺎﯾﯽ ﺗﺸﺨﯿﺺ ﺧﻮﺷﻪﻫﺎ ﺑﺎ ﭼﮕﺎﻟﯽ ﻣﺘﻔﺎوت ﺑﺮﺗﺮي اﯾﻦ اﻟﮕﻮرﯾﺘﻢ ﻧﺴﺒﺖ ﺑﻪ اﻟﮕﻮرﯾﺘﻢ ﻣﻮرد ﻣﻘﺎﯾﺴﻪ اﺳﺖ. ﻧﺘﺎﯾﺞ ﻧﺸﺎن ﻣﯽدﻫﺪ ﮐﻪ اﻟﮕﻮرﯾﺘﻢ MR-VDBSCAN ﻣﯽ ﺗﻮاﻧﺪ ﻋﻤﻠﮑﺮد ﺑﻬﺘﺮ را از ﺳﺎﯾﺮ اﻟﮕﻮرﯾﺘﻢﻫﺎ ﻓﺮاﻫﻢ ﮐﻨﺪ. ﺑﻪ ﻃﻮر ﺧﺎص، ﺷﺒﺎﻫﺖ اﻟﮕﻮرﯾﺘﻢ ﭘﯿﺸﻨﻬﺎد ﺷﺪه 97. ﺑﺮاي ﻣﺠﻤﻮﻋﻪ دﯾﺎﺑﺖ اﺳﺖ. ﻧﺘﯿﺠﻪ ﮔﯿﺮي ﻧﺘﺎﯾﺞ ﻧﺸﺎن ﻣﯽدﻫﺪ ﮐﻪ ﮐﻪ اﻟﮕﻮرﯾﺘﻢ MR-VDBSCAN ﻧﺴﺒﺖ ﺑﻪ اﻟﮕﻮرﯾﺘﻢ K-means ﺧﻮﺷﻪ-ﺑﻨﺪي ﺑﻬﺘﺮي را اﻧﺠﺎم ﻣﯽدﻫﺪ و ﻣﯽﺗﻮاﻧﺪ ﺑﯿﻤﺎران را در زﯾﺮﮔﺮوﻫﻬﺎﯾﯽ ﻗﺮار دﻫﺪ ﮐﻪ ﭘﺰﺷﮑﺎن را در ﺗﺠﻮﯾﺰ ﯾﺎري ﻧﻤﺎﯾﺪ. ﻧﺘﯿﺠﻪ ﭘﯿﺶﺑﯿﻨﯽ ﺷﺪه ﺑﺮاي ﺗﺸﺨﯿﺺ اﯾﻨﮑﻪ ﮐﺪوم ﮔﺮوه ﺳﻨﯽ و ﺟﻨﺴﯿﺖ ﺑﯿﺸﺘﺮ ﺗﺤﺖ ﺗﺎﺛﯿﺮ دﯾﺎﺑﺖ ﻗﺮار دارﻧﺪ، اﺳﺘﻔﺎده ﻣﯽﺷﻮد.

Introduction Diabetes is a metabolic disorder in the body that is impaired by the ability to produce insulin hormone. The main purpose of the present study is to discover the hidden knowledge in the data of diabetic patients, which can assist clinicians in clustering new patients and prescribing appropriate medication according to each cluster. Methods: In this paper, we use MR-VDBSCAN algorithm. The implementation of this algorithm is based on the map-reduce framework of Hadoop. The main idea of the research is to use local density to find the density of each point. This strategy can prevent clusters from joining at different densities. Results: The algorithm is based on the selected dataset, tested and evaluated, and the results show high accuracy and efficiency. The results were compared with the results of k-Means clustering, The MR-VDBSCAN algorithm has a higher execution speed than that of the algorithm and has the ability to detect clusters with different density of superiority of this algorithm than the comparable algorithm. The results show that the MR-VDBSCAN algorithm can provide better performance than other algorithms. In particular, the similarity of the proposed algorithm is 97% for the diabetes set. Conclusion: The results show that the MR-VDBSCAN algorithm performs better clustering than the K-means algorithm and can place patients into subgroups that assist physicians in prescribing.