اراﺋﻪي روش ﺟﺪﯾﺪ ﯾﺎﻓﺘﻦ ﻫﺴﺘﻪ ﺟﻮاﻣﻊ ﺑﺮاي ﺑﻬﺒﻮد اﻟﮕﻮرﯾﺘﻢ اﻧﺘﺸﺎر ﺑﺮﭼﺴﺐ در ﺷﺒﮑﻪﻫﺎي اﺟﺘﻤﺎﻋﯽ

A New Core Discovering Method for Improving Label Propagation Algorithm in Social Networks

ﮐﺸﻒ و ﺗﺤﻠﯿﻞ اﻃﻼﻋﺎت ﭘﻨﻬﺎن در ﺷﺒﮑﻪﻫﺎي ﭘﯿﭽﯿﺪه، ﻣﺨﺼﻮﺻﺎ ﺷﺒﮑﻪﻫﺎي اﺟﺘﻤﺎﻋﯽ ﺗﻮﺟﻪ ﺑﺴﯿﺎري از ﻣﺤﻘﻘﺎن را ﺑﻪ ﺧﻮد ﺟﻠﺐ ﮐﺮده اﺳﺖ. ﺗﺸﺨﯿﺺ ﺟﻮاﻣﻊ اﺑﺰاري ﺑﺴﯿﺎر ﻣﻬﻢ ﺑﺮاي ﭘﯽ ﺑﺮدن ﺑﻪ اﻃﻼﻋﺎت ﺷﺒﮑﻪ و ﺳﺎﺧﺘﺎر آن ﻣﯽﺑﺎﺷﺪ. روشﻫﺎي ﺑﺴﯿﺎري در اﯾﻦ زﻣﯿﻨﻪ اراﺋﻪ ﺷﺪه اﺳﺖ اﻣﺎ ﻫﻤﭽﻨﺎن اراﺋﻪي روشﻫﺎي ﺟﺪﯾﺪ ﺑﺎ دﻗﺖ و ﮐﯿﻔﯿﺖ ﺑﺎﻻ ﯾﮑﯽ از ﻣﺒﺎﺣﺚ ﺟﺬاب و ﭼﺎﻟﺶ ﺑﺮاﻧﮕﯿﺰ در زﻣﯿﻨﻪي ﺗﺸﺨﯿﺺ ﺟﻮاﻣﻊ اﺳﺖ. اﻟﮕﻮرﯾﺘﻢ اﻧﺘﺸﺎر ﺑﺮﭼﺴﺐ )LPA(، ﯾﮑﯽ از اﻟﮕﻮرﯾﺘﻢﻫﺎي ﻣﺸﻬﻮر در زﻣﯿﻨﻪي ﺗﺸﺨﯿﺺ ﺟﻮاﻣﻊ اﺳﺖ. وﺟﻮد ﻣﺎﻫﯿﺖ ﺗﺼﺎدﻓﯽ و ﻋﺪم ﺛﺒﺎت در ﻧﺘﺎﯾﺞ اﯾﻦ اﻟﮕﻮرﯾﺘﻢ، ﺳﺒﺐ ﺷﺪه اﺳﺖ ﮐﻪ LPA اﻟﮕﻮرﯾﺘﻢ ﻏﯿﺮ ﻗﺎﺑﻞ اﻃﻤﯿﻨﺎﻧﯽ ﺑﺎﺷﺪ. ﺗﺸﺨﯿﺺ درﺳﺖ ﮔﺮهﻫﺎي ﻫﺴﺘﻪ و اﺗﺨﺎذ ﯾﮏ ﺗﺮﺗﯿﺐ ﻣﺸﺨﺺ از ﮔﺮهﻫﺎ ﺑﺮ اﺳﺎس اﻫﻤﯿﺘ ﺸﺎن، ﭘﺨﺶ ﺻﺤﯿﺢﺗﺮ ﺑﺮﭼﺴﺐ ﮔﺮهﻫﺎي ﻫﺴﺘﻪ و ﺑﻬﺒﻮد ﻧﺤﻮه ي اﻧﺘﺨﺎب ﺑﺮﭼﺴﺐ ﺻﺤﯿﺢ ﻣﺒﺎﺣﺜﯽ ﻫﺴﺘﻨﺪ ﮐﻪ در اﯾﻦ ﻣﻘﺎﻟﻪ ﺑﻪ ﺑﺮرﺳﯽ آﻧﻬﺎ ﻣﯽﭘﺮدازﯾﻢ. در اﯾﻦ ﻣﻘﺎﻟﻪ اﻟﮕﻮرﯾﺘﻢ ﭘﯿﺸﻨﻬﺎدي ﮐﻪ ﺗﺮﮐﯿﺒﯽ از روش ﭘﺨﺶ ﺑﺮﭼﺴﺐ ﮔﺮهﻫﺎي ﻫﺴﺘﻪ و روش اﻧﺘﺸﺎر ﺑﺮﭼﺴﺐ اﺳﺖ ﺑﺎ اﻧﺘﺨﺎب ﮔﺮهﻫﺎي ﻫﺴﺘﻪ و ﭘﺨﺶ ﺑﺮﭼﺴﺐ آﻧﻬﺎ و ﺑﻬﺒﻮد ﻧﺤﻮهي ﺑﺮوزرﺳﺎﻧﯽ ﺑﺮﭼﺴﺐ ﮔﺮهﻫﺎ، ﺗﻮاﻧﺴﺘﻪ اﺳﺖ دﻗﺖ و ﮐﺎراﯾﯽ اﻟﮕﻮرﯾﺘﻢ LPA را ﺑﻬﺒﻮد ﺑﺨﺸﺪ. ﻧﺘﺎﯾﺞ آزﻣﺎﯾﺶﻫﺎي اﻧﺠﺎم ﺷﺪه ﺑﺮ روي دﯾﺘﺎﺳﺖﻫﺎي اﺳﺘﺎﻧﺪارد و ﻣﺼﻨﻮﻋﯽ و ﻣﻘﺎﯾﺴﻪ ﻧﺘﺎﯾﺞ ﺣﺎﺻﻞ، ﺑﯿﺎﻧﮕﺮ ﻋﻤﻠﮑﺮد ﺑﻬﺘﺮ اﻟﮕﻮرﯾﺘﻢ ﭘﯿﺸﻨﻬﺎدي ﻧﺴﺒﺖ ﺑﻪ اﻟﮕﻮرﯾﺘﻢ LPA و NIB-LPA اﺳﺖ.

Discovering and analyzing of hidden information in complex networks and especially social networks has attracted the attention of many researchers. Community detection is a very important tool for understanding network information and its structure. Many approaches have been proposed, but proposing new methods with high accuracy and high quality are still one of the most exciting and challenging topics in community detection field. Label Propagation Algorithm (LPA) is one of the most popular algorithms in the field of community detection. The random nature and instability in the results of this algorithm make the LPA unreliable. Determining the core nodes and adopting a specific order of nodes based on their importance, more accurate diffusion of the core node label, and improving the way of selecting the correct label, are the main topics that we will discuss in this paper. In this paper, the proposed method which is combination of label diffusion of core nodes and label propagation method, has been able to improve the accuracy and efficiency of LPA by selecting and diffusing the core nodes label and improving the label update strategy. The results of experiments performed on standard and synthetic datasets and comparing the results show that the proposed algorithm performs better than the LPA and NIB-LPA algorithms.