International Journal of Advanced Engineering Research and Science (IJAERS) [Vol-4, Issue-2, Feb- 2017] https://dx.doi.org/10.22161/ijaers.4.2.11 ISSN: 2349-6495(P) | 2456-1908(O) www.ijaers.com Page | 57 Email Analysis for Community Detection - Multi-User Perspective S. Karthika, P. Saranya Department of Information Technology, SSN College of Engineering, Chennai, Tamil Nadu, India Abstract— Email classification has the capacity to group emails and user as community which is based on communication arrangement. Personalized network is used to understand the behavior of each user in an email and it analyze the various or different aspect of community structure e.g. the person who is having the same likeliness in a various social relations. The proposed system extracts single user or multi user from the email corpus using statistical analysis. This methodology uses a multi-user personalized email community detection method and tracks the email user it should be grouped. It also depends on their structural and semantic intimacy. Multi-user personalization concept used to find out the structure of community with fuzzy information i.e., an incomplete set of email details. The interactions are visualized as social graphs. Keywords— Community Detection, Multiuser Community, Pattern of Interest, PI-Net I. INTRODUCTION A social network is a social structure, which inter-relates nodes that are commonly individuals or organizations (such as friends, and co-workers) connected by interpersonal relationships . It is a platform that is used by people to build social relation with other people. Multi-user (or multi-account) represents the person who is having more than one email account. Multi-user personalization concept provides an approximation to the entire network community structure with an email corpus. The focus of this paper is used to group more than one email account information using related communication behavior [2]. It uniquely constructs an Undirected Weighted Graph (UW-graph) using emails meta-data from more than one account. These user grouping is done based on their similar communication patterns [12]. Community detection is used to detect the behavior of the person from multi-user email account. Each user personal email is represented as undirected weighted graph for structural and semantic intimacy. It analyzes community structures using multi-user information, i.e., email from multiple accounts, and it can be used to understand the network. People belonging to the same community are expected to have similar community behavior. The identified communities can be used to classify emails and determine prominent users [10]. It reflects similar neighborhood structure of email communication, e.g., frequent email exchanges with neighbors. The integration of semantic and structural information is necessary for community analysis [13]. The meta-data from an email's content includes subject length, text size, and attachment size, and TAG is the set of attribute and their labels. In order to extract the Communication Patterns of Interest (CPI) an email network is constructed using outgoing email i.e., email that contains sender information in metadata [9]. Using this meta-data it identifies the communication pattern behavior, for example, usually users exchanging the email in a certain time period. The rest of this paper is structured as follows. In Section 2, the different supporting work for community detection is presented. In Section 3, the framework for multi-user personalized email community is discussed and Section 4 explains the methodology for multi-user personalization. In Section 5, experimental study is discussed and in the Section 6 the paper is concluded with the future work. II. RELATED WORK The Communities are a union of nodes in a dark network which are identified to have common properties like interests between each other with denser connectivity than to the other nodes of the network. Such communities are likely to form a functional unit of a network and exhibit some interactions and knowledge exchange with each other as discussed by [1]. The evolution of communities [2] and various approaches opted for dealing with overlapping communities [3] [4] are important for the analysis of communities in social network. The existence of hierarchical structures in networks becomes a challenging issue in community detection where there is a possibility of a community being a part of another larger community. [5] Introduced a measure for evaluating the goodness of partitioning known as modularity which states that it is better to investigate community structure by provisioning nested hierarchy rather single community partitioning. The most prevailing and predominant technique which sociologists use in their