Empirical Analysis of Community Detection over Social Network using Intelligent Machine Learning Strategies

Abstract

Now-a-days Social Network and the Communication mediums are getting a drastic development according to its usage and the provision it offers to users for accumulating it. In a general concern, social network is the advanced boon to the information technology people as well as other commercial and non-commercial organizations to improve their marketing and attain some good benefits based on that. The term communication leads a world by connecting one another in a binding way without any range restrictions. Both these communication and social networking environments are tied up together to provide lots of benefits to the clients without any interference. In Social networking field, the main concern to take care with is Community Detection, because the communities leads a social networking platform in a huge pathway and provides great improvement in it. Usually communities are considered to be a group or set of associated peoples or modules/applications. The main purpose of this paper is to design a customized community detection scheme based on machine learning strategies. Generally this community detection schemes are developed a lot by several researchers but all are pointing out a single model of identifying community, for example, blood donor community identification, covid people community identification and so on. But this paper modify the way of community detection by introducing a new approach with the help of Artificial Intelligence, which is called as Deep Learning Enabled  Community Detection Scheme (DLeCDS), in which the proposed approach is quite different compare to all other classic approaches. This paper introduced an advanced DLeCDS to identify the community efficiently, in which the proposed scheme offers dynamic dataset appliance model, so, that the user can dynamically select the required dataset to process. The machine learning model is trained with some pre-defined datasets for Covid-19, Blood Donor Group Dataset, Traffic Violence Group Dataset and so on. These all are properly trained and once the user select the appropriate dataset, the proposed approach can easily process the result according to the selection of testing data with respect to selected training model. Simply, the proposed model of DLeCDS provides an efficient way to user to accumulate the required outcome based on the dynamic dataset selection and processing model. This approach guarantees the outcome accuracy at the range of 97 to 98% based on the testing data accuracy as well as the proposed paper results and discussion section proves the outcome with graphical view and estimations. The proposed model of Deep Learning Enabled Community Detection Scheme is compared with the classical techniques called Support Vector Classifier, Random Forest Classifier and the Neural Network algorithm, in which compared to all three the proposed approach outcome assures the resulting accuracy in good manner. The proposed model of DLeCDS has novelty in the range of above 95% with respect to its robustness and outcome stability, in which the time complexity is considered as the main concern over this approach. The time complexity can be eliminated by means of Fuzzy Time Estimation Model (FTEM), in which the estimation of time to detect the proper communities over the social network can be achieve with the help of this FTEM. The benefits of proposed approach is clearly illustrated on the resulting section with proper graphical proofs as well as the accuracy level comparisons with respect to classical machine learning models.