Provide a method for recognizing trust in social networks according to individual and personal characteristics using a compatible neural-fuzzy inference system method

Abstract

In this article, we presented a method to detect trust in social networks according to individual and personal characteristics with the help of the adaptive neuro-fuzzy inference system method. The current research required a dataset, for this purpose we designed an online questionnaire and collected 1000 records with the variables of age, gender, occupation, hours of activity in the virtual space, the type of use of the virtual space and the type of relationships in the virtual space, this dataset is as reference data can be used for similar analyzes and has a high level of data security. First, we evaluated and descriptively analyzed the data set, for this purpose we used Excel and SPSS software, we modeled and analyzed using MATLAB simulation. To introduce change limits and fuzzy behavior for variables, dataset parameters were introduced to the algorithm using Bayesian membership functions. Due to the uncertainty of the type of membership functions, coverage of the space under control, less computational volume, reduction of analysis time and increase of accuracy, we used the deductive clustering method and trained the network using feedforward neural network and trained the network with data. We continued the training until we reached full convergence. We entered the test and check data and using the squared error performance function, we came to the conclusion that with the method used in this research, it is possible to predict people's trust in each other in virtual space with an error of less than 1.5%.

Keywords

• social networks
• trust
• fuzzy system
• trust reversal