Article Preview
Top1. Introduction
With the development of an open network environment, a dynamic, real, interconnected but difficult to accurately predict complex network environment is provided. This makes the agent technology face great challenges when dealing with open and dynamically distributed, complex issues. Agents in the real world represent different interest entities, and the goal of each agent is to maximize the interests of the entity they represent. Sometimes agents will inevitably produce various forms of trust and security issues in order to complete tasks or maximize their own interests. Therefore, it is very important to study the trust model based on complex network environment.
P2P lending refers to complete lending transactions between individuals that occur directly through a lending platform without the involvement of a bank and other financial institutions. The world’s first P2P lending platform, Zopa, was launched in the UK in 2005. Zopa acts as an information proxy, in which a loan application will be released to the website after approved by the platform, and investors choose to lend money after reading the relevant information on the website. Prosper is the first P2P lending platform in USA and is a pure mediation platform. It requires borrowers who wish to apply for a loan to have a minimal 640 FICO credit score. Prosper will assess the credit rating of borrowers by aggregating data from credit institutions, historical trading data, and personal credit score. Prosper will set the lending rates in advance based on the default risk which reflects the credit rating of the borrowers. The loan information released via the Prosper platform includes the loan amount, loan interest rates, borrowers’ debt/income ratio, credit rating, social networks, and historical transaction records, all anonymously. The borrowers’ personal credit record will be affected if payments are overdue for two months once a loan has been obtained.
The earlier Beth model (Javanmardi, 2014) used the concept of experience for the first time to describe and measure trust relationships. The model divides the experience into affirmative experience and negative experience according to the success or failure of the task, distinguishes direct trust from recommended trust, and gives two comprehensive calculation formulas of trust. Although the influence of different recommendation paths on the recommendation trust is discriminated, it is not advisable to use the 100% trust recommendation node in calculating the recommendation trust. The subjective trust model takes into account the uncertainty of subjective trust, and extends the binary logic and probability theory to more accurately describe human perception of trust. However, it does not distinguish between positive and negative evidence, which is not conducive to punishing malicious behavior (Meng, 2018).
The main objective of this paper is to develop an investor trust model and implement it using appropriate technology and validate it. This will provide a theoretical reference framework for conducting subsequent research in studying investor behavior. The main contribution of this paper is the development of an agent-based complex network testing investor trust model to explore the characteristics of investors' trust change in a network environment. The advantage of the agent-based implementation is that it can effectively improve the success rate of executing tasks and minimize the distance between cooperative agents.
Our work is organized as follows: The literature review is introduced in Section 2. The node-linking mechanism in P2P networks is described in in Section 3. Simulation studies are stated to verify the effectiveness of the design in Section 4. Indicators are discussed in the simulation experiment to measure the performance of the algorithm in Section 5. Finally, Conclusions are given in Section 6.