Information Identification, Diffusion And Asset Pricing On Complex Networks

The development of the Internet and social media significantly changed theinteraction of information among individuals in the financial markets from threeaspects:1. Emerging a large number of "individual information sources" who haswide influence. Individual faces more information channels than ever before whenmaking decision;2. Market information becomes heterogeneous, instant andinteractive. The information transmission speed in stock market is greatly improved,which also brought the problems such as false information diffusion and informationmanipulation. Individuals must pay more attention on information identification whenmaking decision;3. Individuals have a wide range of information search ability,which help them to change information sources freely in the decision making process.Information diffusion network structure in financial market presents a complexdynamic evolution. These changes has produced important influences on the useageof information to make a decision process in financial markets, and further affect theformation of asset prices.From the point of view in this kind of change, it needs to answer two related questionsbelow at least to accurately describe the information diffusion process in financialmarket: How the agents send, transmit, distort the information and identify it in theprocess of information diffusion? What kind of information network structuresfollowed by stock investors changing information? So far, current studies only focuson one of the two questions. In this paper, using the advantage of agent-basedmodelling, we studied information network and heterogeneous information sourcesinfluence on uninformed agents in the artificial stock market. We established twoartificial stock markets. One of which is information diffusion in fixed networkstructure and the other is information diffusion in a dynamic evolution networkstructure. All of agents update their strategies through reinforcement learning andpredict next period price with genetic algorithm. In order to make this informationtransmission networks closer to the real market, agents interact with each otherthrough the Small-world Network in basic model. We tested the changing ofuninformed agents identification ability and asset price by controlling networkstructure conditions and heterogeneous information source conditions.Because of the introduction of heterogeneous information sources and complexnetwork structure, we obtained conclusions very different from the traditionalcheap-talk game model. We discover that the informationd diffusion networkstructure between individuals and heterogeneous sources can improve the uninformedagents identification ability. Uninformed agents can improve their identificationability with heterogeneous information sources. However, the threshold effects existin heterogeneous information sources, which means uninformed agents will notdistinguish dishonest informed agent with few heterogeneous information sources.Their identification ability will have a great improvement only if there exists enoughheterogeneous information sources in stock market. This article also examined the accuracy of information sources, the diffusion time, the information search ability andcosts to identify information sources. Finally, our studiy provides the necessarytheoretical basis and policy suggestions for false information diffusion preventing andinformation manipulation preventing.