Research On Recommendation Algorithm And Cooperation Behaviour Based On The Complex Network

With the rapid development of society and economy, great changes have taken place in our life-styles. The analysis and understanding of human behavior has become an important research topic. As we all know, the real world is composed of many complex systems, and they can be abstracted into various kinds of complex networks. In recent years, as the rapid development of complex networks, it has become an important research tool in revealing and settling some problems of nature and society. In addition, the collecting, statistics, computing, simulation and modeling of lots of data have become possible with the development and maturity of computer technology, which provides new ideas and analysis methods for other areas. In this thesis, we introduce some background knowledge of complex network firstly, and then we focus on the research of recommendation algorithms and cooperation behavior.With the explosive growth of information, it is more difficult to find the useful information for some one. The personalized recommendation is considered to be an effective way to sovle the information overload problem. Personalized recommender systems use the personal information of user (the historical record of his activities and possibly his profile) to uncover his habits and to consider them in the recommendation. Then analysis the characteristics of the data and establish the appropriate recommendation algorithm, which can give a recommendation list and help the given user find the information of interest. We propose an influence-based approach to investigate network-based recommendation systems. Different from the previous mass diffusion approach, we give a new expression of initial resource distribution and take into account the influence of resources associated with the receiver nodes, that is to say the relative resources is more important than the absolute resources. In the new algorithm, we introduce a tunable parameter to investigation the relation between the response of the resources receiver and his degree. According to ranking score and two measures about the personalization, we demonstrate that our method can outperform the previous methods greatly. It's found that there is an optimal initial resource distribution that leads to the best algorithmic accuracy and personalization strength. The optimal initial resource distribution indicates that we should increase the initial resource located on popular objects, rather than decrease them. In addition, we propose a novel method to compute the similarity between congeneric nodes in bipartite networks. Different from the standard cosine similarity, we take into account the influence of the public properties of these nodes, and suppose that the more popular of the public properties the smaller contribution to their similarity. The more unpopular public properties are more important for the similarity. Substituting this new definition of similarity for the standard cosine similarity, we propose a modified collaborative filtering (MCF). Based on a benchmark database, we demonstrate the great improvement of algorithmic accuracy for both user-based MCF and object-based MCF.Many complex systems can be abstracted into specific networks, and some interactions on the network can be abstracted as game among individuals, which can help people to understand the cooperation behavior in nature and society. The instabilities and fluctuations of individual fitness induced by some uncertain factors are ubiquitous in the real world. In this thesis we study how the individual fitness fluctuations affect the evolution of cooperation taking place in two-dimensional square lattices and Newman-Watts small-word networks. We introduce two tunable parametersαandβto control the fluctuation scope of agents and the fluctuation amplitude of fitness, respectively. Interestingly, the cooperation can be promoted in such a situation thatβis not too high andαis not too low, which indicates that the fluctuations play a significant role in the emergence of cooperation and the proper fluctuations can improve the cooperation of a system. The promotion of cooperation is explained by the feedback mechanism. Our results may be helpful in understanding the role of fluctuations in emergence of cooperation in the real world. In addition, we study the effect of heritability on the evolution of spatial public goods games and spatial personer's dilemma game. As we all know, heritability is ubiquitous within most real biological and social systems. Heritable traits can be fitness, strategy, and the way of strategy adoption for evolutionary games. In our model, the fitness of players is determined by the payoffs from current interactions and their history. Based on extensive simulations, we find that the density of cooperators is enhanced by increasing the heritability of players over a wide range of the multiplication factor. We attribute the enhancement of cooperation to the heritability that can stabilize the fitness of players, and thus prevents the expansion of defectors effectively. Finally, we study the effect of geographic distance on the evolution of the naming game. It is well known that the development of human language is a complex and long process. Languages are different in different ares. With the development of society, different cultures and languages infiltration have become more and more universal. The geographic distance is an important factor in the formation and development of language. We study the effect of geographic distance on a simple naming game, and simulation results show that the proper length of added shortcuts will lead to a faster convergence. In addition, we have reported the dynamical behavior of statistical quantities, including the total memory (inventory) of agents, success rate of negotiations, the number of total different invented words, and the convergence time.