The Research Of Artificial Immunological Methods On Context-Aware Communication

The Context-Aware Communication(CAC) is a new mode for daily communications. It uses sensors and related technologies, to make the communication terminals and communi-cation systems have the ability to perceive the context of the people, and to make reasonable anticipation and response by adjusting their behavior and environment to serving and improv-ing people’s communication process. This improve the user experience of communication, which can take great benefits for the industry and scientific research.The CAC needs lots of sensors and data transmission network throughout the communica-tion space, and powerful computing devices. With the development of embedded chip circuit technology and Internet of Things and Mobile Internet, more powerful cluster computing and storage capacity emerged. All of these changes are speeding up the CAC’s developement. But the brain of CAC is still not well resolved. That is because the scattered and dynamic chang-ing characters of context, which make the traditional software engineering ineffective in that situation. Meanwhile, the fusion and reasoning of context need much artificial experience and expert knowledge, but which is hard to work in the real problems. The research area of CAC is still new, the problems are different with traditional engineering methods, so, we need some more novel ideas and views.To overcome obstacles, this paper analyzed the existing context-aware computing frame-works and related applications. Then from the characterstics of communication service, we eastablished a context-aware domain model for the CAC systems. Meantime, this paper intro-duced a context fusion method based on the artificial antigen presenting cells (APCs) which inspired by the principle of APCs in the innate immune system initegration of different antigens and signals. For the other question of context reasoning, a collaborative filtering method based on artificial immune network was proposed. This paper completed the following work:1) Built a domain model of CAC systems:By analysis of commonality and variability between CAC applications, and the difference between CAC and traditional communications services, applications and other CAAs. We made the division of CAC system boundary, ex-tracted the entities and relationships between them, built the domain model. Effects of cloud computing, social networking were also disscussed. This model is helpful for designing com-puting framework of CAC, to improve the degree of software reuse.2) Built a computing framework for CAC systems:A framework was proposed for the CAC systems, which was based on the domail model discussed before. The framework was inspired by other context-aware computing and applications, but especially situable for the CAC systems. It combined the communication terminals, the scheduling center and people in the communication platform.3) Proposed a Danger Theory based context fusion method:For fusing the complex, dy-namic and changing context, this paper draw on the principle of danger thoery in biological immunonlogy, as well as the mechanism of APCs in the innate immune system on efficient processing of different antigens and signals. This paper proposed a DT based fusion method for the communication context fusion. This method has the advantages of needless of artificial experience and expert knowledge, and makeing quick disscussion of user’s current context. Because of depending on the cell swarm intelligence, this alogrithm also has a good noise tol-erance.4) Proposed a context-aware collaborative filtering recommendation algorithm for context reasoning:The context reasoning need to consider the interest preference differences between users which can be targeted to make a reasonalble response to meet the needs of individual users. In this paper, we used the assumption that users had similar preferences within the plat-form should have similar preferences for different services in similar contexts, and proposed an immune network-based context-aware collaborative filtering method, which was used to predict the preference of target user in the new context for the various service response. This can be auxiliary for context reasoning to make appropriate responses. The method draws mu-tual stimulation and suppression between the antigens and antibodies in immune network for the memory mechanism, which helped the generation of the neighbor users for collaborative filtering, can effectively solve the sparsity problem. During processing the composite context of the target user, a divide and conquer method was used to make this method can be easy used offline or parallel processed, and improve the efficiency of the method.5) Three experiments were designed to verify the effectiveness of methods proposed in this paper:In experiment1, we stimulated the activities about watching movie, including activities of a target person from driving to the theater, watching movie, and leaving. Context fusion method based on artificial APCs was tested, and the experiment result shows that it can effectively determine the entire situation of different time and positions, which is helpful for the determination of incoming calls, and switching of phone scenario mode can be done automatically. In experiment2, by artificially increasing or reducing10-100%noise data in the simulation of experiment1, the compared result shows that the fusion results were affected little by the noise, so it has a strong capacity of noise tolerance. The third experiment to verify the immune network-based context-aware collaborative filtering recommendation method in a standard data set MovieLens. The comparative analysis shows that this method has maintain a higher recommendation accuracy.This paper starts its research from the computing framework and two important intelligent problems of context-aware communication. The analysis work of characterstics of this domain was done, and framework proposed in this paper is beneficial to design and develop applications of this field. The new information processing methods inspired from the biological immune have a better dynamic adapative capacity.