Generalized Information Quality And Its Application

Artificial intelligence technology is constantly evolving,nowadays.We are already in a sea of information.But when we make decisions and investments,it is an inevitable problem to synthesize multiple levels and aspects of information to make the wisest and most appropriate decision for the situation at hand.In the application of multi-source information fusion,how to process the data collected by sensors is a crucial step for information fusion.First of all,the collected sensor information has to be processed in advance in that to ensure that the final fusion result is intuitive and obey common sense.However,in the actual application,due to the failure of the sensor itself,the physical characteristics of sensor or environmental uncertainty will lead to conflicting and uncertain data collected,so the collected data need to be evaluated to determine how much valid information is contained in it.If we do not distinguish directly take it and use it,it will lead to wrong decisions,investment failure or resulting in irreparable errors.Therefore,the main research content of this paper is how to evaluate the information quality,and effectively carry out the weighted allocation,and make the proposed method applicable to the information fusion and related fields.Generalized information quality(GIQ)measures the magnitude of uncertainty of the information and illustrates the amount of valid information that can be provided by considering the degree of mutual support degree between the information from the collected each information source.In the field of multi-source information fusion,Dempster-Shafer evidence theory(D-S theory)has been widely used because of its excellent ability to model uncertain information.In addition,the Dempster's combination rule,provided in Dempster-Shafer evidence theory,conducts information fusion without prior knowledge.Therefore,this paper focuses on a study about how to effectively perform weighted fusion based on generalized information quality under the framework of D-S theory,and the specific research contents and methods are as follows.(1)Research on weighted fusion based on all kinds of situations on the basis of dynamic weight assignmentA method of dynamically assigning weights to multiple information sources is proposed.With the aid of dynamic weight assignment method,the magnitude of different information source can be reflected and it can be seen as their effectiveness in influencing the final decision.The first step is to analyze the collected information and determine the relationship between the data collected from each source.Next,assigning weights by the degree of mutual support and GIQ.Finally derive the final fusion results with the help of Dempster's combination rule.(2)Research on weighted fusion method on the basis of Tsallis entropyAn algorithm of measuring the certainty,included in the collected probability distributions,is proposed.The algorithm is on the framework of Tsallis entropy and GIQ.Inspired by the idea of multiple fractals,Tsallis presented the idea of Tsallis entropy.The Tsallis entropy is a generalization of the Boltzmann-Gibbs-Shannon(BGS)entropy,which degenerates to the BGS entropy when a certain condition is met.This paper gives the definition of information volume which is the description of the certainty the information contained.Then,considering the usability of information from the perspective of the interrelationship between information.Last but not the least,the weight allocation method is determined by information volume and the value of GIQ.The proposed algorithm model is compared with other weighted fusion methods by constructing arithmetic examples.Besides,the results of the comparison experiments show that the newly proposed algorithm has better performances and higher accuracy.Furthermore,we conducted a set of comparison experiments based on publicly available datasets to further highlight the superiority of the presented novel methods.By constructing different levels of conflicting information to achieve the comparison effect,the experimental results demonstrate that our proposed algorithm has a broader range of application,and not only can achieve effective fusion of information in a conflicting environment,but also converge faster and with higher accuracy.