Three-Way Decision Model And Application Under Human-Computer Collaboration Mechanism

Machine intelligence and human intelligence have different concerns and working principles when performing the same type of task.Therefore,a reasonable cooperation mechanism can take the advantage of the two entities,and accomplish tasks efficiently.The Three-Way Decision(3WD)model considers the uncertainty and the risk in the decision process.It distinguishes the relatively certain area from the uncertain region,which are given different types of decisions.Such strategy is not only consistent with human behavioral habits,but also provides a basis for people to collaborate with machine.In humanware,humans are added into the system as service with corresponding mechanisms and interfaces.The task is distributed for human and machine based on the 3WD.Then the cooperation mechanism of humanware service and software service is constructed to reduce decision risk and improve efficiency.In term of facial image recognition,human and machine have their own strengths.If the identity authentication task is completed through the human-machine collaboration,both advantages can be utilized and the task can be performed effectively.Hence,the 3WD-based human-machine collaboration mechanism is utilized to complete the identity authentication task.In order to evaluate and select appropriate Agents from the intuitionistic fuzzy information system,an Agent evaluation method is constructed based on the intuitionistic fuzzy TOPSIS.Voting is a common way to synthesize the reviewers' opinions.Voting results are Intuitionistic Fuzzy Numbers(IFNs),which are stored in an intuitionistic fuzzy information table.The proposed method aims to comprehensively evaluate the Agents' capabilities using TOPSIS based on the principals of IFNs.The selected candidate will participate in the collaboration and complete the task together with the machine intelligence.How to achieve the minimum total cost is a pivotal problem in the collaboration.Accordingly,the working areas of human and machine are sparated based on 3WD.The relatively certain region of the algorithm is set as the working region of machine intelligence,and the uncertain region(boundary region)as the human working region.By adjusting the boundary region in the three-way decision,the total decision cost is minimized.In the task execution phase,an appropriate cooperation mechanism is established to realize the interaction between human intelligence and machine intelligence.Considering the human preferences and interests,the sample in the human working area is pre-processed by software,which will improve the efficiency of the entire decision-making process.It requires different amount of information to distinguish different facial images.Since the acquisition of information often takes time in reality,the decision is given gradually with the supplement of information.The face recognition process of human beings is a sequential decision process based on multi-granularity facial images.However,the existing sequential three-way decision is not applicable for the images.To simulate the human decision process,a multi-granular image features definition based on the total order relation is proposed.Then the sequential three-way decision model for facial images is put forward based on the features.The multi-granular features are extracted by subspace methods and the sequential decision process is constructed on the multi-granular features.As the amount of information increases,the decision-making cost of machine intelligence decreases while the time cost increases.The sequential three-way decision will comprehensively take into account two kinds of costs and cease the decision process at the appropriate information granularity to achieve the minimum total cost.In the sequential decision-making process,how to rationally distribute the task is the key issue in the collaboration.In the procedure,as the boundary region becomes smaller,the necessary workload of experts and the cost of decision-making will decrease.To accomplish the identity authentication task,a human-machine cooperation mechanism based on the sequential three-way decision is presented in this work.At each granularity,the total decision cost is achieved by adjusting the borderline of the two working areas.Afterwards,the sequential decision balances the decision cost and time cost.It will stop at the appropriate granularity,so that the total cost is the smallest.In the sequential decision-making process,the risk costs and decisions in the previous steps also provide a reference for people's decision at the current information granularity.Multi-category is a common scenario in reality,but the existing three-way decision models cannot be directly applied to multi-category problems.To solve the problem,a multi-category three-way decision model is put forward and the corresponding human-machine collaboration mechanism is established.On the basis of comparing the decision risks,the decision regions and decision rules are constructed directly in the multi-category scenario.In order to pursue the minimum total cost,the algorithm adjusts the task division through the trust threshold,which will affect the total decision cost.The boundary decision distinguishes the possible and impossible options.It provides a good reference for human decision making and reduces the workload of executor on each single sample.The human-machine cooperation mechanism is explored based on the three-way decision model,which is applied to the face recognition-based identity authentication task.The static two-category scene is expanded to the multi-granular and multi-class scene.The proposed model can effectively reduce the overall risk loss and improve the decision-making efficiency.It expands the development of three-way decision models and provides new ideas for human-machine collaboration.In the identity authentication task based on the face recognition,the proposed models take the advantage of machine intelligence,which also improve the customer satisfaction.