Research Of Interval-valued Fuzzy Cognitive Maps And Its Application On Ozone Prediction

A fuzzy cognitive map(FCM)can be understood as a graphical representation of the knowledge about or the perception of a given system.Fuzzy cognitive mapping is a combination of fuzzy logic and cognitive mapping.Cognitive mapping is based on graph theory,which is also the basis of most calculations and indices.Up to now,FCM has been successfully utilized in a number of fields,such as time series analysis,pattern recognition and risk assessment.However,there are some open problems on FCMs including quantifying the dynamic causal connections between concepts,describing the uncertain information,aggregating different maps.Interval-valued fuzzy cognitive map(IVFCM),a generalization of FCM,can provide more reasonable inference results by utilizing interval-valued fuzzy sets instead of fuzzy sets to represent the state of concepts and their mutual connections.Regarding the challenges of FCMs,we investigate the topic of ensemble IVFCMs and employ them to solve the ozone prediction problems.The main content of this thesis is as follows:⑴ Propose an ensemble IVFCM model.Firstly,we augment all IVFCMs into samesize maps.Secondly,a multi-objective model is established to objectively determine the weights of different maps.Thirdly,all the augmented maps are aggregated by all the augmented connection matrices and their weights.Finally,a number of examples are utilized to validate the performance the proposed model.⑵ Propose an IVFCM-based ozone prediction model.Concerning the prediction problem in which the data is derived from one station,we determine an IVFCM by learning the data of environmental monitoring from one station and predict the future ozone data.Regarding the prediction problem in which the data from multiple station,we design an ensemble-IVFCMs-based model to predict future ozone data.The detailed experiments indicate the feasibility of the proposed prediction models.