An evolving Takagi-Sugeno model based on aggregated trapezium clouds for anomaly detection in large datasets

Anomaly detection is an important task for applications involving Big Data. Comparing with traditional method, anomaly detection in Big Data confronts growing amounts of data with high dimensionality and complex structures, which require more real-time analysis. This paper presents a fuzzy input-output system for anomalous data using electronic consumer records (ECR), a trapezium-cloud-map-filtration (TCMF) framework and a value mining model. ECRs are used to add or remove criteria based on consumers' consumption. In addition, MapReduce framework and trapezium clouds generated from each subsample are aggregated by using the aggregated trapezium cloud as a filter for each subsample. Then, a fuzzy logic-based value mining model is proposed based on Takagi-Sugeno model (T-S model) and trapezium clouds. This paper establishes a system that can improve decision-making accuracy by filtering large-scale data, and an illustrative example using a hotel booking situation is presented to verify the validity and feasibility of the proposed model. Finally, a comparative analysis is conducted between the proposed approach and existing methods.