Online credit card fraud prediction based on hierarchical temporal memory model
Recent studies of the human brain have brought about a new understanding of the structural and algorithmic property of the neocortex. This understanding gave birth to the Hierarchical Temporal Memory (HTM) which holds a lot of promises in the area of time-series prediction and anomaly detection problems. This paper demonstrates the behaviour of an HTM model with respect to its learning and prediction of online credit card fraud. The model was designed using the object oriented analysis and design methodology. Java programming language was used for implementation and Matlab was used to carry out simulations. The resulting model demonstrated learning like that of the human brain using sparse data, hence, the model required less resources to evaluate big data. The model recorded higher accuracy of 92.3% compared to Artificial Neural Network model that recorded 89.6%, hence reducing cases of misclassification.
Keywords: Duty Cycle, Boost Value, Synapses, Learned Index