Classification of Sequence Data Using LSTM: An Application on Chaotic Sequences

Abstract

Abstract Chaos theory and chaotic systems are focused on the study of nonlinear dynamic systems. The chaotic sequences are patterns generated from the recursive nature of chaotic maps. Identification and classification of chaotic systems from a given chaotic sequences are the primary need for any chaotic system analysis. Also, to model a dynamic system that depicts the behavior of an existing system, the chaotic behavior followed by that system needs to be identified. So the problem of classifying these chaotic sequences is prime necessity to address many problems in chaos theory and related domain. In this work, we have used a deep learning model, which is called long short-term memory (LSTM) for this classification task. The model was applied to two different chaotic sequence datasets, one with 1-D sequences scattered over two classes, and the other holds 2-D sequences distributed among three classes. The model gave an average of 90% for one-dimensional sequence dataset and 95% for two-dimensional sequences.KeywordsChaos theoryChaotic sequencesSequence classificationLSTMDeep learning