Artificial intelligence-based approach for Univariate time-series Anomaly detection using Hybrid CNN-BiLSTM Model

Abstract

Anomaly detection is a critical issue that has been extensively researched across a wide range of study fields and application domains. The main purpose of an anomaly detection model is to determine which instances stand out as being dissimilar to all others. Such instances are known as anomalies. Due to the common need for analyzing huge real-world data sets, automated data-driven models based on artificial intelligence techniques have become the most trending field of research. Deep learning techniques outperform other traditional machine learning and statistical methods as the scale of data increases. They have the power to learn very complex hierarchical feature relations within high-dimensional input data. In this paper, we suggested a hybrid deep learning technique that combines one-dimensional convolutional neural networks (1D CNN) with bidirectional long short-term memory (BiLSTM) for the detection of anomalies in univariate time series. The suggested model (1D CNN-BiLSTM) was tested and verified by the benchmark datasets and compared to state-of-the-art algorithms. The experimental results ensured that the proposed method delivers on its promises.