International Conference on Computer and Knowledge Engineering

Home / 15th International Conference on Computer and Knowledge Engineering

REMA: Reinforced Exponential Moving Average for Real-Time Anomaly Detection in Sensor Data

Authors :

Mohammad Hossein Jafari Naeimi¹ Ali Norouzi² Athena Abdi³

1- Computer engineering Department K.N.Toosi University of technology Tehran, Iran 2- Computer engineering Department K.N.Toosi University of technology Tehran, Iran 3- Computer engineering Department K.N.Toosi University of technology Tehran, Iran

Keywords :

Real-time Anomaly Detection،Embedded Systems،Sensor Data Anomalies،Statistical Methods،Adaptive Moving Average

Abstract :

This article introduces the REMA, a real-time Reinforced Exponential Moving Average utilized for anomaly detection in time series sensor data. The conventional statistical EMA model is augmented by reinforcement learning to improve its effectiveness. it employs a smoothing factor, a threshold, and a simple reward-punishment mechanism, which is applied to the smoothing factor. These attributes render REMA a precise and very rapid model, capable of detecting anomalies in the minimum possible time, recovering data in the event of sensor failure, and ensuring the uninterrupted functioning of the system. REMA exhibits robust anomaly detection capabilities, achieving an average accuracy of 99.6% with normal data and 92.5% with anomalous data. Due to its low computational complexity and high generalization ability, it serves as an exemplary choice in scenarios where speed is critical