Detection and quantification of anomalies in communication networks based on LSTM-ARIMA combined model

Abstract

The anomaly detection for communication networks is significant for improve the quality of communication services and network reliability. However, traditional communication monitoring methods lack proactive monitoring and real-time alerts and the prediction effect of a single machine learning model on communication data containing multiple features is not ideal. To solve the problem, A prediction-then-detection anomaly detection method was proposed, and quantitative assessment of network anomalies was developed. Specifically, anomaly-free data was obtained by eliminating outliers, and the long short-term memory (LSTM) and autoregressive integral moving average (ARIMA) were combined via residual weighting to predict the future state of the key performance indicators (KPI) without outliers. Anomalies were identified using the error comparison between the prediction and actual values, and the network condition was quantified using the scoring method. It is observed that the proposed LSTM-ARIMA hybrid model has better prediction effect, which can well represent the performance of KPIs of the future state, and the prediction-then-detection anomaly detection method has excellent performance on both precision and recall.

Hot Deformation Behavior of Hastelloy C276 Alloy: Microstructural Variation and Constitutive Models

“…Firstly, the forgetting gate ( ) cleans up the useless input data from the last cell status, which can be expressed as [ 53 ] where is the activation function; is the inputting data; and represent the weight and bias of the forgetting gate, respectively.…”

“…The individual positions of a PSO algorithm can be updated by tracking the individual extreme value ( ) and the group extreme value ( ). The velocity and position equation of particle can be, respectively, expressed as [ 53 ] where is the speed of the particles, illustrates the inertia factor, the indicates a random number from 0 to 1, is the current position of the particles, and and are the learning factors.…”

“…Neural networks layer Operation point Vector transfer Firstly, the forgetting gate ( f t ) cleans up the useless input data x t from the last cell status, which can be expressed as [53]…”

“…The individual positions of a PSO algorithm can be updated by tracking the individual extreme value (pbest) and the group extreme value (gbest). The velocity and position equation of particle can be, respectively, expressed as [53]…”

Do Large Datasets or Hybrid Integrated Models Outperform Simple Ones in Predicting Commodity Prices and Foreign Exchange Rates?

“…Recent research has garnered significant interest from academics and practitioners due to the emergence of hybrid techniques that combine classical models with machine learning models. Hybrid prediction models have been utilized in various research fields, including meteorology, hydraulics, and exhaust emissions, for forecasting purposes (Chang et al 2019;Liu et al 2018;de O. Santos Júnior et al 2019;McNally et al 2018;Sadefo Kamdem et al 2020;Selvin et al 2017;Xue et al 2022;Sun et al 2022;Wu et al 2021;Wu and Wang 2022;Yu et al 2020;Zhang et al 2018Zhang et al , 2022Zolfaghari and Gholami 2021;Ma et al 2019;Dave et al 2021;Zhao et al 2022;Moustafa and Khodairy 2023;Zolfaghari and Gholami 2021). This study proposes several approaches that integrate machine and deep learning models with conventional statistical models, based on the assumption that time series can be decomposed into linear and nonlinear components or into time-dependent sums of frequency components and noise.…”