BiLSTM-MLAM: A Multi-Scale Time Series Prediction Model for Sensor Data Based on Bi-LSTM and Local Attention Mechanisms

Yongxin Fan,Qian Tang,Yangming Guo,Yifei Wei

DOI: https://doi.org/10.3390/s24123962

IF: 3.9

2024-06-19

Sensors

Abstract:This paper introduces BiLSTM-MLAM, a novel multi-scale time series prediction model. Initially, the approach utilizes bidirectional long short-term memory to capture information from both forward and backward directions in time series data. Subsequently, a multi-scale patch segmentation module generates various long sequences composed of equal-length segments, enabling the model to capture data patterns across multiple time scales by adjusting segment lengths. Finally, the local attention mechanism enhances feature extraction by accurately identifying and weighting important time segments, thereby strengthening the model's understanding of the local features of the time series, followed by feature fusion. The model demonstrates outstanding performance in time series prediction tasks by effectively capturing sequence information across various time scales. Experimental validation illustrates the superior performance of BiLSTM-MLAM compared to six baseline methods across multiple datasets. When predicting the remaining life of aircraft engines, BiLSTM-MLAM outperforms the best baseline model by 6.66% in RMSE and 11.50% in MAE. In the LTE dataset, it achieves RMSE improvements of 12.77% and MAE enhancements of 3.06%, while in the load dataset, it demonstrates RMSE enhancements of 17.96% and MAE improvements of 30.39%. Additionally, ablation experiments confirm the positive impact of each module on prediction accuracy. Through segment length parameter tuning experiments, combining different segment lengths has resulted in lower prediction errors, affirming the effectiveness of the multi-scale fusion strategy in enhancing prediction accuracy by integrating information from multiple time scales.

engineering, electrical & electronic,instruments & instrumentation,chemistry, analytical

What problem does this paper attempt to address?

### Problems Addressed by the Paper This paper proposes a multi-scale time series prediction model based on Bidirectional Long Short-Term Memory (Bi-LSTM) and Local Attention Mechanism (LAM), named BiLSTM-MLAM. The main objective is to address the time series prediction problem in sensor network data and improve prediction accuracy. Specifically, the paper addresses the following key issues: 1. **Multi-scale Feature Fusion**: Through a multi-scale patch segmentation module, the model can capture data patterns and features at different time scales. 2. **Long Sequence Dependency**: Utilizing the Bi-LSTM structure, the model can consider both past and future contextual information at each time step, thereby better capturing long-term dependencies. 3. **Local Feature Extraction**: By introducing a local attention mechanism, the model effectively extracts local features in subsequence time series, enhancing the model's understanding of local features. Experimental results show that BiLSTM-MLAM outperforms six baseline methods on multiple datasets. Particularly, in predicting the remaining life of aircraft engines, it improves the Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) by 6.66% and 11.50%, respectively, compared to the best baseline model. Additionally, significant improvements were also achieved on the LTE dataset and load dataset. ### Summary This paper proposes a novel multi-scale time series prediction model named BiLSTM-MLAM, aiming to improve the time series prediction accuracy of sensor network data through multi-scale feature fusion, Bi-LSTM structure, and local attention mechanism. Experiments validate the superior performance of this model in various tasks.

BiLSTM-MLAM: A Multi-Scale Time Series Prediction Model for Sensor Data Based on Bi-LSTM and Local Attention Mechanisms

A dual-stage attention-based Bi-LSTM network for multivariate time series prediction

Effective LSTMs with Seasonal-Trend Decomposition and Adaptive Learning and Niching-Based Backtracking Search Algorithm for Time Series Forecasting

Fusion k-means clustering and multi-head self-attention mechanism for a multivariate time prediction model with feature selection

Time Series Prediction Based on LSTM and High-Order Fuzzy Cognitive Map with Attention Mechanism

Temporal Self-Attention-based Conv-LSTM Network for Multivariate Time Series Prediction

TE-LSTM: A Prediction Model for Temperature Based on Multivariate Time Series Data

Remaining Useful Life Prediction of Aero-Engine Based on Multi-Scale Channel Enhanced BiLSTM

SALSTM: Segmented Self-Attention Long Short-Term Memory for Long-Term Forecasting

RAdam-DA-NLSTM: A Nested LSTM-Based Time Series Prediction Method for Human–Computer Intelligent Systems

Multi-Resolution LSTM-Based Prediction Model for Remaining Useful Life of Aero-Engine

Atmospheric Temperature Prediction Based on a BiLSTM-Attention Model

Long Time Series Deep Forecasting with Multiscale Feature Extraction and Seq2seq Attention Mechanism

Novel double-layer bidirectional LSTM network with improved attention mechanism for predicting energy consumption

A Time Series Prediction Model Based on Long Short-Term Memory Networks

Multi-Scale Remaining Useful Life Prediction Using Long Short-Term Memory

Multi-LSTM: A Multiscale Long Short-Term Memory-Based Framework for Predicting Time-Series Transcriptomic Gene Expression

Multivariate Time Series Forecasting Based on Sliding Attention Correction LSTM Network

A deep LSTM‐CNN based on self‐attention mechanism with input data reduction for short‐term load forecasting

Spatiotemporally Multidifferential Processing Deep Neural Network and Its Application to Equipment Remaining Useful Life Prediction

Remaining Useful Life Prediction Based on Multi-channel Attention Bidirectional Long Short-term Memory Network