Attentional Bi-LSTM for Multivariate Time Series Forecasting on Edge Devices: A Case Study on NanoPi Neo Plus2

Le résumé fourni par la source

This study presents a resource-aware deep learning pipeline for next-step multivariate time series forecasting, featuring an attentional bidirectional long short-term memory architecture. The proposed model is designed to capture both forward and backward temporal dependencies while dynamically focusing on the most salient time steps using a Bahdanau-style attention mechanism. We first evaluate the method on the widely used Jena Climate dataset, then extend the study to a larger real-world meteorological dataset from California. This combination provides both a standard benchmark and a more challenging real-world test case, where the model demonstrates its superior predictive accuracy compared to state-of-the-art models, including CNN-RNN, CNN-LSTM, and Stacked-LSTM baselines. To ensure practical applicability in real-world, resource-constrained environments, the entire model is optimized and deployed on a NanoPi Neo Plus2 board—an ARM-based 64-bit single-board computer with limited computational resources. Our implementation leverages lightweight inference techniques and efficient model quantization to enable on-device prediction without cloud connectivity. The resulting system achieves competitive forecasting performance with minimal latency and power consumption, showcasing the feasibility of edge-AI solutions for environmental monitoring and smart sensing applications. Both quantitative and qualitative analyses confirm t he e ffectiveness a nd i nterpretability o f the proposed approach.The source code for this project is publicly available at: https://github.com/NavidH95/attentional-bilstm-for-edge-forecasting/tree/main.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.