Problem
Research questions and friction points this paper is trying to address.
stock return prediction
neuroevolutionary architecture search
edge computing
online learning
ensemble methods
Evolve on the Host, Predict on the Edge: Deploying Online Neuroevolutionary Architecture Search for Cross-sectional Stock Return Prediction
🤖 AI Summary
This study addresses the challenges of large model sizes, high online update costs, and fixed architectures in predictive modeling by proposing a cloud-edge collaborative paradigm characterized as "host evolution, edge prediction." Methodologically, the ONE-NAS algorithm is employed on the host to perform online neural architecture search for compact recurrent networks, combined with an island-model population ensemble to enhance generalization. The optimized models are deployed via TCP/IP onto Raspberry Pi devices for real-time inference. Experimental results demonstrate that this architecture achieves single-sample prediction in merely 24.6 ms with a net strategy return of +27.5%, significantly outperforming LSTM and GRU baselines as well as single optimal models. By overcoming the limitations of conventional static architectures, this work enables efficient, real-time stock return prediction on resource-constrained edge devices.
📝 Abstract
Accurate forecasting models are usually large, expensive to update online, and fixed in architecture once trained. We apply ONE-NAS, an online neuroevolutionary architecture search that evolves a population of small recurrent networks as each window of data arrives, to daily cross-sectional stock return prediction, and pilot it on a host and endpoint pipeline: the host runs the search and ships each generation's champion genomes over TCP/IP to a Raspberry Pi 4B, which predicts online. On the Pi a single champion predicts a 50-stock window in 24.6~ms and the ensemble of 40 island champions in 556~ms, far inside the daily decision cycle. On four panels of US mid-cap equities over 2022--2024, reading the population as a rank-mean ensemble of island champions returns $+27.5\%$ net of realised transaction costs, against $+11.3$ to $+14.8\%$ for online LSTM, online GRU and monthly-retrained LSTM baselines and $+4.5\%$ for the single best genome used in prior ONE-NAS work.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Online Neuroevolutionary Architecture Search
Edge Computing
Cross-sectional Stock Return Prediction
Rank-mean Ensemble
Host-Edge Pipeline
🔎 Similar Papers
Optimizing Time Series Forecasting Architectures: A Hierarchical Neural Architecture Search Approach
2024-06-07arXiv.orgCitations: 0
