Long-Memory Reservoir Computing for Data-Scarce Dengue Forecasting

2026-07-14 · Source: stat.ML updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Health & Medical Research · Depth: Expert, extended

Summary

A novel long-memory reservoir computing framework, comprising Fractional ESN (fESN) and Wavelet ESN (wESN), addresses challenges in data-scarce dengue forecasting. Dengue incidence series are often short, noisy, non-stationary, nonlinear, and exhibit long-range temporal dependence, which traditional models like ARFIMA or deep neural networks struggle to capture simultaneously. The proposed framework integrates dedicated long-memory and short-memory Echo State Network (ESN) reservoirs with a ridge-regression readout. fESN incorporates fractional-differencing dynamics directly, while wESN extracts stable low-frequency components through wavelet smoothing. Theoretical analysis confirms standard ESNs induce short-memory processes, whereas fESN and wESN generate polynomially decaying dependence, consistent with statistical long memory. Empirical evaluations across multiple dengue datasets and forecasting horizons demonstrate that fESN and wESN consistently outperform statistical and deep learning baselines. The "memory-esn" Python package provides an implementation, and conformal prediction offers calibrated uncertainty intervals for public health decision-making.

Key takeaway

For public health analysts and data scientists tasked with forecasting dengue incidence in data-scarce regions, traditional linear or deep learning models often fall short due to complex data characteristics. You should consider adopting the Fractional ESN (fESN) or Wavelet ESN (wESN) frameworks. These models provide superior accuracy by jointly capturing long-range dependence, non-stationarity, and nonlinear dynamics, even with limited data. Furthermore, combining them with conformal prediction offers crucial calibrated uncertainty intervals for robust decision-making and resource allocation.

Key insights

Long-memory reservoir computing (fESN, wESN) effectively forecasts data-scarce, nonlinear time series with long-range dependence.

Principles

• Standard Echo State Networks exhibit short-memory behavior.
• Fractional differencing and wavelet smoothing encode statistical long memory.
• Reservoir computing offers data-efficient nonlinear time series modeling.

Method

Integrate dedicated long-memory and short-memory ESN reservoirs. fESN uses fractional-differencing for memory input; wESN applies wavelet smoothing before fractional filtering. Both train a ridge-regression readout.

In practice

• Utilize the "memory-esn" Python package for implementation.
• Combine with conformal prediction for calibrated uncertainty intervals.
• Wavelet ESN is robust to wavelet filter and decomposition level.

Topics

• Dengue Forecasting
• Reservoir Computing
• Echo State Networks
• Fractional Differencing
• Wavelet Transform
• Conformal Prediction
• Long-Range Dependence

Best for: AI Scientist, Research Scientist, Data Scientist

Related on AIssential

• fTNN: a tensor neural network for fractional PDEs — fTNN is a deterministic tensor neural network method for fractional PDEs, using geometry-adapted integration and singularity-aware trial functions for high accuracy.
• RhyMix: A Lightweight Adaptive Multi-Rhythm Network for Long-Term Time Series Forecasting — RhyMix combines explicit cyclic priors and multi-scale convolutions with adaptive gating for efficient, accurate long-term time series forecasting.
• Neural Negative Binomial Regression for Weekly Seismicity Forecasting: Per-Cell Dispersion Estimation and Tail Risk Assessment — Seismic count data's overdispersion requires per-cell Negative Binomial modeling for accurate extreme event risk assessment.
• Kolmogorov-Arnold Reservoir Computing — KARC bridges reservoir computing and KANs, using basis-function expansions for efficient, high-fidelity dynamical system forecasting and generative modeling.
• A time-series classification framework for individual-level absenteeism prediction under severe class imbalance — A Time Series Classification framework proactively predicts individual absenteeism by analyzing historical attendance sequences, outperforming traditional methods.
• JA-SIREN: Deterministic Initialization for Sinusoidal Networks via Spectral Matching — Deterministic initialization via spectral matching for sinusoidal networks significantly enhances INR reproducibility and performance.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by stat.ML updates on arXiv.org.