Toward the use of neural networks for influenza prediction at multiple spatial resolutions

Abstract

Mitigating the effects of disease outbreaks with timely and effective interventions requires accurate real-time surveillance and forecasting of disease activity, but traditional health care–based surveillance systems are limited by inherent reporting delays. Machine learning methods have the potential to fill this temporal “data gap,” but work to date in this area has focused on relatively simple methods and coarse geographic resolutions (state level and above). We evaluate the predictive performance of a gated recurrent unit neural network approach in comparison with baseline machine learning methods for estimating influenza activity in the United States at the state and city levels and experiment with the inclusion of real-time Internet search data. We find that the neural network approach improves upon baseline models for long time horizons of prediction but is not improved by real-time internet search data. We conduct a thorough analysis of feature importances in all considered models for interpretability purposes.

Influenza’s Plummeting During the COVID-19 Pandemic: The Roles of Mask-Wearing, Mobility Change, and SARS-CoV-2 Interference

“…We explored two self-correcting regularized multiple regression models to forecast the weekly influenza activity. Both regression models are dynamically trained and regularized using the LASSO method; unlike autoregressive integrated moving average models [53] , [54] , they allow the self-selection of multiple lags (up to 52) of influenza activities as model inputs. The regression models are described as follows:…”

The Incoming Influenza Season — China, the United Kingdom, and the United States, 2021–2022

“…To forecast influenza activity under different NPIs, we explored a self-correcting regularized multiple regression. The approach used time series data to predict future points in the series and has been widely used to forecast influenza activity ( 6 - 7 ). Unlike the conventional autoregressive integrated moving average method, it allowed for self-selection of multiple lags of past observations as model inputs.…”

Non–von Neumann multi-input spike signal processing enabled by an artificial synaptic multiplexer

“…Information on the surrounding environment is obtained using sensor networks, and thus, the amount of obtainable information is determined by the number of sensors in the network (4). This implies that the active and preemptive response of AI devices relies on the densification of the sensor network (5)(6)(7)(8)(9)(10)(11), which inevitably causes incremental system complexity (4,(12)(13)(14)(15)(16). In this sense, various strategies using the conventional complementary metaloxide semiconductor (CMOS) devices have been proposed to relieve system complexity upon the densification (16)(17)(18)(19)(20)(21) including a multivariable proportional integration differential (PID)-based control system (22)(23)(24) and a parallel-to-serial converter (25)(26)(27)(28), but they have a fundamental limitations of the serial processing (29,30).…”