Forecasting Food Security with Reservoir Computing

Kurzfassung

Early warning systems are an essential tool for effective humanitarian action. Advance warnings on impending disasters facilitate timely and targeted response which help save lives, livelihoods, and scarce financial resources. We present a quantitative methodology based on Reservoir Computing (RC) to forecast levels of food consumption for 60 consecutive days, at the sub-national level, in four countries: Mali, Nigeria, Syria, and Yemen. The methodology is built on publicly available data from the World Food Programme’s integrated global hunger monitoring system (https://hungermap.wfp.org/). We compare the performance of the RC model to various algorithms including ARIMA, XGBoost, LSTMs and CNNs spanning from classical statistical to deep learning approaches. Our findings highlight Reservoir Computing as a particularly well-suited model for this task given both its notable resistance to over-fitting on limited data samples and its efficient training capabilities. This work constitutes a successful application of RC on high-dimensional, heterogenous, real data and has been submitted to Nature Communications.