Limited Data Forecasting for Dengue Propagation

Dengue menace, which has been present in Sri Lanka since the early 1960s, has become a major public health concern. Climatic factors such as average rainfall, minimum temperature, and relative humidity greatly influence the epidemiological pattern of Dengue. The objective of this study is to develop an effective model to forecast future Dengue propagation in Colombo district in Sri Lanka, with the intention of raising awareness amongst the authorities and the public to take necessary steps to minimize the spread. In this paper, Long Short-Term Memory based few-shot learning (LSTM-FSL) model which makes use of limited data in the target task is proposed. The proposed model is compared with the Naive LSTM model and Vector Autoregression (VAR) model in terms of Mean Absolute Deviation (MAD) and Root Mean Square Error (RMSE). The LSTM-FSL model performs with an RMSE of 108.62 and MAD of 97.64, outperforming the other compared models.

I. Introduction

Dengue is a mosquito-borne viral infection that has soared over the last few decades in Sri Lanka. The deadly nature of the disease has posed a massive challenge to national health authorities for several decades. Hence, the government needs to allocate a considerable amount from the annual budget for the preventive and curative measures of this disease.