基于BiLSTM模型的新冠肺炎预测研究

A Study on the Prediction of COVID-19 Based on the BiLSTM Model

摘要: 新型冠状肺炎疫情的持续蔓延给人类社会带来了深远影响，时至今日，疫情仍在世界范围内传播，新冠疫情发展趋势预测是一大研究焦点。传统传染病模型基于一系列数学假设进行建模预测，未考虑人口流动与反馈机制的影响，很难对新冠疫情传播趋势进行可靠的预测；统计与机器学习模型单纯依据已有数据进行预测，难以有效提高预测精度。本文基于深度学习方法，构建多特征下的双向长短期记忆网络(BiLSTM)模型针对不同场景下的国家与地区的新冠疫情的发展趋势进行预测。选取决定系数R2与平均绝对百分比误差(MAPE)作为评价模型的指标，并与传统长短期记忆网络(LSTM)模型进行比较分析，实验结果表明BiLSTM模型具有更好的预测性能和实用性。

Abstract: The continuous spread of coronary pneumonia in COVID-19 has brought far-reaching influence on human society; Up to now, the epidemic is still spreading all over the world. The prediction of the development trend of the COVID-19 epidemic is a major research focus. Traditional predicative models of infectious disease, based on a series of mathematical assumptions, failed to take into the account of population mobilization and feedback mechanism, and thus it is difficult to reliably pre-dict the spread of COVID-19; statistical and machine learning models make predictions based solely on existing data, making it difficult to effectively improve prediction accuracy. This paper aims to conduct a prediction targeted on the developing trend of COVID-19 pandemic in different scenarios of countries and regions based on the bidirectional long and short term memory network (BiLSTM) model with multiple features under deep learning. Decision Coefficient R2 and Mean Absolute Per-centage Error (MAPE) were selected as the indexes of the evaluation model, and compared with the traditional Long Short-Term Memory network (LSTM) model, experimental results show that the BiLSTM model has better prediction performance and practicability.