Urban Covid-19 Analysis and Prediction Based on Machine Learning and LSTM Models

Abstract:
To find out the urban factors affecting the development of the epidemic and to predict the epidemic more accurately and quickly, a series of experiments were conducted by machine learning model and Long Shot-Term Memory (LSTM) model. The most important features were first identified by the k-means clustering model and then the "epidemic development indicator" features were determined by the principal component analysis (PCA) model. The experiments show that this "epidemic development indicator" feature can reduce the amount of data while successfully classifying epidemic devel-opment in cities. To analyze the contribution of city data to epidemic development, a decision tree method was used to analyze and visualize the results. The results showed that the most important factor was population size, followed by population density and latitude, and the least influential factor was the annual temperature. To predict and respond to the epidemic more quickly and accurately, a Term Memory (LSTM) model was introduced and a low error was achieved.