Forecasting COVID-19 Confirmed Cases in Major Indian Cities and Their Connectedness with Mobility and Weather-related Parameters

Abstract

Coronavirus disease 2019 (COVID-19) outbreak that was declared as a pandemic by the World Health Organization (WHO) on 11 March 2020 has already had severe consequences in all aspects of people’s lives worldwide. The pandemic has affected over 200 countries and has become a major concern. India also faced a stiff challenge in terms of controlling the virus outbreak and through some strict measures such as nationwide lockdown was able to control the further spread of COVID-19 towards the latter part of 2020. Therefore, it is imperative to predict the spread of this virus along with causality analysis of parameters that play a significant role in its spread. The present study employs a series of univariate and multivariate time series forecasting techniques namely MSARIMA, ARMAX and extended VAR models to predict COVID-19 cases in New Delhi, Mumbai and Bengaluru. Besides, providing a robust forecasting performance for COVID-19 cases, the study also deals with finding the causal relationship of the spread of COVID-19 with various mobility and weather parameters. Outcomes of our study establish that the spread of COVID-19 can be associated with mobility and weather parameters apart from the various precautions that are taken by the people to reduce community transmission. However, the type of mobility (residential, retail and workplace) and type of weather conditions (air quality, temperature and humidity) associated with the causality differ with cities. For New Delhi, air quality, residential, retail are the parameters affecting the spread of the COVID-19 cases, whereas masks, temperature, residential and workplace were the significant mobility and weather parameters for Mumbai. In addition, for Bengaluru, the statistically significant causal variables were air quality, masks and residential. Outcomes of this study would help the concerned authorities to predict and contain future COVID-19 spreads in Indian cities efficiently.

Vector Autoregression for Forecasting the Number of COVID-19 Cases and Analyzing Behavioral Indicators in the Philippines: Ecologic Time-Trend Study

“…In epidemiology, public interest on COVID-19 can be a marker of ongoing but undetected infection over a specific period and location, or it can be a proxy measure for the population’s risk perception and intention to practice self-protection measures. Krishna [ 21 ] and Milani [ 23 ] both demonstrated the possibility of using internet search volume for words related to protective measures as signs of precautions taken by the population over a period of time. On the other hand, Walker and Sulyok [ 33 ], Lin et al [ 34 ], and Higgins et al [ 35 ] have shown the correlation between web-based traffic search trend and the number of reported cases and highlighted the potential of “infodemiology” for surveillance.…”

“…Daily number of cases, change in mobility, and public interest on the risk of contracting COVID-19 were treated as endogenous variables since previous literature suggests the presence of a bidirectional relationship among these 3 variables [ 21 , 23 ]. These variables were also chosen to demonstrate the use of publicly available data for surveillance.…”

“…As the governments design their respective lockdown exit strategy, it is crucial to maintain the awareness of the population about the persistence of the threat due to the virus and sustain its interest in disease prevention and control measures [40]. The importance of providing attention to public interest was supported by the change in statistical significance of public interest for the prediction of the number of COVID-19 cases over time and the weaker performance of model 1 as public interest becomes important in predicting the number of cases for September [15][16][17][18][19][20][21][22]2021, and January 28 to February 4, 2022. Friedman and Kuttner [41] recommended the assessment of the innovation accounting of variables once Granger causality hints a change in the dynamics of variable.…”

“…Scholars have used VAR to describe the interaction of a variety of variables with the spread of SARS-CoV-2 using real-world data. A study conducted by Krishna [ 21 ] showed that population density improved the prediction of the number of COVID-19 cases in all major cities of India and suggested the possible variation in the environmental characteristics of each city as the relationship of temperature, air quality, and humidity with COVID-19 cases was found to be inconsistent. The interconnectedness of countries in terms of cross-border transmission was also exhibited in the works of Fitriani et al [ 22 ] and Milani [ 23 ].…”