Predict Spread of Infectious Diseases Across Regions via Transport Systems in Kenya Using Machine Learning Models: Case of Truck Crews and Bus Passenger Movement

Abstract

The recurrent outbreak of infectious diseases has led to a global health catastrophe with substantial socioeconomic interruptions, morbidity, and deaths. The coronavirus pandemic began in December 2019 in Wuhan, China, and has since transmission throughout the world. It is the most recent outbreak of a viral illness. Understanding and predicting the factors that contribute to COVID-19 spread is critical for providing insights to public health decision-makers in order to inform interventions. The spread of COVID-19 could be propagated by human mobility. People travel for several reasons such as business, education, tours, and work. Buses and matatus are the common modes of transport in Nairobi which contributes to 41.99 % according to KIHBS 2015/2016. The immediate drastic measures taken by most governments is to restrict movements between countries, counties, and towns which has a significant impact on the economy. Developing accurate machine learning models to demonstrate the relationship between the movement of people and the spread of COVID-19 is needed to help come up with measures to reduce the spread and prepare surveillance for future occurrences. Predicting the spread of COVID-19 is currently being done using various machine learning and epidemiological. However, disease modelling based on mobility patterns remains complex due to the siloed data hence the need for innovative methods to enhance prediction accuracy. This research used Logistic Regression and K-Nearest Neighbor machine learning algorithms to forecast the transmission of COVID-19 using modelled data. The training and testing data came from a variety of sources, including truck traffic obtained from NCTTCA reports containing average daily weighted traffic captured at weighbridges. The bus transportation data was modeled using the linear regression trip generation model and parameters from the NUTRANS report (The study on Master Plan for Urban Transport in The Nairobi Metropolitan Area in The Republic of Kenya), with the base year derived from socioeconomic attributes in the 2019 census report. The projections have been made based on the population growth rate derived from the world's development indicators. COVID-19 data was derived from the World Health Organization. Logistic regression demonstrated a testing accuracy score of 0.64 while K-Nearest Neighbor achieved a testing accuracy score of 0.643 using the combined truck and bus daily traffic data. Daily bus data demonstrated an accuracy score of 1 while daily truck data demonstrated an accuracy score of 0.83. We conclude that using modeled data retrieved from various sources, we can forecast the spread of viral illnesses through human mobility patterns.