T-Ridership: A web tool for reprogramming public transportation fleets to minimize COVID-19 transmission

As the outbreak of novel coronavirus disease (COVID-19) continues to spread throughout the world, steps are being taken to limit the impact on public health. In the realm of infectious diseases like COVID-19, social distancing is one of the effective measures to avoid exposure to the virus and reduce its spread. Traveling on public transport can meaningfully facilitate the propagation of the transmission of infectious diseases. Accordingly, responsive actions taken by public transit agencies against risk factors can effectively limit the risk and make transit systems safe. Among the multitude of risk factors that can affect infection spread on public transport, the likelihood of exposure is a major factor that depends on the number of people riding the public transport and can be reduced by socially distanced settings. Considering that many individuals may not act in the socially optimal manner, the necessity of public transit agencies to implement measures and restrictions is vital. In this study, we present a novel web-based application, T-Ridership, based on a hybrid optimized dynamic programming inspired by neural networks algorithm to optimize public transit for safety with respect to COVID-19. Two main steps are taken in the analysis through Metropolitan Transportation Authority (MTA): detecting high-density stations by input data normalization, and then, using these results, the T-Ridership tool automatically determines optimal station order to avoid overcrowded transit vehicles. Effectively our proposed web tool helps public transit to be safe to ride under risk of infections by reducing the density of riders on public transit vehicles as well as trip duration. These results can be used in expanding on and improving policy in public transit, to better plan the scheduled time of trains and buses in a way that prevents high-volume human contact, increases social distance, and reduces the possibility of disease transmission (available at:http://t-ridership.com and GitHub at: https://github.com/Imani-Saba/TRidership).


Introduction
The COVID-19 pandemic is changing the world significantly by affecting all economics, education, and transportation. The transmitting property of COVID-19 between individuals shows the need for immediate attention. After the COVID-19 became an official pandemic, barreling through all countries and the spread is infiltering every aspect of our lives, a response was setup with considerations of facing recurring outbreaks over the next decades [1]. COVID-19 virus can be seen as a wake-up call to improve outbreak responses, and lessons from this pandemic can be used to keep diseases like coronavirus under control before spreading widely. While the world is experiencing an expanding outbreak of COVID-19, the necessity of public transport to guarantee access and continuity of essential services is explicit. Public transportation, which is one of the principal arteries of the cities, also makes them dangerous in a pandemic. It has been proved that human-to-human transmission of COVID-19 infection has occurred among close contacts. To control outbreaks happening anytime and anywhere else, considerable primary efforts to avoid transmission are required. Many research institutes are trying to provide physical and public places for recovery, besides working to defeat the virus by vaccination and identifying places and populations with the highest need.
The highest risk of infectious diseases in public transportation is sitting or standing in high proximity within a closed environment [2], and consequently, among the non-pharmaceutical prevention measures, the deployment of physical distancing has the most significant role in ensuring public transportation safety [3].
As a result of the experience gained after outbreaks like COVID-19, we should now focus more and think of the different ways to define a reliable transit system. As there are various factors that are effective in transmitting infection diseases (eg. Crowded spaces, poor ventilation, lack of social distancing), considering public transit factor will be a big step forward in future plans and decisions to prevent the disease from becoming a pandemic. In the current pandemic situation, an increasing number of agencies have begun to define new policies for transportation systems to keep running them at their maximum efficacy and are taking steps to incorporate social distancing into the transportation planning process.
It is urgent to find concrete strategies to reduce the spread of the virus on public transportation to limit the spread of the disease.
In this study, we aim to explore the risks of riding public transit during a pandemic and suggest measures that can be taken in response. The risk of infection on public transit depends on several factors, including the number of people riding on public transport, air exchange in the system, and hygiene practices. Socially distanced public transit settings that are under the control of transit operators can minimize the risk of infection spread.
This study reviews methodologies for public transportation decision making, their performance, and demonstrates a framework for incorporating social distancing considerations in transportation planning. A decision on effective transportation schedule for the present is made through statistical analysis of annually collected data of previous years. This would make the public transit system safe to travel for the people during any infectious diseases. In the first part of our study, we are aiming to understand how to measure the performance of a transit line density. Through a case study using data from the Metropolitan Transportation Authority (MTA), transforming them into a form suitable for analysis and then creating a map of the chosen MTA line stations, the study evaluates competing for our new method for reprogramming public transit and decision making based on previously collected data records using novel normalizing T-Ridership web tool analyses.
In the next part, a decision support web tool, namely T-Ridership, is proposed to analyze and visualize the data and the evaluation to effectively reflect density changes. The proposed framework should help decision-makers to incorporating social distancing considerations into transportation planning.
The novelty of this approach is expressed in the study and assessment of values as transit principles and benchmarks that guide public transport agencies in their decision-making process when planning to optimize public transit for safety against contiguous disease to avoid mass density applied to the public transit vehicles automatically.
The policy for physical distancing investigated in this paper requires transport operators to reduce the number of passengers per vehicle, which our proposed web tool, T-Ridership, can be used to help public transportation operations plan better time scheduled of trains and buses to minimize the likelihood of exposure by decreasing the proximity and density of crowded transit vehicles. Besides socially distanced public transit settings, our proposed approach considerably optimizes the duration of exposure by reducing the duration of time spent in a particular setting. This research is suggesting a profound intervention in the way cities approach their transport policy, even after the virus subsides, to play a highly significant role in unpredictable future outbreaks and promotes the low risk of spread of the virus.

Literature review
Several studies have detected the transmission of COVID-19 is through contact, droplet, airborne routes, and even studies generate that a considerable rate of people are infected by COVID-19 respiratory infection through tidal breathing which is majorly transferred using various forms of public transportation that inevitably puts them in relatively close contact with high proximity and density of potential carriers of the virus. Many transit agencies have established frequent sanitation. However, evidence suggests that airborne transmission of the virus poses a greater danger than other modes of transmission which indicates the efficiency of social distancing and relatively density control in crowded transit vehicles in preventing virus dissemination in exhaled breath. These causative agents, explaining the large community outbreak of COVID-19, emerge the public transit agencies the necessity of pandemic response measures including physical distancing, and operations to avoid crowded transit for prevention of infection [4][5][6][7][8].
Different modes of public transportation are used daily by millions of people, and they often carry passengers above their capacity, especially in peak hours. This might contribute to the spread of infectious diseases among public transport users. Researchers found an association between respiratory infection in winter and train use, emphasizing that the use of public transport in the winter potentially exposes travelers to respiratory viruses [9]. Studies on estimating the relative risk reduction of domestic dissemination of COVID-19 in Japan due to travel restrictions on the public transportation network indicate that milder travel restrictions could have a similar impact instead of a lockdown that might seriously damage the economy on controlling the domestic transmission of COVID-19 [10]. Travel behavior and mode choice preferences have been transformed significantly compared to the pre-pandemic period, even after controlling the COVID-19 burden by vaccination. In order to avoid the collapse of the public transport system during a pandemic and even after, it is necessary to keep public transit running through optimal infrastructure planning and raise safety against infectious disease transmission to create incentives for citizens to prefer public transportation to individual transport. Among the recommendations provided by the Government, the results obtained from several pieces of research showed that reducing the level of proximate contact between people and decreasing the duration of time spent in a particular setting are highly favorable preventive measures to encourage continued use of public transit [11].
Many studies explored people's intentions to use public transport during the COVID-19 pandemic while adhering to precautionary measures by collecting responses through questionnaire surveys and the results of studies showed high intentions of users to continue using public transport during the COVID-19 pandemic, which provides insights for policymakers and public transport operators about further necessary actions to promote safe public transport use during the current and any possible future pandemics [12].
During the COVID-19 outbreak, epidemiologists encouraged social distancing and spending less time in an enclosed place with the crowd [3]. The greatest risk for infectious diseases in public transportation is the proximity and density of potential carriers of the virus in a closed environment [2].
Scientifically suitable approaches that may lead to strong policy decisions on the transportation systems in terms of the COVID-19 pandemic are deficient. Hence, even after the lockdown is lifted across countries, restrictions and passenger safety in public transport should be prioritized as well as optimizing the number of passengers in public transit vehicles. Strict traffic control should be introduced everywhere to avoid vehicle jams and mass gatherings [13].
Preventive measures, such as social distancing restrictions in mass urban transportation and public areas, need to be implemented effectively by government agencies and other responsible institutes. These restrictions can play a highly significant role in unpredictable future outbreaks (epidemic and pandemic disease) to slow down the virus spread [14].
Based on this research study, there is evidence that humanto-human transmission has occurred among close contacts. To control outbreaks happening anytime and anywhere else, considerable primary efforts to avoid transmission are required. Procedures to prevent or reduce transmission should be implemented in populations at risk, and public places, especially public transport that is the lifeblood of our cities and should continue running as long as possible to transport essential workers. If we have to travel using public transportation, the preference is to use the least-crowded bus or subway car possible. To an extent, a public health crisis like the coronavirus outbreak, is an urban planning problem, as public spaces impact how we interact with one another, and how a disease like COVID-19 can spread [15]. Urban places' responses to the pandemic are revealing that mandated distancing, physical, and social structures of communities significantly influence places' ability to cope with the immediate crisis.
The qualities that make urban places vulnerable during COVID-19 are the keys to containing the disease's spread. Public transportation, which is one of the principal arteries of the cities, also makes them dangerous in a pandemic.
Public health research indicates that the risk of contagion infection for public transit riders depends on a variety of factors, including the characteristics of the contagion in terms of its means, ease of transmission, and disease prevalence in the community, the possibility of exposure to the contagion, duration of exposure to the virus, passenger densities and occupancy levels in public transit vehicles and the public health interventions implemented in public transit [16,17]. As the virus continues to spread, the physical density and design of public places will continue to influence how well they overcome the pandemic. For many communities and populations, the impacts of COVID-19 are likely to be dire, but efforts that nurture and support workers will be vital to mitigating the fallout. Crowded and high-density public transportation systems furthered the spread of the virus. Despite serious downsides, the quality of densely connected places like public transportation can also make a life amid social distancing more livable, and this is causing any public transit agencies across the world to seek to ramp up plans for contagious virus response.
Virus infections play an important role in the public transit fleet. Analyzing subway entries as measured by turnstile entry records [18] and the number of new cases reported by the New York City Department of Health [19] shows that newly reported cases finally decreased around March 16th, 2021, as subway use was cratering. It is expanding the idea of the relation between the two trends of turnstile entries and coronavirus incidence [20]. Also, in the simulations shown in other recent research, controlling virus transmission in the public transportation system reduces the speed of virus spread and the height of the peak by approximately 20%. Evidently, this does not indicate a reaction of reducing public transport capacities but more willingly use it to plan lower densities of public transportation and thus reduced infection rates. In addition, at home or in private vehicles, one interacts with the same people repeatedly, effectively decreasing the infection probability. However, public transport is much different: One sits or stands next to other people every day. In consequence, public transportation is increasing possible infection paths between different people every day, especially while squeezed together. Since they run back into their respective homes, public transport results in strong mixing [21].
However, public transportation is highly vulnerable to facing outbreaks such as COVID-19, it is essential to continue running to move needed workers, as well as considering controlling the spread of the virus in public transport is a big step in confronting the disease [22]. The mass gathering of people in public places is a significant concern in the spread of the coronavirus. This has led to restrictions on public transportation [23]. An overview of the speed of the coronavirus spread in the US suggests that in cities where public transportation is more commonly used, such as New York and other big, crowded cities, the outbreak is higher compared to cities where public transportation is used less [20].
COVID-19 spreads massively among people who are in close contact for a prolonged period. There are many guidelines to follow for practicing and considering social distancing options to travel safely when there are demands for running errands or commuting to and from work using public transit-for example, when you are selecting seats on a bus or train to keep at least 6 ft distance. But many people have personal circumstances or situations that present challenges with practicing social distancing to prevent the spread of COVID-19, so there should be more strict rules that automatically apply social distancing in public places such as public transportation, apart from people's personal preferences.

Hybrid optimized dynamic programming
In this paper, we designed a hybrid structure based on the inspiration from neural network (NN) algorithms. Our idea is motivated by the design of the neural network structures to make our network structure easy to analyze public transportation trips. We propose a hybrid model that may lead to an optimized algorithm of neural network structure (Fig. 1).
To make our network structure easy to generalize to other datasets, our methodology separates the structure design and weights search. Taking inspiration from the neural network, that has become a powerful and most successful tool in machine learning, our presented optimized training algorithm evolves the performance of the fully connected model of neural network which just works with some prepared functions. This flow chart outlines the methodological steps taken in the study, including data collection, tool development, implementation, and data analysis. It also includes a case study demonstrating the effectiveness of the T-Ridership tool in reducing passenger density in public transit. The flowchart illustrates the flow of information and data between these steps.
The presented study embraced the neural network method along with dynamic programming to analyze public transportation trip flows and minimize Coronavirus exposure by using public transportation.
Neural network is one of the emerging technologies, which interprets novelty detection in transportation and urban planning area. Among all different methodologies for modeling transportation parameters and available data analysis research tools, a neural network has been suggested as the most successful alternative. Neural network, an extremely popular class of Constructive interference (CI) models, has been widely applied to evaluate various transportation issues [24][25][26][27][28][29][30][31][32].
In standard feed-forward neural network modeling that is applied to this paper, there is one set of connections fully connecting the input layer to the next layers, resulting in the activations of the nodes until getting the output. In this procedure, neurons are the basic common features in neural network composition, Multilayer perceptron hybrid dynamic model. A schematic representation of the process involved that each station is considered as a neuron and connected to the next layer with its own weight, which is the number of passengers between two stations. After calculation, If the number exceeds the Max, then the recursion function is used to go back and repeat itself, to finally program 2 lines with allowed capacity consideration in public transit vehicles.
which take data in and join to the next transmission process by weighted connections to compute an output [33].
The defined method in this paper is inspired by the neural network to implement programming and suggest a hybrid system using dynamic programming which lets the errors in analysis correct themselves through a recursive function and enables the function to repeat itself several times, to eventually output the result [34].
We used the supervised machine learning model to validate the process of evaluating a trained model on a test MTA dataset. This method provides the generalization capability of a trained model. The input data on T-Ridership clean and straightforward to process and can easily interpret data features. Therefore, data is not subjected to pre-processing. Based on the maximum amount of data available in MTA, we have considered a series of restrictions for entering initial data in T-Ridership by the user. So, we do not encounter extreme values during the process. We split the data into training and test data sets. The main idea of splitting the dataset into a validation set is to prevent our model from overfitting. The T-Ridership is processed on a small dataset. Therefore, we used k-fold cross-validation with an independent test set. As input data, we have the number of passengers entering stations, and we need to find how many passengers are in a vehicle between two stations. The T-Ridership measures the number of allowed passengers inside the vehicle by k-fold crossvalidation. The average values are computed in the loop from the first station to the vehicle's current station. T-Ridership looks for the best score for each station (node) to optimize two runs for one line. Since this is a classification problem, we defined an accuracy score algorithm for getting optimized.
Since we are faced with a classification problem, T-Ridership uses the Logistic Regression model. T-Ridership is looking to discover new runs based on the maximum capacity of a station on the line (whether the vehicle has to stop in the current station or skip it); as a result, we used this technique in our algorithm. According to the Tridership method to optimize a line, the cut-off point is the fixed value the user should input in the initial step.
The study is carried out under the motivation of knowing that modeling daily trips for traffic flow prediction will optimize the issues affecting mass density in public transit modes.
Moreover, the suggested hybrid neural network approach and dynamic recursion function improved the proposed models for public transit planning, and different data analysis indicates more robust models compared to other transportation programming systems.
The model is calibrated based on the passenger's origin, destination, and population density in each station. This reduces passengers' indoor wait time, congestion, population density, and person-to-person contact. This approach is willing to find a solution to reduce the number of passengers in the same direction with different destinations during peak times. The Neural network takes the processing elements from data resources using a transfer function joined by weights connections; data flow along with these connections and wherever it finds the statistics exceeding the intended density, a recursion function through dynamic programming lets the errors correct themselves from the beginning, figures out the most populated stations and skips the stations with high density and transfer it to an excessive run to divide passengers in a common route into different trains, and repeats the procedure to finally get to the output with the least skipped stations. Using an algorithm to achieve the best travel estimation for less population traffic is the basis of this study.
As shown in Fig. 2, in the neural network model, the neuron is the basic component that is in place of stations, and a multilayer perceptron structure is a fully interconnected set of layers of neurons. Each neuron (N1· · ·n) of a layer is connected to each neuron of the next layer, from the input layer to the output layer. Fig. 2 is a schematic representation of the process involved.

Implemented T-Ridership web tool
The idea and the created algorithm are followed by an implemented web tool that is freely available to be used online. There are no similar web tools compared to our designated one, which can automatically normalize the density in public transportation vehicles with social distancing consideration, which has become one of the top priorities in the recent outbreak of COVID-19. Our web tool is flexible to apply to any kind of public transit such as taxi, bus, or train.

T-Ridership
Identifying components of the public transit system most likely to exacerbate disease spread is critical for public health authorities and public transit agencies to be able to plan considering epidemics or pandemics and control their spread. In this work, we propose a method to help the transit network use a free web tool to find the timetable of public transportation system to avoid mass density especially in peak time and crowded routes.
The algorithm of the T-Ridership web tool works through an iterative process of calculating transit data. The T-Ridership is implemented in C# (Microsoft Visual Studio 2019), and also several other languages were used.
T-Ridership is a free web tool for reprogramming the public transportation fleet to reduce COVID-19 infection transmission by social distancing consideration. While coding our database, we use the neural network algorithm. Each station is considered a neuron and connected to the next layer with its own weight, which is the number of passengers between two stations. By considering the capacity of each bus or each train car with the social distancing application, a maximum number is fixed, and the system automatically skips the stations that exceed the fixed number.
The designated T-Ridership web tool works with two methods; the first method calculates the number of passengers on a route; when we just have passenger statistics in each station without knowing their origin and destination station. In other words, when the only available data is the number of passengers entering and exiting each station, without knowing each passenger's specific origin and destination and how many stations are traveled by each passenger.
In our second method which is designed to be used by knowing the number of passengers with their specific origin and destination by creating an n × n matrix that n is the number of stations in one route, and contains our input data consisting of the columns, showing the number of passengers exit in different stations from a specific origin, and the rows showing the number of exit passengers from diverse origins but in a particular destination. The last column of the matrix indicates the number of passengers transiting between two continuous stations.
The significance of this method is that besides having its complexity for collecting input data and calculations, it is highly accurate. Method A determines the number of passengers on a route based on the entrances and exits at each station, without taking into account the number of stations traveled by each passenger. Conversely, Method B calculates the number of passengers on a route based on the entrances and exits at each station, while taking into account the number of stations traveled by each individual passenger (Fig. 3).
As the output of the designated web tools, the result is shown as text and graphical illustration, featuring all stations divided into different runs with their stops and skipped stations (Fig. 4).
In the first step, the main aim is to find the real number of passengers traveling inside a public transportation vehicle on a route, which is essential in our research as it is affecting virus transmission. The first step capacity calculation is based on real data extracted from available turnstile open data. Users can input any time limit data in the web tool and get the result.
A public transit route has n stations. The number of passengers who are expected to get on and off at the ith station during a given time period is known, denoted as x i and y i respectively.
At each station i, the total number of passengers leaving for the next station is denoted as P i Then suppose social distancing policy allows no more than S passengers in the vehicle at any given time, and a feed-forward process is done: (1) Ridership feed-forward parameters (n, x i , y i , P i , S) (2) for i = 1 to n (3) do (4) if (P i > S) (5) skip station i //in first train P i+1 = Pi -y i -x i //subtract the number of passengers exiting and entering the station i from the next station (7) end loop when reaching the last station (8) return skipped stations.
With this algorithm run 1 is set with its stops and skipped stations.
Then for the next run, we use the following algorithm: (1) Recursion algorithm to find next train stops (2) given: skipped stations from first train m [1..j] (3) for i = m to 1 //until the first station or the previous skipped station (4) do (5) find the biggest P i and skip it //for the current train (6) call feed-forward //to check the maximum capacity (7) return skipped stations (see Fig. 5)

Input data set
As the basis of input data, automatic passenger counting can be done by contact-type counters, optical sensors, vision-based systems, and Integrated Circuit (IC) cards. All these counters can be applied in many public places with entrances, such as subways and bus stations. Subway passenger flow has obvious time-varying characteristics with easily identifiable peaks [35].
We are using the open transit data Toolkit [18] that is helping to utilize transit data. Through the analysis of IC card data, we can understand the time distribution characteristics of bus passenger flow, assist in planning the frequency of shifts, and improve the operating efficiency of the entire transit system.

Case study dataset
In this section, we present the dataset used for our study, as well as the pre-processing step conducted to make statistical analysis of the data obtained from the Metropolitan Transportation Authority (MTA) that is a public benefit corporation responsible for public transportation in the U.S. state of New York [35], which has the most available and updated data in the State of New York and through the US. As the first step, we used the Open Transit Data Toolkit for measuring and analyzing the transit systems to be used in the next step, which is programming the prepared data and visualizing transit data. These data are not clear enough to use directly in T-Ridership web tools. At first, we did the clarifying to be used for the next step. After examining the peak hour and density of each station, we have mapped each station's density. However, there were different days and months' data that had been examined and had the same result; we looked into using data for the week of Feb 19, 2022, to illustrate the results in charts. We captured data for the Q Line in New York City Subway in Brooklyn, entering the turnstile at the Prospect Park stop, located within the 40.66161 latitudes and −73.96225 longitude, to the station at the Ocean Pkwy stop at the end of the line, from Feb 19th to Feb 25th, 2022, knowing that missing or incomplete data is a frequent challenge in data wrangling. Fig. 6 superimposes the stops that tens of thousands of passengers take every day back and forth between stations in downtown Brooklyn.

Result of selected line stations' density
As shown in Fig. 7, each station in line Q consisting of 15 stations from Feb 19, 2022, to Feb 25, 2022, is analyzed individually. As a result, obviously, neighborhoods with more population and We considered two ways to find route updates in our algorithm. First, when users have access to detailed data including each passenger's origin and destination, and second when users just have access to entering and exiting passengers of each station. Based on what kind of data is available we go through this algorithm to gain the first run, and relatively re-reading the data for having the second run. density had more crowded stations. As shown in Fig. 6, the King's Highway and Church Av had the most passengers and density.
Policy determiners can use the result of the current study to make decisions and change the public transit ordination to avoid overcrowded stations and trains. By detection of the T-Ridership web tool, identified stations with high density should change their scheduled timetable and frequency.

Results of T-Ridership
We use statistics and information of ridership to illustrate the lines' density, and by comparing the target lines, we figure out the most ridership lines. Therefore, there is a high-resolution human contact in those routes that play a fundamental role in infectious disease transmission.
The data resulting in the previous section can be performed in the measurement that trains transfer within a defined period of time. From the result, the stations that each train should stop at its scheduled time or the stations that needs to be skipped can be determined.
By the location coordinates of the stations, the following chart shows our sample of the line in the Brooklyn neighborhood in the New York subway. As explained before in this section, we chose a sample with real data for Feb 2022 containing 15 sequence stations, while evaluating their entry and exit for one week, and as a result, two stations were recognized as the most crowded in comparison with others as shown in Fig. 7. We then calculated the average of our initial results and normalized the data to input it into the T-Ridership web tool, which demonstrated the best performance in reducing the number of passengers in overcrowded stations. Fig. 8 illustrates the effectiveness of the T-Ridership web tool in decreasing the number of passengers on a route by dividing them into two different transit timetables.

Conclusions
This study aimed to investigate the association between traveling on public transport and the risk of spreading infectious diseases and what public transit agencies can and should do in response to an outbreak of infectious diseases like COVID-19. We conducted searches on Metropolitan Transportation Authority (MTA) database in order to include real data from transportation to show the efficiency of our proposed method on minimizing the number/proportion of people riding on public transit, especially during peak hours, to ensure promoting physical distance protocols and avoid riding on high-density public transit vehicles. If we apply socially distanced public transit settings, it can be used as a preventive means of transmission of any future high-risk infectious disease like COVID-19 before becoming a worldwide outbreak.
When COVID-19 upended travel across the world, governments issued directives on physical distancing and the use of face masks to curb community spread of the disease. Given that public transport has been identified as a high-risk environment for transmission, it was of interest to ascertain user and operator compliance to guidelines for public transport operations during this period to measure its risk level. The policy on the use of face masks within vehicles is adhered to; however, the policy on the physical distancing in vehicles requires stricter enforcement. The policy for physical distancing requires transport operators to reduce the number of passengers per vehicle.
In this paper, we use the inspiration from neural network to design an optimized hybrid dynamic programming that can capture sequence analysis tasks. The effectiveness of the proposed methodology was evaluated using a real dataset issued from the public transport network of New York in the US. We also investigated the effect of this information on the transportation system on the social distancing accuracy. The results obtained from the T-Ridership web tool that used both subway and bus data, showed significant improvement on all the data set prediction performances. All investigations have been carried out to enhance and consolidate these results. Particular attention has been given to reducing human-to-human contact. Our simulation system and T-Ridership web tool, with implications particularly for the postlockdown phase, are to minimize the contact of passengers in indoor places like buses and trains by preventing overcrowding that results in reducing the spread of coronavirus outbreak.
Thus, allowing for better optimization of public transportation setup, this methodology can be extended to the whole public transport network, and policy determiners can use the result of the current study to make decisions and change the public transit ordination to avoid over-crowded stations and trains. By detection of the T-Ridership web tool, identified stations with high density should change their timetable scheduled time and frequency ( Supplementary Fig. S2).
The main limitation of this research was the capacity of adding vehicles to the public transportation system, due to limited transportation earmarked funding. Although adding more vehicles on high-density routes could help the better performance of our work, it is not practically possible. Based on this, our research's main idea is not to add any vehicles but to change the frequency and timetabling of trains in a route. In this method, the only problem is that the waiting time for some limited passengers in skipped stations increases; however, the other passengers will wait less for the train and also spend less time inside the vehicle. Despite the increase in waiting time in some stations, this is still beneficial in the outbreak to avoid shutting down public transportation entirely.
If we can solve the funding problem, we intend to add a maximum of one vehicle to optimize the increase in waiting time. As the idea generates revenue, the decision makers are motivated to apply the web tool to the system and reap the benefits.

CRediT authorship contribution statement
Saba Imani: Analyzed the data, Writing -original draft, Supervision. Majid Vahed: Participated in designing and coordinating the study. Shreya Satodia: Participated in designing and coordinating the study. Mohammad Vahed: Initiated and implemented the software.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.