 Well, thank you for being until the last presentation of the day. I'm going to talk a little bit on how we use data to analyze mobility in our cities. And OK. OK, sorry. So first, why we want to try to understand mobility and transportation? Because we think it's important because people need to access urban opportunities. And basically, when we are talking about urban opportunities, we're talking about jobs, health, education, recreations, and everything we have to do when we live in a city. So the question of how we do that. So basically, we have two options. The first one is when we have data, we can analyze the data. And there are a lot of opportunities to analyze mobility using data, especially with the cell phone data, with new sensors that are located in vehicles, GPS systems, et cetera. But the problem is sometimes it's not easy to get the data. So we have some other options when we use simulations. So I'm going to talk about a couple of projects where in the first project, I was analyzing actual data. And in other projects, when we don't have the data, we use simulations. So basically, this is a project that we're working with the Inter-American Development Bank. This is something we've been doing with another colleague, that is Felipe Gonzalez. And this is an open source package in Python that can be used to analyze transit data, basically the transactions that are created every time you go into the public transportation systems. Basically, these type of cards, we have the SUBE card, which is the card that you can use to pay for buses, metro system, or train systems. And you have all around the world these type of systems. Let me tell you something a little bit about the data. Basically, on the basic level, it could be a little bit more complex, but on the basic level, you have a data set with transactions. Basically, each car has an ID, a unique ID. And every time you pay for the ticket, you get a transaction. It has the vehicle number. It has a data field, the mode of transportation. And you have the type of ticket that you pay. Sometimes you have different prices, for example, when you are students. And then each vehicle has another data set that is a GPS data that basically each of these machines records the location, they launch it and let it do 30 seconds, 32 minutes, or something in between. It depends. So in the basic level, this is the data that you, they minimum require data that you need in order to use this package. The code is published in the Inter-American Development Bank Code for Development website. And of course, it's available for any city that has the data and wants to use it. So there is a lot of literature of how to work with this data. Of course, the main problem, as you might know, is that you can combine this GPS data and this ticket data. And you have an idea of where the trip started, but you don't know anything on where the trip ended. So the literature discuss this. And for many years, we've been working on how we can basically recreate or infer what we call the trip chain, basically, is what every card or every person that every user does during the whole day, basically from the beginning of the day when you go out of your house and enter into the system and then do whatever you have to do within the city. And basically, the way to infer this is you use the origin of the next transaction and you do a set of validations in order to know if this could be possible. And for example, if you take a bus line at the beginning of the day to go to downtown and take the second trip of the day, you take another bus. You have to check whether there is a bus station of the same bus number that you took in the beginning. And once that's validated, you can infer or you can assume that the trip is possible. Then another thing you have to do is trying to understand whether the trip is a trip or is a leg of a trip, or basically if you are doing a transfer. So for doing this, we use a window of time using the timestamp. So in these particular cases, you can see you have three transactions, but it's only one trip. So basically, the origin of the trip is the latitude and longitude of the first trip. And the destination of the first trip is the latitude and longitude of the fourth transaction. And of course, we assume the last trip of the day go back to the first trip, which is presumably home. I'm not going to discuss the assumptions because it's going to take too long, discuss in the literature, and many things already very interesting. So once we have this data set that is processed, we can create these origin destination matrices. And we can visualize them, something you can do also. You can provide a shape with different type of songs. And the system is going to take these songs and it's going to create these matrices based on these songs. If you don't provide anything, the system is going to propose based on data as sonification that you can use. You can also provide as many sonifications as you want. And this is also very technical, but this is the way we are using these matrices, is a way you can plan for better public transit systems. It also creates different indicators that are commonly used to understand transport. This is like the distribution of distances. Basically, we infer the trip chain, and we know the origin of the destination of each trip. And we are using open stream maps to calculate the distances of each trip using the open stream API. And once you have that, you can start doing some analysis. And I'm going to show a few examples of this analysis. This is an analysis where we are looking at the train line that go into the south of Buenos Aires, the Roca train line. So we are looking at the trips that originated or ending in the train stations. And we are trying to analyze the short trips. Basically, in order, this has been done for the Inter-American Development Bank, and they were doing a project, and they wanted to invest in micro-mobility infrastructure. So basically, the idea of this project is trying to understand the route where you have potential for micro-mobility users. So basically, we assume that if you have a lot of short trips, and trips that are shorter than five kilometers, there is an opportunity for improving this type of infrastructure. So this is also published at the Inter-American Development Bank website. If you're interested about how it was done. Then the package also provides some indicators about the supply of transport, basically about the routes of the buses. So you can also provide a shape with the route geometry. But if you don't have it, the system using the GPS data is going to create a line as close as possible. Of course, it's not going to be perfect. And you can analyze the demand in each part of the route. And this is another work we did with this package. Basically, we wanted to analyze how people from informal settlements move, and how is the difference between the travel patterns, between the low-wing populations, and the rest. So basically, we're using the first trip of the day, which is presumably, as I said before, the closest to their homes. We identify the trips that started in a radius of the informal settlements. And we create the origin destination matrices and these graphs that help us analyze how this population moves. And we can get some of the main conclusions we infer from these studies. The average distances are higher for low socioeconomic levels. Trips of low socioeconomic level uses require more transfers than the trips from higher economic levels. Most of the trips from lower socioeconomic levels are in buses. They preferably use metro, for example. And so basically, that's the conclusions that we have. And also, if you want to take a look at the work, it's in the Inter-American Development Bank website. So this is a second work. Also, I've been doing this with the Latin American Development Bank. This is CAF. And in this particular case, it's also a Python package that is called, I mean, the other one was called Urban Trips. This is called, it's not called Urban Python, it's PyOMU. I'm sorry, I think I made a mistake here. Because it's the Mobility Observatory that is a joint initiative between the Latin American Development Bank and the Inter-American Development Bank. And basically, what you need to, we are working with census data at the radio and at the census track level. And we are using open stream maps to identify, basically, the census data help us analyze where people live at the census track level, how the population is distributed. Then we are using open stream maps in order to get the equipments within the city. Basically, from restaurants, public buildings, you name it. We use a clustering technique in order to identify areas of high density of activity, which, basically, we assume those are where people want to go. Because there is high density activities there. So we have our origins, we have our destinations. And once we have that, we can simulate trips and in order to simulate trips, we are using the open stream maps APA. And we are using the Google Maps APA. And we get a lot of information about the simulated trips. Like, basically, we can simulate distances, different auto transportation. We have traffic prediction with the Google API. We know transfer locations. We even have, for some cities, we have estimated time that you have to wait in the stations and so many things. We do this work for 30 Latin American cities. And some of the things you can get is what we call the isochronous of travel times. This is an example for Mexico City where you can compare how the travel times using public transit and using cars. And also, the frequency distribution of the same travel times with cars and with public transit. This is a simulation of the whole day. So you know how velocities and travel times varies along the day. We have here, like, a Wednesday and then we have Saturday and Sunday. And then having information from the census, we infer as a socioeconomic level at the census track level. So we can compare how people from different socioeconomic level move. And also, you can do some analysis for a particular equipment. For example, you can get health facilities and nice travel times to health facilities. And you can identify areas where you have good accessibility to this particular equipment and the areas where you don't. And you can do a particular set of policies or decide in which type of equipment you need in different parts of the city. And of course, you can do some differences in travel time for each socioeconomic level. And well, thank you very much. Thank you for that amazing talk. We do have time for some questions. Hi, thank you. It's somewhat a common practice here that someone will ask you to pay for the trip with your car and give you money. How often do you see that? Or how do you try to correct for that if it's even a problem for you? OK, when we basically, when you have, in some cases, we have in the same cart a trip that starts in the same place. So basically, we split. And we basically create a new ID number. So if I go with my kids to school and then I come back with my kid, I'm going to have two different trips that are exactly the same. And they are counted. In many cases, I can get the first one and the second one is lost. We have some percentage of the cases are lost and we cannot do anything. For example, if I do one trip during the day, you cannot do anything. That trip is lost. We get to recover about 70% of the trip change. But yes, in the case you are doing, I mean, if I go with my kids to school and then I go to work, and then someone else is going to pick my kid from the school, but the second trip is lost. I'm going to analyze the first one. If I go with my kids to school, then I go to work, and I come back, and I pick up my kid, the two trips are going to work fine. I know it's as clear. Thank you so much, Sebastian. It's really wonderful to see all the beneficial uses of this data to understand public service uses and things like that. Also, with my imagination, I can think of some harmful ways to use this data, some funny, some not so funny. And I wonder who governs this data, and if you have measures to protect individual privacy and security. Well, I mean, we have the data that we get from the government isn't masqueraded. They are not the real IDs, so we can track the users, but we don't know anything about the user. We just know the latitude and longitude of where someone took the trip. And we have, there are like 15 million records day, so it's very difficult. But there are also ways to, you can work with samples. And if you, I mean, actually, it's not that easy to get the data. We got some data using like a public information request, but eventually there are ways you can use or the government can use to provide this data and not have the problems of privacy information. But as I tell you, there are so many data, there's a lot of data, and we don't know anything about the user, we only, we have like a masqueraded ID card. I have a question. Can you tell us more about the data cleaning part of this? You just mentioned that you were able to get the data from some like Freedom of Information Act requests. How clean or messy is it? How much time do you spend on it? Well, actually, the data set is large, but it's not that complicated, because I mean, if you want to analyze many data at once, but at the basic level, this is the data set. So it's not that complicated. It's that you have the ID and information about the trip, and then you have the GPS system. So basically, you try to locate using the timestamp and the data ID, the closest one. And so you have to clean a little bit some of the data, because as I said before, in some cases you have only one trip a day, so you have to, but I mean, it took some time. But this is something that we validated a lot. We tried this, not only for Buenos Aires, we tried this for the city of Córdoba, which has a different system. It doesn't have, I mean, in Buenos Aires, we have the SUBE. But in the city of Córdoba, they have other enterprise, and it also works. But I mean, we try to do as much of the cleaning as possible in the package. So the package handles that. Very nice. Hopefully, well. All right, we have time for one more quick question if anyone has one. All right, thank you so much for that talk. Can we have a big hand? Thank you. Thank you so much.