[Script Info]
Title: 
[Events]
Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
Dialogue: 0,0:00:00.00,0:00:16.56,Default,,0000,0000,0000,,{\i1}hacc preroll music{\i0}
Dialogue: 0,0:00:16.56,0:00:21.34,Default,,0000,0000,0000,,Herald: And a lovely welcome back to the\Nhaccs stage on the third day this
Dialogue: 0,0:00:21.34,0:00:26.65,Default,,0000,0000,0000,,Congress, we are here with a talk on "A\Nfew quantitive thoughts on parking in
Dialogue: 0,0:00:26.65,0:00:35.76,Default,,0000,0000,0000,,Marburg" by Martin L. He's interested in\Ndata analytics now and infrastructure and
Dialogue: 0,0:00:35.76,0:00:41.06,Default,,0000,0000,0000,,traffic in general. And because of that,\Nhe started scraping publicly available
Dialogue: 0,0:00:41.06,0:00:46.51,Default,,0000,0000,0000,,parking data in Marburg and just went on\Nand analyzed it and found a lot of
Dialogue: 0,0:00:46.51,0:00:52.36,Default,,0000,0000,0000,,interesting things which he is going to\Npresent in this talk to you right now. In
Dialogue: 0,0:00:52.36,0:00:59.26,Default,,0000,0000,0000,,case you didn't know, there is IRC client\Non the live.hacc.media where you can ask
Dialogue: 0,0:00:59.26,0:01:04.23,Default,,0000,0000,0000,,questions later or with the #rC3hacc tag\Non Twitter.
Dialogue: 0,0:01:04.23,0:01:09.65,Default,,0000,0000,0000,,Martin Lellep: Welcome to my talk "A few\Nquantitative thoughts on parking in
Dialogue: 0,0:01:09.65,0:01:15.67,Default,,0000,0000,0000,,Marburg". I am delighted to speak here on\Nthis Congress because I love the yearly
Dialogue: 0,0:01:15.67,0:01:20.13,Default,,0000,0000,0000,,conferences. Also, thank you to the\Norganizing team for making all this
Dialogue: 0,0:01:20.13,0:01:25.99,Default,,0000,0000,0000,,possible. You do an absolutely fabulous\Njob. Now, the first question that you
Dialogue: 0,0:01:25.99,0:01:32.54,Default,,0000,0000,0000,,should ask is: why? The following is a\Npurely hobby project question, I came up
Dialogue: 0,0:01:32.54,0:01:37.14,Default,,0000,0000,0000,,with a question because transportation is\Nimportant, but unfortunately, it's also
Dialogue: 0,0:01:37.14,0:01:43.07,Default,,0000,0000,0000,,difficult. The most popular vehicles these\Ndays are cars and hence the question, how
Dialogue: 0,0:01:43.07,0:01:50.05,Default,,0000,0000,0000,,do people park in Marburg? Who am I? My\Nname is Martin, and I analyze publicly
Dialogue: 0,0:01:50.05,0:01:58.55,Default,,0000,0000,0000,,available data. I live close to Marburg,\Ntherefore the parking in Marburg. Now, a
Dialogue: 0,0:01:58.55,0:02:04.70,Default,,0000,0000,0000,,little bit of background regarding\NMarburg, it's a small picturesque, vibrant
Dialogue: 0,0:02:04.70,0:02:10.12,Default,,0000,0000,0000,,university town. There are a few highlights,\Nsuch as the castle, the old town and the
Dialogue: 0,0:02:10.12,0:02:16.56,Default,,0000,0000,0000,,river, just to name a few. It has around\N80,000 residents and a somewhat dense core
Dialogue: 0,0:02:16.56,0:02:22.45,Default,,0000,0000,0000,,around the old town. You can see a few\Npictures here of the castle, the old town
Dialogue: 0,0:02:22.45,0:02:29.67,Default,,0000,0000,0000,,and the river, respectively. Now, at this\Npoint, I would like to give my props to
Dialogue: 0,0:02:29.67,0:02:34.17,Default,,0000,0000,0000,,David Kriesel because all this work was\Ninspired by his amazing data science
Dialogue: 0,0:02:34.17,0:02:39.33,Default,,0000,0000,0000,,talks. You can find them on YouTube. And I\Nabsolutely encourage you to look for the
Dialogue: 0,0:02:39.33,0:02:49.21,Default,,0000,0000,0000,,Bahnmining, Spiegelmining and the Xerox\Nstory talks. OK, so if you have questions,
Dialogue: 0,0:02:49.21,0:02:53.84,Default,,0000,0000,0000,,then please ask, I will be there live\Nduring the Q&A of this conference and also
Dialogue: 0,0:02:53.84,0:03:01.51,Default,,0000,0000,0000,,you can send me an email with whatever you\Nlike, essentially. OK, so first of all, I
Dialogue: 0,0:03:01.51,0:03:06.53,Default,,0000,0000,0000,,would like to give a quick introduction to\Nthe data source. Now, the data, the
Dialogue: 0,0:03:06.53,0:03:13.89,Default,,0000,0000,0000,,parking data from Marburg is publicly,\Nwell it's published live on a system that
Dialogue: 0,0:03:13.89,0:03:18.24,Default,,0000,0000,0000,,is implemented by the city, by the city\Ncouncil, I believe . It's called
Dialogue: 0,0:03:18.24,0:03:26.23,Default,,0000,0000,0000,,Parkleitsystem Marburg or PLS for now, and\Nit publishes the data such as the parking
Dialogue: 0,0:03:26.23,0:03:31.60,Default,,0000,0000,0000,,decks, the number of free parking spots\Nand the location. The address here is
Dialogue: 0,0:03:31.60,0:03:39.07,Default,,0000,0000,0000,,pls.marburg.de. And let's see how it\Nlooks. Yeah, so obviously it's still
Dialogue: 0,0:03:39.07,0:03:44.32,Default,,0000,0000,0000,,online and you can see here the parking\Ndeck names listed, the number of free
Dialogue: 0,0:03:44.32,0:03:52.07,Default,,0000,0000,0000,,parking spots. Color coded is if it is\Nrather full or if it's rather empty, you
Dialogue: 0,0:03:52.07,0:03:58.39,Default,,0000,0000,0000,,can see here all of them are in the green.\NThe green color coding here, it's
Dialogue: 0,0:03:58.39,0:04:03.12,Default,,0000,0000,0000,,because it's probably close to Christmas.\NNobody wants to really park in the city.
Dialogue: 0,0:04:03.12,0:04:09.05,Default,,0000,0000,0000,,And the only one that's this one here, the\NMarktdreieck Parkdeck that it has some
Dialogue: 0,0:04:09.05,0:04:15.65,Default,,0000,0000,0000,,load to it. Then also there's a button\Ncalled route. So whenever you click on the
Dialogue: 0,0:04:15.65,0:04:22.69,Default,,0000,0000,0000,,on this button, say we we pick the\NErlenring-Center button, we are redirected
Dialogue: 0,0:04:22.69,0:04:31.64,Default,,0000,0000,0000,,to Google Maps and we can see here the\Nlocation of this parking deck, for
Dialogue: 0,0:04:31.64,0:04:37.23,Default,,0000,0000,0000,,example. Let's go back. Last but not\Nleast, there's also the maximum vehicle
Dialogue: 0,0:04:37.23,0:04:44.04,Default,,0000,0000,0000,,allowance and of course, the time stamp of\Nthe data. OK, back to the presentation
Dialogue: 0,0:04:44.04,0:04:49.04,Default,,0000,0000,0000,,now. This is a very simple website, so of\Ncourse it's easy to scrape and that's what
Dialogue: 0,0:04:49.04,0:04:58.23,Default,,0000,0000,0000,,I did. Regarding the scraper, I used a\NLinux computer and a docker container. And
Dialogue: 0,0:04:58.23,0:05:05.74,Default,,0000,0000,0000,,this scraper, you can see a small sketch\Nhere to the left, it simply visits the
Dialogue: 0,0:05:05.74,0:05:11.36,Default,,0000,0000,0000,,website every 3 minutes inside the docker\Ncontainer and writes the data into I
Dialogue: 0,0:05:11.36,0:05:17.51,Default,,0000,0000,0000,,believe it was csv files, which are\Nsubsequently used for the data analysis.
Dialogue: 0,0:05:17.51,0:05:24.70,Default,,0000,0000,0000,,All of it, the scraper and the analysis\Nscripts are written in Python. OK, the
Dialogue: 0,0:05:24.70,0:05:33.22,Default,,0000,0000,0000,,data format is pretty simple, it's\Nprocessed internally with data frames,
Dialogue: 0,0:05:33.22,0:05:37.37,Default,,0000,0000,0000,,with the package panda. Everybody who\Nknows Python probably knows panda, anyway.
Dialogue: 0,0:05:37.37,0:05:42.62,Default,,0000,0000,0000,,It's the data format is as follows. The\Nrow corresponds to the time. The column
Dialogue: 0,0:05:42.62,0:05:47.46,Default,,0000,0000,0000,,corresponds to the specific parking deck,\Nand the cell corresponds to the number of
Dialogue: 0,0:05:47.46,0:05:53.76,Default,,0000,0000,0000,,free parking spots at that time of that\Nparking deck. Now, in order to make the
Dialogue: 0,0:05:53.76,0:05:59.63,Default,,0000,0000,0000,,numbers a bit more usable, I transformed\Nthe number of free parking spots to the
Dialogue: 0,0:05:59.63,0:06:05.55,Default,,0000,0000,0000,,number of used parking spots by\Nsubtracting it from the maximum along the
Dialogue: 0,0:06:05.55,0:06:13.89,Default,,0000,0000,0000,,time. OK, now the intro is just to get\Nused to the data, we'd like to take a look
Dialogue: 0,0:06:13.89,0:06:19.53,Default,,0000,0000,0000,,at the locations of the of the park houses\Nor the park decks. This is a screenshot.
Dialogue: 0,0:06:19.53,0:06:26.60,Default,,0000,0000,0000,,There's an interactive version. Let me\Nopen it here. It's a interactive map. You
Dialogue: 0,0:06:26.60,0:06:33.50,Default,,0000,0000,0000,,can see two types of markers, the first\None red, the second one green, and that's
Dialogue: 0,0:06:33.50,0:06:39.11,Default,,0000,0000,0000,,because the red ones are the ones that are\Ngiven, well they are encoded in the links
Dialogue: 0,0:06:39.11,0:06:45.81,Default,,0000,0000,0000,,of the PLS system, and they\Nare actually wrong. So when you click on
Dialogue: 0,0:06:45.81,0:06:54.46,Default,,0000,0000,0000,,the for instance. Erlenring-Center parking\Ndeck that I've done before, the location,
Dialogue: 0,0:06:54.46,0:06:59.47,Default,,0000,0000,0000,,longitude and latitude are actually\Nincorrect and, um, Google Maps corrected
Dialogue: 0,0:06:59.47,0:07:04.41,Default,,0000,0000,0000,,on the fly. And therefore, I have shown\Nhere the ones given on the website that
Dialogue: 0,0:07:04.41,0:07:11.26,Default,,0000,0000,0000,,are incorrect in red and the ones shown\Nthat are correct. So you can safely focus
Dialogue: 0,0:07:11.26,0:07:17.07,Default,,0000,0000,0000,,only on the green ones. Um, a quick\Noverview here is the train station region,
Dialogue: 0,0:07:17.07,0:07:22.34,Default,,0000,0000,0000,,there are two. And then they are scattered\Naround the city. Um, sometimes there are
Dialogue: 0,0:07:22.34,0:07:30.16,Default,,0000,0000,0000,,two parking decks very close by, for\Ninstance, these two and these two. And
Dialogue: 0,0:07:30.16,0:07:34.18,Default,,0000,0000,0000,,that's because it's essentially one\Nparking deck with two parking sections
Dialogue: 0,0:07:34.18,0:07:40.46,Default,,0000,0000,0000,,typically inside the building and on top\Nof the building. OK, let's go back to the
Dialogue: 0,0:07:40.46,0:07:49.09,Default,,0000,0000,0000,,presentation. With that in place, we or we\Ntake a look at the joined data, meaning I
Dialogue: 0,0:07:49.09,0:07:55.64,Default,,0000,0000,0000,,accumulate the number of used parking\Nspots across all the parking decks. You
Dialogue: 0,0:07:55.64,0:08:00.11,Default,,0000,0000,0000,,can see that here now, so it's a quite\Ncomprehensive picture, I started data
Dialogue: 0,0:08:00.11,0:08:06.48,Default,,0000,0000,0000,,scraping in August 2019 and stopped it at\Nthe end of February 2020.
Dialogue: 0,0:08:06.48,0:08:13.18,Default,,0000,0000,0000,,This data here is a different resample\Nfrequency of the original and raw data. I
Dialogue: 0,0:08:13.18,0:08:17.37,Default,,0000,0000,0000,,started with a resample of one hour. So\Njust a reminder, it's the true frequency
Dialogue: 0,0:08:17.37,0:08:23.28,Default,,0000,0000,0000,,is three minutes. Again, I resampled here\Ninto one hour. It's not very easy to
Dialogue: 0,0:08:23.28,0:08:27.84,Default,,0000,0000,0000,,understand on that scale here. Then to one\Nday it's the orange now and lastly on one
Dialogue: 0,0:08:27.84,0:08:33.49,Default,,0000,0000,0000,,week and we can learn different things\Nfrom it. So in particular, the orange
Dialogue: 0,0:08:33.49,0:08:38.76,Default,,0000,0000,0000,,curve of one day shows that there might be\Nsome periodicity in the signal. And the
Dialogue: 0,0:08:38.76,0:08:43.87,Default,,0000,0000,0000,,green one shows that there are times or\Nweeks that are particularly... where
Dialogue: 0,0:08:43.87,0:08:47.64,Default,,0000,0000,0000,,there's particularly little parking\Ndemand, for instance, here around
Dialogue: 0,0:08:47.64,0:08:56.49,Default,,0000,0000,0000,,Christmas 2019. OK, so again, from the\Norange signal, you can see that there's
Dialogue: 0,0:08:56.49,0:09:01.57,Default,,0000,0000,0000,,probably some periodicity, and in order to\Nquantify that, I plotted the or computed
Dialogue: 0,0:09:01.57,0:09:06.66,Default,,0000,0000,0000,,the auto correlation function. The auto\Ncorrelation function essentially takes a
Dialogue: 0,0:09:06.66,0:09:11.57,Default,,0000,0000,0000,,time signal and computes the overlap\Nbetween the time signal and the same
Dialogue: 0,0:09:11.57,0:09:17.74,Default,,0000,0000,0000,,signal shifted by some time and whenever\Nthere's a large overlap. That points
Dialogue: 0,0:09:17.74,0:09:23.71,Default,,0000,0000,0000,,towards the periodicity, and here we can\Nsee that the periodicity maximum or the
Dialogue: 0,0:09:23.71,0:09:28.44,Default,,0000,0000,0000,,auto correlation maximum, the first one\Ncorresponds to one week and therefore the
Dialogue: 0,0:09:28.44,0:09:33.69,Default,,0000,0000,0000,,periodicity can be safely assumed to be at\Nseven days. Of course, when there's
Dialogue: 0,0:09:33.69,0:09:41.06,Default,,0000,0000,0000,,periodicity and a signal at seven days,\Nfor instance, there's also periodicity. In
Dialogue: 0,0:09:41.06,0:09:44.68,Default,,0000,0000,0000,,14 days and in 21 days, but the\Ncorrelation coefficients, they decay
Dialogue: 0,0:09:44.68,0:09:53.47,Default,,0000,0000,0000,,typically. OK, now we have the periodicity\Nwith respect to days in place. Now let's
Dialogue: 0,0:09:53.47,0:09:59.01,Default,,0000,0000,0000,,take a look at the day and hour demand.\NAnd for that, I computed a two dimensional
Dialogue: 0,0:09:59.01,0:10:05.95,Default,,0000,0000,0000,,histogram with the day Monday to Sunday on\Nthe one axis and the other axis
Dialogue: 0,0:10:05.95,0:10:11.78,Default,,0000,0000,0000,,corresponds to the hour. And here we can\Nclearly see that the majority of the
Dialogue: 0,0:10:11.78,0:10:17.82,Default,,0000,0000,0000,,parking demand is around the noon hour. So\Nstarting from 11 to to approximately,
Dialogue: 0,0:10:17.82,0:10:26.27,Default,,0000,0000,0000,,let's say, 5 p.m. or so. Interestingly.\NThat was a point where I was surprised is
Dialogue: 0,0:10:26.27,0:10:30.74,Default,,0000,0000,0000,,that Sunday's is a day where there's\Nlittle parking demand in Marburg, I
Dialogue: 0,0:10:30.74,0:10:35.72,Default,,0000,0000,0000,,wouldv'e guesstimated that Sunday when\Neverybody has spare time, they typically
Dialogue: 0,0:10:35.72,0:10:40.27,Default,,0000,0000,0000,,rush into the city. But that's obviously\Nnot the case. Another interesting fact is
Dialogue: 0,0:10:40.27,0:10:44.84,Default,,0000,0000,0000,,that Monday morning seemed to be very\Ndifficult to get up because you can see
Dialogue: 0,0:10:44.84,0:10:55.34,Default,,0000,0000,0000,,the parking demand is smaller than on on\Nother mornings. OK, now, after that, I
Dialogue: 0,0:10:55.34,0:11:01.49,Default,,0000,0000,0000,,come to the separated... separate and\Nanalysis where I take a look at the
Dialogue: 0,0:11:01.49,0:11:07.52,Default,,0000,0000,0000,,individual parking decks. So first of all,\Nagain, the times series, it's it's a bit
Dialogue: 0,0:11:07.52,0:11:11.54,Default,,0000,0000,0000,,dense and it's very hard to see. So there\Nare a few things to learn from the
Dialogue: 0,0:11:11.54,0:11:16.41,Default,,0000,0000,0000,,picture. So first of all, the green\Nsignal that corresponds to the Erlenring-
Dialogue: 0,0:11:16.41,0:11:22.32,Default,,0000,0000,0000,,Center. Reminder, I just opened it. In the\Nvery beginning of this talk seems to be
Dialogue: 0,0:11:22.32,0:11:32.50,Default,,0000,0000,0000,,the dominant one, then there are quite a\Nfew data gaps. So take for instance. Well,
Dialogue: 0,0:11:32.50,0:11:37.37,Default,,0000,0000,0000,,it's very apparent here for the violet\None, the Furthstraße-Parkdeck, this one
Dialogue: 0,0:11:37.37,0:11:44.53,Default,,0000,0000,0000,,here. And that's an extreme case. It had\Nobviously some kind of problem. It was
Dialogue: 0,0:11:44.53,0:11:49.28,Default,,0000,0000,0000,,open for some time and then closed for\Nsome other times. Typically, park houses
Dialogue: 0,0:11:49.28,0:11:54.28,Default,,0000,0000,0000,,or parking decks are either open 24/7, but\Nthere are also quite a few that are that
Dialogue: 0,0:11:54.28,0:12:07.03,Default,,0000,0000,0000,,close overnight. OK, next I was interested\Nin the statistics of parking demand for
Dialogue: 0,0:12:07.03,0:12:13.92,Default,,0000,0000,0000,,individual parking decks, so I\Nconcentrated only on, say, one parking
Dialogue: 0,0:12:13.92,0:12:20.68,Default,,0000,0000,0000,,deck and computed the histograms of the\Nused parking spots also, depending on the
Dialogue: 0,0:12:20.68,0:12:27.53,Default,,0000,0000,0000,,time. Let's focus here on the Oberstadt,\Nit's the old town and you can see that the
Dialogue: 0,0:12:27.53,0:12:37.13,Default,,0000,0000,0000,,overall parking demand peaks at around,\Nlet's say, maybe 20 used parking spots, so
Dialogue: 0,0:12:37.13,0:12:42.47,Default,,0000,0000,0000,,that's the average, but that's not for all\Ntimes when we make that statement,
Dialogue: 0,0:12:42.47,0:12:45.60,Default,,0000,0000,0000,,depending on the time, for instance, the\Nmorning we can see that's approximately
Dialogue: 0,0:12:45.60,0:12:51.63,Default,,0000,0000,0000,,the same. But when we go towards noon, we\Ncan see that the number of parking spots
Dialogue: 0,0:12:51.63,0:12:58.68,Default,,0000,0000,0000,,or used parking spots increases. There are\Neven a few times when it's at the maximum
Dialogue: 0,0:12:58.68,0:13:05.50,Default,,0000,0000,0000,,around noon. Now, when we go towards later\Nhours, the maximum shifts towards smaller
Dialogue: 0,0:13:05.50,0:13:11.68,Default,,0000,0000,0000,,values again. Now, this this behavior of\Nthe maximum shifting, so clearly,
Dialogue: 0,0:13:11.68,0:13:17.15,Default,,0000,0000,0000,,depending on the hour, is not not apparent\Nfor all the parking decks. For instance,
Dialogue: 0,0:13:17.15,0:13:24.01,Default,,0000,0000,0000,,the Parkdreieck here ... Marktdreieck,\Nsorry, that doesn't show the signal as
Dialogue: 0,0:13:24.01,0:13:32.89,Default,,0000,0000,0000,,clear as the Oberstadt one. OK, from this\Nall now we can quantify also the, I call
Dialogue: 0,0:13:32.89,0:13:37.66,Default,,0000,0000,0000,,it integral parking demand, simply it's\Nthe the number of parking spots that have
Dialogue: 0,0:13:37.66,0:13:44.76,Default,,0000,0000,0000,,been provided per parking deck. Now the\Npicture here, it's normalized to the
Dialogue: 0,0:13:44.76,0:13:49.70,Default,,0000,0000,0000,,maximum and one can see from this picture\Nhere very easily that the Erlenring-
Dialogue: 0,0:13:49.70,0:13:54.20,Default,,0000,0000,0000,,Center, as we've estimated or guessed\Npreviously already is the one that's
Dialogue: 0,0:13:54.20,0:14:01.72,Default,,0000,0000,0000,,dominating the whole city. It's providing\Nthe most parking spots by a large margin,
Dialogue: 0,0:14:01.72,0:14:07.69,Default,,0000,0000,0000,,actually. The next one is the Lahn-Center\Nand then maybe the Oberstadt and the other
Dialogue: 0,0:14:07.69,0:14:12.75,Default,,0000,0000,0000,,ones follow after these. Another\Ninteresting point here is that the
Dialogue: 0,0:14:12.75,0:14:20.61,Default,,0000,0000,0000,,proportion of parking spots provided on\Nweekends differs for the different parking
Dialogue: 0,0:14:20.61,0:14:25.30,Default,,0000,0000,0000,,decks. For instance, here you can see this\None here is quite a big portion, the
Dialogue: 0,0:14:25.30,0:14:29.74,Default,,0000,0000,0000,,Erlenring-Center, also on weekends.\NContrary, the Marktdreieck-Parkdeck has
Dialogue: 0,0:14:29.74,0:14:38.42,Default,,0000,0000,0000,,only a very small portion of, um, of\Nparking spots provided on weekends. It
Dialogue: 0,0:14:38.42,0:14:43.52,Default,,0000,0000,0000,,might be interesting to know that this\Nparticular parking station is ... it's the
Dialogue: 0,0:14:43.52,0:14:47.88,Default,,0000,0000,0000,,one that is used if you want to go to a\Ndoctor, because it's very close. So many
Dialogue: 0,0:14:47.88,0:14:51.09,Default,,0000,0000,0000,,doctors are not open on Sundays, on\NSaturdays, and therefore probably the
Dialogue: 0,0:14:51.09,0:14:56.13,Default,,0000,0000,0000,,parking demand is quite low. Now, there's\Na temporal version also where I rendered a
Dialogue: 0,0:14:56.13,0:15:02.23,Default,,0000,0000,0000,,small video that I'm opening now, and you\Ncan see essentially the same as in the
Dialogue: 0,0:15:02.23,0:15:07.70,Default,,0000,0000,0000,,previous graph, but against time. Again,\Nit's very apparent that there's a
Dialogue: 0,0:15:07.70,0:15:15.59,Default,,0000,0000,0000,,periodicity and here my scraper crashed\Nand it's back in business again, and I
Dialogue: 0,0:15:15.59,0:15:23.10,Default,,0000,0000,0000,,found it interesting to see that there are\Nparking decks that have cars... well that
Dialogue: 0,0:15:23.10,0:15:28.04,Default,,0000,0000,0000,,host cars, even at night, for instance,\Nhere the Erlenring-Center again in the
Dialogue: 0,0:15:28.04,0:15:33.50,Default,,0000,0000,0000,,Lahn-Center, the ones that are the largest\None, they offer parking also overnight.
Dialogue: 0,0:15:33.50,0:15:42.36,Default,,0000,0000,0000,,And there are some cars in there,\Nprobably. OK, let's close that again. Now,
Dialogue: 0,0:15:42.36,0:15:51.00,Default,,0000,0000,0000,,I come lastly to the prediction part now.\NThe goal here is to measure the parking
Dialogue: 0,0:15:51.00,0:15:56.34,Default,,0000,0000,0000,,demand through the parking decks, but then\Nto interpolate between the parking decks,
Dialogue: 0,0:15:56.34,0:16:02.17,Default,,0000,0000,0000,,so I would like... so I have ...say the\NOberstadt the old town and the, I don't
Dialogue: 0,0:16:02.17,0:16:05.54,Default,,0000,0000,0000,,know, the Erlenring, which was the largest\None. I would like to know what's the
Dialogue: 0,0:16:05.54,0:16:12.07,Default,,0000,0000,0000,,parking demand in between, for instance.\NFor doing so, I use a spatial fit and I
Dialogue: 0,0:16:12.07,0:16:16.42,Default,,0000,0000,0000,,use a machine learning model for that, in\Norder to do that spatial fit. It is now,
Dialogue: 0,0:16:16.42,0:16:22.14,Default,,0000,0000,0000,,in this particular case, a non parametric\Nmodel called Gaussian Process Regression.
Dialogue: 0,0:16:22.14,0:16:27.17,Default,,0000,0000,0000,,And the nice thing about that is that it\Nalso returns the uncertainty. Because say,
Dialogue: 0,0:16:27.17,0:16:31.40,Default,,0000,0000,0000,,for instance, you would like to use these\Nmodel, machine learning predictions to
Dialogue: 0,0:16:31.40,0:16:37.67,Default,,0000,0000,0000,,say, build some kind of parking deck or to\Nget rid of one. All these operations, all
Dialogue: 0,0:16:37.67,0:16:41.89,Default,,0000,0000,0000,,these derived actions would be very\Nexpensive. So you would like to know if
Dialogue: 0,0:16:41.89,0:16:45.80,Default,,0000,0000,0000,,the uncertainty is large or small for\Nwhatever the machine learning model
Dialogue: 0,0:16:45.80,0:16:51.45,Default,,0000,0000,0000,,predicts. Just for the math oriented\Npeople. If you're interested in that
Dialogue: 0,0:16:51.45,0:16:57.14,Default,,0000,0000,0000,,model, definitely take a look at the, I\Nwould call it, Gaussian process bible by
Dialogue: 0,0:16:57.14,0:17:05.89,Default,,0000,0000,0000,,Rasmussen. It's amazing to read. Yeah,\Nthere are two, um, evaluations now, I did.
Dialogue: 0,0:17:05.89,0:17:09.55,Default,,0000,0000,0000,,The first one is based on the whole data\Nset, so there's no spatial or..sorry...
Dialogue: 0,0:17:09.55,0:17:16.71,Default,,0000,0000,0000,,there's no temporal resolution. And what I\Ndo, I did well, I rrendered a video and I
Dialogue: 0,0:17:16.71,0:17:22.92,Default,,0000,0000,0000,,would like to explain you the outcome of\Nthat while it is running. The top picture
Dialogue: 0,0:17:22.92,0:17:26.93,Default,,0000,0000,0000,,here shows you the prediction by the\Nmachine learning model. And the the bottom
Dialogue: 0,0:17:26.93,0:17:35.25,Default,,0000,0000,0000,,picture shows you the uncertainty. The\Ntraining data, meaning the parking decks,
Dialogue: 0,0:17:35.25,0:17:41.17,Default,,0000,0000,0000,,is denoted by the black points. Now, first\Nof all, the uncertainty, you can see that
Dialogue: 0,0:17:41.17,0:17:46.22,Default,,0000,0000,0000,,wherever there is training data, the\Nuncertainty goes down. So the model is,
Dialogue: 0,0:17:46.22,0:17:50.62,Default,,0000,0000,0000,,um, certain about its prediction that\Nbecause, well, there's training data and
Dialogue: 0,0:17:50.62,0:17:57.69,Default,,0000,0000,0000,,in between the uncertainty rises again.\NNow the prediction, you can see some small
Dialogue: 0,0:17:57.69,0:18:05.64,Default,,0000,0000,0000,,hill. It's exactly the Erlenring-Center,\Nwhich was the largest one. Now, what is
Dialogue: 0,0:18:05.64,0:18:10.72,Default,,0000,0000,0000,,shown in the video is it's rotating. You\Ncan see the coordinates of Marburg on the
Dialogue: 0,0:18:10.72,0:18:16.35,Default,,0000,0000,0000,,on the plane, on the bottom plane. And at\Nsome point, the view rotates upwards and
Dialogue: 0,0:18:16.35,0:18:21.14,Default,,0000,0000,0000,,gives you a top down perspective with a\Ncorresponding color bars or corresponding
Dialogue: 0,0:18:21.14,0:18:28.85,Default,,0000,0000,0000,,color map. So, again, here's the the\Nmaximum, the Erlenring-Center. And I did
Dialogue: 0,0:18:28.85,0:18:37.15,Default,,0000,0000,0000,,that because next we would like to finally\Nmeasure the parking demand between
Dialogue: 0,0:18:37.15,0:18:44.57,Default,,0000,0000,0000,,stations. OK, there's another small video\Nagain, and now we start right from the top
Dialogue: 0,0:18:44.57,0:18:48.61,Default,,0000,0000,0000,,down, color coded view and again, the\Nblack points are the... is the training
Dialogue: 0,0:18:48.61,0:18:53.12,Default,,0000,0000,0000,,data, but now the red points are, is kind\Nof test data, meaning positions in
Dialogue: 0,0:18:53.12,0:18:59.74,Default,,0000,0000,0000,,between. I concentrated now on the Mensa\Nbecause I have a special relation with the
Dialogue: 0,0:18:59.74,0:19:04.73,Default,,0000,0000,0000,,Mensa, the physics department, the\Nuniversity library, the train station and
Dialogue: 0,0:19:04.73,0:19:09.79,Default,,0000,0000,0000,,the cinema. And just to demonstrate from\Nthis spatial fit, we can derive the
Dialogue: 0,0:19:09.79,0:19:14.74,Default,,0000,0000,0000,,parking demand at these positions also.\NHere, this yellow pike, it's the
Dialogue: 0,0:19:14.74,0:19:21.50,Default,,0000,0000,0000,,Erlenring-Center again. Now, that's only a\Nqualitative result, of course, I don't
Dialogue: 0,0:19:21.50,0:19:25.17,Default,,0000,0000,0000,,want to derive any quantitative at this\Npoint, it's just a proof of concept that
Dialogue: 0,0:19:25.17,0:19:32.40,Default,,0000,0000,0000,,it is possible to derive something like\Nthat from the publicly available data.
Dialogue: 0,0:19:32.40,0:19:36.50,Default,,0000,0000,0000,,Now I forgot to mention the beginning that\Nthere's a bonus and I would like to come
Dialogue: 0,0:19:36.50,0:19:43.53,Default,,0000,0000,0000,,to the bonus now. It is about the Corona\Ncrisis or pandemic, of course. What I did
Dialogue: 0,0:19:43.53,0:19:48.02,Default,,0000,0000,0000,,is, the initial data acquisition phase,\Nhere in black, that's the whole talk was
Dialogue: 0,0:19:48.02,0:19:54.31,Default,,0000,0000,0000,,about that black portion here. I stopped\Nit at around the end of February and I
Dialogue: 0,0:19:54.31,0:20:00.12,Default,,0000,0000,0000,,restarted the whole data acquisition\Nprocess now again at in approximately
Dialogue: 0,0:20:00.12,0:20:05.59,Default,,0000,0000,0000,,April. Just to capture something from the\NCorona crisis as well. And you can see
Dialogue: 0,0:20:05.59,0:20:12.05,Default,,0000,0000,0000,,here again, the time series. I think the\Nmost interesting bit about it and the most
Dialogue: 0,0:20:12.05,0:20:16.74,Default,,0000,0000,0000,,comprehensive bit is the the mean . You\Ncan see the the mean across the whole time
Dialogue: 0,0:20:16.74,0:20:20.82,Default,,0000,0000,0000,,denoted by this dashed line. And you can\Nsee that the mean is smaller. So during
Dialogue: 0,0:20:20.82,0:20:27.38,Default,,0000,0000,0000,,the Corona pandemic fewer people parked in\NMarburg, which is reasonable, I would say.
Dialogue: 0,0:20:27.38,0:20:34.69,Default,,0000,0000,0000,,But there are also times where the number\Nof parking spots decreased significantly.
Dialogue: 0,0:20:34.69,0:20:40.25,Default,,0000,0000,0000,,So for instance, right when the Corona\Ncrisis started in April and now the second
Dialogue: 0,0:20:40.25,0:20:46.34,Default,,0000,0000,0000,,wave in October, November, December, it is\Nvisible that the parking demand decreased
Dialogue: 0,0:20:46.34,0:20:53.22,Default,,0000,0000,0000,,a lot. And I went one step further and\Nwanted to know the the differences between
Dialogue: 0,0:20:53.22,0:20:59.43,Default,,0000,0000,0000,,pre Corona and during Corona also for each\Nof the parking decks, that's what I did
Dialogue: 0,0:20:59.43,0:21:04.73,Default,,0000,0000,0000,,here. It's now not the normalized parking\Ndemand, but the absolute parking demand.
Dialogue: 0,0:21:04.73,0:21:10.54,Default,,0000,0000,0000,,So now we can see also the absolute\Nnumbers, the black black bars you've seen
Dialogue: 0,0:21:10.54,0:21:16.44,Default,,0000,0000,0000,,previously already. Now the red bars is\Nfor the during the Corona crisis. And then
Dialogue: 0,0:21:16.44,0:21:22.34,Default,,0000,0000,0000,,I defined these, the first wave and the\Nsecond wave as serious corona times. So I
Dialogue: 0,0:21:22.34,0:21:25.83,Default,,0000,0000,0000,,also plotted a third bar... set of bars\Nhere. And it's interesting to see that
Dialogue: 0,0:21:25.83,0:21:32.73,Default,,0000,0000,0000,,while most of the parking decks, of\Ncourse, suffered in terms of providing
Dialogue: 0,0:21:32.73,0:21:39.63,Default,,0000,0000,0000,,parking demands or most of them provided\Nfewer parking decks, parking spots. But
Dialogue: 0,0:21:39.63,0:21:44.12,Default,,0000,0000,0000,,there are a few, like, for instance, the\NMarktdreieck-Parkdeck here that, well,
Dialogue: 0,0:21:44.12,0:21:49.71,Default,,0000,0000,0000,,almost increased. We can see during the\Ncorona in general it increased a bit. And
Dialogue: 0,0:21:49.71,0:21:54.19,Default,,0000,0000,0000,,then during the heavy corona, it increased\Neven more. And as I mentioned before, this
Dialogue: 0,0:21:54.19,0:22:00.33,Default,,0000,0000,0000,,is the parking deck that corresponds to,\Nyeah, a whole collection of doctors. So. I
Dialogue: 0,0:22:00.33,0:22:07.22,Default,,0000,0000,0000,,derive that well during Corona times the\Nparking demand in front of doctors even
Dialogue: 0,0:22:07.22,0:22:13.37,Default,,0000,0000,0000,,increased a tiny bit. Yeah, with that, I\Nwould like to come to my conclusions.
Dialogue: 0,0:22:13.37,0:22:18.10,Default,,0000,0000,0000,,Thank you for sticking with me until now.\NSo I scraped publicly available data here
Dialogue: 0,0:22:18.10,0:22:26.84,Default,,0000,0000,0000,,with a small scraper set up. I analyzed\Nit, for instance, for day and hour
Dialogue: 0,0:22:26.84,0:22:30.89,Default,,0000,0000,0000,,patterns. And last but not least, did some\Nmachine learning in order to quantify the
Dialogue: 0,0:22:30.89,0:22:38.78,Default,,0000,0000,0000,,demand in between the stations, there is\Nan accompanying blog article also. You can
Dialogue: 0,0:22:38.78,0:22:42.17,Default,,0000,0000,0000,,find it down here, there all the figures\Nin higher resolution and you can play
Dialogue: 0,0:22:42.17,0:22:47.52,Default,,0000,0000,0000,,around with an interactive map also, if\Nyou like. Um, and to finally now conclude
Dialogue: 0,0:22:47.52,0:22:51.42,Default,,0000,0000,0000,,the presentation. I would like to hear\Nfrom you what you think about this
Dialogue: 0,0:22:51.42,0:22:57.46,Default,,0000,0000,0000,,analysis. I'd like to improve with these\Nkind of mini studies. And therefore, I
Dialogue: 0,0:22:57.46,0:23:03.06,Default,,0000,0000,0000,,would be very interested in your critique\Nregarding the content, the presentation
Dialogue: 0,0:23:03.06,0:23:08.02,Default,,0000,0000,0000,,and general content... comments. Again,\Nyou can email me to this email address
Dialogue: 0,0:23:08.02,0:23:15.92,Default,,0000,0000,0000,,here, or alternatively, I set up a Google,\Num, Google form. So the Google forms
Dialogue: 0,0:23:15.92,0:23:20.51,Default,,0000,0000,0000,,document that exactly comprised of these\Nquestions, and you can simply type them in
Dialogue: 0,0:23:20.51,0:23:24.92,Default,,0000,0000,0000,,if you're interested. Thank you very much.
Dialogue: 0,0:23:24.92,0:23:29.56,Default,,0000,0000,0000,,Herald: All right, first of all, thank you\Nfor this amazing talk, I have a few
Dialogue: 0,0:23:29.56,0:23:33.94,Default,,0000,0000,0000,,questions what have been relayed to me and\NI'm just going to ask them one after the
Dialogue: 0,0:23:33.94,0:23:39.07,Default,,0000,0000,0000,,other. And let's not waste any time and\Nstart with the first one. Have you found
Dialogue: 0,0:23:39.07,0:23:46.89,Default,,0000,0000,0000,,parking decks that are usually heavily\Noverloaded or never completely used?
Dialogue: 0,0:23:46.89,0:23:56.03,Default,,0000,0000,0000,,Martin: Um so. Given that there are only\Naround what was it, 8 or 9 or 10 in the
Dialogue: 0,0:23:56.03,0:24:05.16,Default,,0000,0000,0000,,data set, honestly, I never looked for for\Nthat question. So, um, short answers is:
Dialogue: 0,0:24:05.16,0:24:10.78,Default,,0000,0000,0000,,No. Long answer, yes, I could have or I\Nstill could, I would say.
Dialogue: 0,0:24:10.78,0:24:16.98,Default,,0000,0000,0000,,H: OK. Have you tried prediction in time,\Nso guessing which parking decks will be
Dialogue: 0,0:24:16.98,0:24:21.70,Default,,0000,0000,0000,,exhausted soon?\NM: No, no. So that's obviously it's
Dialogue: 0,0:24:21.70,0:24:26.75,Default,,0000,0000,0000,,like... it's... I would consider that\Nsomething like the predictive maintenance
Dialogue: 0,0:24:26.75,0:24:32.93,Default,,0000,0000,0000,,of traffic business kind of. It's\Ndefinitely a thing that people that have
Dialogue: 0,0:24:32.93,0:24:38.41,Default,,0000,0000,0000,,more time and more are willing to invest\Nmore definitely should do and could do. I
Dialogue: 0,0:24:38.41,0:24:43.77,Default,,0000,0000,0000,,would say I mean, there's lots of lots of\Nadditional data that might be of interest,
Dialogue: 0,0:24:43.77,0:24:49.37,Default,,0000,0000,0000,,like weather data. And, for instance, is\Nit a is it a public holiday, yes or no and
Dialogue: 0,0:24:49.37,0:24:55.58,Default,,0000,0000,0000,,all that kind of stuff. So, again, short\Nanswer.: No. Long answer. Yes. Would be
Dialogue: 0,0:24:55.58,0:25:00.92,Default,,0000,0000,0000,,possible.\NH: OK, so if anyone watching has the time
Dialogue: 0,0:25:00.92,0:25:05.71,Default,,0000,0000,0000,,or energy to do that, they could.\NM: Absolutely. Yes.
Dialogue: 0,0:25:05.71,0:25:10.65,Default,,0000,0000,0000,,H: OK, and the last question I have right\Nnow is, will the code or especially the
Dialogue: 0,0:25:10.65,0:25:16.51,Default,,0000,0000,0000,,scraping part be available publicly or\Nlike in the GitHub or somewhere?
Dialogue: 0,0:25:16.51,0:25:24.58,Default,,0000,0000,0000,,H: Um, I could do that. So I was very I\Nwas quite hesitant with it. So obviously
Dialogue: 0,0:25:24.58,0:25:29.84,Default,,0000,0000,0000,,publishing the data could be problematic.\NI have no experience with it on the legal
Dialogue: 0,0:25:29.84,0:25:34.15,Default,,0000,0000,0000,,side. So I would probably not publish the\Ndata, which is I mean, it's old data
Dialogue: 0,0:25:34.15,0:25:41.04,Default,,0000,0000,0000,,anyway. So and but then regarding the\Ncode, I was just waiting if anybody's
Dialogue: 0,0:25:41.04,0:25:45.11,Default,,0000,0000,0000,,interested. So given that somebody stated\Nthe interest, I would probably publish it.
Dialogue: 0,0:25:45.11,0:25:49.64,Default,,0000,0000,0000,,Yes.\NH: OK, yeah I think that's it from the
Dialogue: 0,0:25:49.64,0:25:52.18,Default,,0000,0000,0000,,question side .\NM: Hmhm.
Dialogue: 0,0:25:52.18,0:25:58.56,Default,,0000,0000,0000,,H: And they were all answered quite\Nnicely. And judging by that, I don't get
Dialogue: 0,0:25:58.56,0:26:04.81,Default,,0000,0000,0000,,any more questions right now. So, yeah, I\Nwould conclude is talk. Maybe you can also
Dialogue: 0,0:26:04.81,0:26:08.37,Default,,0000,0000,0000,,like have a last word. From my side I'm\Ndone here.
Dialogue: 0,0:26:08.37,0:26:14.78,Default,,0000,0000,0000,,M: Yes. So, um, well, thank you very much\Nfor watching the talk. And I try to
Dialogue: 0,0:26:14.78,0:26:19.60,Default,,0000,0000,0000,,improve. I think I said it on the last\Nslide. If I'm right, let me know if you
Dialogue: 0,0:26:19.60,0:26:26.11,Default,,0000,0000,0000,,have any doubts or things to improve\Nessentially on. And then regarding maybe
Dialogue: 0,0:26:26.11,0:26:31.72,Default,,0000,0000,0000,,the last question of publishing it, I\Nbelieve that I put a link there to find my
Dialogue: 0,0:26:31.72,0:26:37.70,Default,,0000,0000,0000,,blog and I would probably just add another\Nblog post stating well there's an github
Dialogue: 0,0:26:37.70,0:26:42.24,Default,,0000,0000,0000,,repository. You can go there and just find\Njust find the code and stuff like that
Dialogue: 0,0:26:42.24,0:26:48.26,Default,,0000,0000,0000,,there. So if you're interested, just, you\Nknow, find my website. My name is Martin
Dialogue: 0,0:26:48.26,0:26:55.69,Default,,0000,0000,0000,,Lellep. Um, and then you will in a few\Ndays, I guess probably in 2021 only. So I
Dialogue: 0,0:26:55.69,0:27:00.08,Default,,0000,0000,0000,,won't be able to publish it in the next\Ntwo days. But then the code will be
Dialogue: 0,0:27:00.08,0:27:06.05,Default,,0000,0000,0000,,public. Yes.\NH: OK, then. Have a great day. Great time
Dialogue: 0,0:27:06.05,0:27:08.74,Default,,0000,0000,0000,,at Congress and byebye.
Dialogue: 0,0:27:08.74,0:27:10.81,Default,,0000,0000,0000,,{\i1}postroll music{\i0}
Dialogue: 0,0:27:10.81,0:27:38.00,Default,,0000,0000,0000,,Subtitles created by c3subtitles.de\Nin the year 2021. Join, and help us!