VAST Challenge 2017 MC1

Master in Data Mining and Knowledge Discovery
Universidad de Buenos Aires

Student Team: Blanco, Daniela; Cabrera, Omar; Centurión, Emmanuel; Demarchi, Agustín;
Garber, Leandro; Liberman, Gaston; Oyola, Diego; Sokil, Juan Pablo

This challenge proposes to find different types of patterns: daily and multiple day.

The approach used to find patterns was based in the quantity of days that vehicles stayed in the park. As a result of that, we decided to subset the dataset in two. On the one hand, those vehicles that entered and exited the park at the same day; on the other hand, those vehicles that entered and exited the park in different days. For example, a set of vehicles that entered one day at 23 pm and left the park at 2 am the next day or later are considered as part of a multiple day pattern. Using that criteria we identified 19,489 vehicles that entered the park within one year (from May 2015 to May 2016), with the following type and monthly distribution.

After this division we found out that the patterns were very different during the months of the year, which seems related with a seasonal behavior. Winter and summer were the seasons with less and more entries respectively.

The first approach to analyze trajectories was using Gephi, creating a directed graph that allows us to see the similarities and differences between daily (left image) and multiple day trajectories (right image).

The graph shows connections between the different points in the park (gates, campings, entrances). The bigger the edge, the more the vehicles that went through that path.

Another approach used for trajectories was the sunburst plot (Ms-Excel). That allows us to see the differences from the daily and multiple day vehicles (at the left we can see the daily trajectories and at the right the multiple day ones) showing the proportion of point-to-point routes. The visualization shows by colour, the starting point of the vehicle, and inside each subdivision the proportion of each ending point in the trajectory.

A visualization was designed to include time reference. We classified the different sensors using a hierarchical cluster (R, ggplot library), and we found that there are different patterns in each month/season of the year.

After that we looked for different kinds of approaches to figure out patterns considering the type of cars trajectories (Tableau)

and also day of the week trajectories (tableau)

and with a directed graph that we designed by the team that allows to investigate all trajectories in a separated way. (D3)

Let’s see what we found:

Daily patterns

First pattern: Tourist buses that visit the park

The following route: Entrance 2-General Gate 5 – General Gate 2 – Ranger Stop 0 – Ranger Stop 2 – General Gate 1-Entrance 0.

Tourist buses (type 5 and 6). They make a 39-minute average trip for the reservation. The park's vehicles are kept 204 minutes in average inside the park, and the rest of the types of cars have a very dispersed behavior (much variability, which does not allow to infer a frequent behavior).

Second pattern: Trucks crossing the Park
The following route: Entrance1-General Gate 7-Entrance 3.
They are vehicles that cross the park without stopping. They have short runs, 27 minutes on average. 88% of these cases are type 2 (trucks with trailers). No specific pattern is observed throughout time or season (flow is even along space-time).

Third Pattern: Camping maintenance

The following route: Ranger Stop 5-gate 4-General Gate 3.

They are almost all 2 p vehicles (93%) leaving the parking a lot and crossing unauthorized spaces until arriving at Camping 8. They do it in the months of December and January and during working hours (8 am to 8 pm). They carry out the maintenance of that camping.

Fourth pattern: Park extension works.

The following route: Entrance1-Gate 2 – Ranger Stop 1

96% of them are 2 p vehicles. They travel between the 5 and 10 pm.

The vehicles cross the Gate2 and quickly arrive (5 minutes) to the RANGERSTOP1 (they do not stop), there they stay, they go and they come, pass repeatedly inside the sensor without traversing another one. They are kept within that sensor for about 1:30 hours. They are employees of the park who are working on the extension of the venue, probably making a route to improve the connections.

Fifth pattern: Picnic

The following route: Entrance 2-General Gate 3 – Camping 8

People who go camping during the day, go to camping 8 which is the most frequented and the closest. The turnout increases on the weekends, and during the summer months considerably increases the amount of concurrence (both on weekends and in the rest of the days). In most cases they are type 1 cars.

Sixth pattern: Passing vehicles

They carry out the following routes:

Entrance 0-General Gate 4-General Gate 7-Entrance 1

Two behaviors were clearly defined. In the first case, used by all types of vehicles (although to a greater extent of cars or two-axle trucks), crossing from 0 to 1, avoid the camping area. They delay between 30 and 40 minutes, taking on average 12 minutes for the first and last tranche, and another 7 for the intermediate, so it is assumed that they do not stop. They use the road to travel.

The second case, in the opposite direction and the same way, averages 32 minutes, and is also a frequent pattern, used by all types of vehicles.

Multiple day patterns

First Pattern: Becoming one with the nature

Entrance2-General-Gate2-ranger-Stop0-ranger-Stop2-General-gate1-general-gate4-general-Gate7-Camping5-Camping5-General-Gate7-Entrance3,

Vehicles that arrive on Saturday morning to the camping 5 and stay until Sunday noon. They cross all the park to the most remote camping. There are mostly Type 1 vehicles (50%) and the rest is divided between type 2 and 3 vehicles.

Second Pattern: Wild Weekend

They carry out the following routes: Entrance1- camping4 - camping4 - general-gate7 - entrance3,

These are vehicles that come and stay in the first camping closest to entrance 1, probably they are young people that go at night to party at the park. They are not interested in the nature. They just want to be free of adults.

Third Pattern: Camping Weekend

They carry out the following routes: Entrance1 - camping2 - camping2 - general-gate7 - entrance3,
These are vehicles that arrive at any time, especially during summer months, they don’t have much time, but they enjoy though. Probably, they are families that goes on a camping weekend, probably family with kids.

Fourth pattern: Camping with accommodations

Entrance0 - general-gate1 - ranger-stop2 - ranger-stop0 - general-gate2 - general-gate5- camping6 - camping6 - general-gate5 - general-gate2 - ranger-stop0 - ranger-stop2 - general-gate1

They cross along the park and go camping close to the ranger base. Probably, they want to stay in a common area where they can buy food, drinks and other stuff. In that place you can rent a car, so maybe they leave their cars there and get a transportation service to travel within the park.

Fifth pattern: Bird´s fans
Entrance3- general-gate7 - camping5 - camping5 - general-gate7 - entrance3
They go directly to the camping5 (the most remote one). This is probably one of the best points to sightsee birds in their natural habitat, so they go there and camp to see the birds at dawn.

Sixth pattern: A week in the forest

There are those who prefer to spend a whole week in the park, mainly motorbikes and cars but they prefer different campings to stay: bikes are quite fond of campings 4 and 5, but cars of 2 and 3.

Unusual patterns

First pattern: 2 Axle-Trucks at dawn by prohibited zone
On Tuesdays and Thursdays 2 axle-trucks have appeared in not allowed trajectories from 2 am to 5 am, probably with the intention to dump wastes.

Second pattern: vehicle that have stayed in the park in the whole period
Every month they stayed in a different camping.

Third pattern: Night camping fan.

It is a type 1 vehicle that enters at 22-23 p.m. on Sundays from entrance 0 and stays for 2 days in camping 6, and then leaves from entrance 0, after 5 days they come back to the park and do the same trajectory. The curiosity about this case is that the carID does not change when returning to the park.

Fourth pattern: Weekend camping fan.

Type 1 Vehicle that stays from Friday’s afternoon to Monday’s midnight, repeating this sequence for more than 3 months. For the first time they entered through entrance 1 and stayed at camping 4 for the weekend, then left through entrance 4. Since then, it has entered through entrance 4 and stayed in the same camping and leaving through the same entrance they entered.

Fifth pattern: Group of type 1 vehicles stopping at a not allowed stop.

On 10/07/2015 at 10 a.m. - 6 Type 1 vehicles spent the afternoon at the ranger-stop 1 checkpoint which is not an allowed stop.

Sixth pattern: Vehicles that avoid the gate2 sensor

These are vehicles that don’t go through the sensor, probably they cross the park for a not allowed area. They also share the same route between 7 hours and 13 hours a day 20/07/2015. The road corresponds to Entrance1-RangerStop0-Entrance1.

Dangerous patterns

Unusual pattern No. 1: Trucks at dawn by zone prohibited, probably unloading garbage.

Frequent daily pattern No. 6: Passing vehicles. The constant passage of vehicles generates sound pollution that keeps birds away.

Unusual pattern No. 6: Vehicles that bypass the sensor. This indicates that they access to protected areas and disturb the birds.