Entry
Name: "NTU-Wall-MC2"
VAST
Challenge 2014
Mini-Challenge 2
Team Members:
Lina Wall, Nanyang Technological University Singapore, CA0001LL@e.ntu.edu.sg PRIMARY
Jing Guo, Nanyang
Technological University Singapore, GUOJ0020@e.ntu.edu.sg
Student Team: YES
Analytic Tools Used:
Excel
Access
Matlab
MySQL
ArcMap (10.2.2)
Approximately how many hours were spent
working on this submission in total?
Around 120
May we post your submission in the
Visual Analytics Benchmark Repository after VAST Challenge 2014 is complete? Yes
Video:
https://www.dropbox.com/s/2h1xqnqrswl57z6/MiniChallenge2-LinaJing.wmv
Questions
MC2.1 – Describe common daily routines for GAStech employees. What does a day in the life of a typical GAStech employee look like? Please limit your response to no more than five images and 300 words.
Answers:
The typical daily routine of a GasTech
employee starts with a breakfast break before going to work. After work the
employee drives back home before going out again to a restaurant or bar. After
dinner there is time for additional activities like shopping.
We analysed, based on the employees with a company car, the average time
spent on different activities separately for the four different job types:
Executives, Engineers, IT-Staff and Security.
The duration of the lunch time varies between the people of different
jobs. While the Engineers have a longer lunch break and spend more time at
home, the Executives in general have longer working days. During the weekend
the Executives go for golf courses, regularly. For CarID31, which belongs to Sten Sanjorge Jr., the
President/CEO of GasTech, only data for the last 3
days are available; in this time he stays at the hotel.
Img: Daily Routines
Img:
Example of Movement Profile for Executives
The routes of the trucks are slightly different between each working
day, but consist of regular repeated movement pattern. Normally the truck
drivers are using the same truck on different days. Only for truck No. 107 we
could detect a regularly exchange between two drivers: Fridays it is operated
with two shifts, morning shift and afternoon shift, and is used by two
different truck drivers, respectively. From the credit card transactions we
know that truck drivers have payments mainly at industrial places like Carlyle
Chemical Inc., Abila Airport, Stewart and Sons Fabrication, while the
transactions of the other employees are related to shops, restaurants, clothing
store, or supermarkets. The information from the credit card transaction like
name and function was merged with the time related gps coordinates of the
credit card holders. Through the analysis of all movement profiles we
identified the home location for each driver and found some overlaps.
Img:
Truck Routes
Img:
Truck Drivers Duty Table
Img:
Map with all identified locations
MC2.2 – Identify up to twelve unusual events or patterns that you see in the data. If you identify more than twelve patterns during your analysis, focus your answer on the patterns you consider to be most important for further investigation to help find the missing staff members. For each pattern or event you identify, describe
a. What is the pattern or event you observe?
b. Who is involved?
c. What locations are involved?
d. When does the pattern or event take place?
e. Why is this pattern or event significant?
f. What is your level of confidence about this pattern or event? Why?
Please limit your answer to no more than twelve images and 1500 words.
Answers:
Table:
Overview of abnormal events
|
No |
Pattern |
Who |
Where |
When |
Signific |
Conf. |
|
1 |
Invitation
@Home2 |
IDs 1, 2,
3, 5, 6, 7, 8, 9, 11, 14, 18, 19, 25, 26 (28 aligned) |
Home2 |
10 Jan evening |
Middle |
High |
|
2 |
Visitors
@Home10 |
CarID10, CarID15,
CarID16 |
Home10 |
6/7 Jan |
High |
High |
|
3 |
“Homes” of
ID21 |
CarID21 |
Home13/15/16,
Home14/18 |
often |
High |
High |
|
4 |
Special
nights |
CarID15,
CarID16, CarID21, CarID24 |
Home4,
Home10, Home32, Home35 |
6/7, 8/9,
10/11, 13/14 |
High |
High |
|
5 |
Special
locations |
CarID13,
15, 16, 21, 24 |
36.0502,24.9004 36.0529,24.8494 36.0589,24.8928 36.0632,24.9002 36.0695,24.8415 36.0806,24.8469 |
|
low |
low |
|
6 |
Interaction
@Capitol |
CarID 15,
22, 25, 34 |
Capitol |
11 Jan + 18 Jan |
High |
High |
|
7 |
Short
stops @Kalami Kalfenion |
CarID 15,
35 CarID 10,
11, 30 |
Kalami
Kalfenion |
18 Jan |
Middle |
Low |
|
8 |
Mismatch
CC and gps-location |
CarID1, |
Frydos
Autosupply n' More |
13 Jan |
High |
Mi |
|
9 |
High
Workload |
CarID 107 |
|
16 Jan |
Middle |
High |
|
11 |
Late work |
CarID 104 CArID 107 |
|
16 Jan 17 Jan |
Middle |
High |
|
12 |
Missing
data CarID9 |
CarID9 |
several |
often |
Middle |
Low |
No1: We condensed the gps
tracking information for each car in so called movement profiles. By browsing
through this simplified location maps, we detected that the place we identified
as Lars Azada’s Home (Home2) was found in nearly in all Engineers’ profiles and
in some of the IT staff. Comparing the date and time for this event we found
that they had a meeting (maybe a birthday celebration) during the evening of
January 10th.
Img:
Movement profiles for IT Staff
No2: In the movement
profiles we found an overlap between ID10, ID15, and ID16. In the night from 6th
to 7th of January Isia Vann (CarID:16) arrives at 23:09 at Ada
Campo-Corrente’s home (labelled as Home10) and stays there over night. At 3:20
of 7th Loreto Bodrogi (CarID:15) drives from his home, which is identically with the
home location for Isis Vann, to Ada’s home and arrives there at 3:35. In the
morning, Ada starts 10min earlier than the other two. All three have a stop at
Brewed Awakening; Ada leaves after 18 minutes (overlapping time with Isia and
Loreto less than 10 min), while Isia stays at the coffee shop for over 40
minutes.
Both, Isia Vann and Loreto Bodrogi, are employed at GasTech as security
(site and perimeter control); Ada Campo-Corrente belongs to the executives
(SVP/CFO).
The gps data for Elsa Orilla (ID28,
Engineer) are showing her on the 18th and 19th at the
place of Home 4. But ID28’s tracking signal is very noise: after aligning the
trajectory of ID28 there is no relation to Home 4 anymore.
This event is highly suspicious because it has an unusual timing and the
fact, that Ada leaves earlier than her guests, is against the expected
behaviour of a host.
Img:
Visitors at Home 10, 6/7 Jan
No3: While it was easy to
detect the home location in most of the movement profiles, for Hennie Osvaldo
(CarID: 21) it is not obvious where he lives. He has some overnight stays
during the weekend and on Wednesdays at the same location like Inga Ferro
(ID13), Loreto Bodrogi (ID15), and Isia Vann (ID16). The other nights he spent
at the same place like Lidelse Dedos (ID14) and Birgitta Frente (ID18). In the
night of 11th January, he drives at 3:32 to the home of Willem
Vasco-Pais (ID35; Executive), where 10 minutes earlier Isia Vann (ID16) has
left (arriving there 23:07, leaving time 3:23).
Img:
Home Locations of ID21
No4: Other suspicious
night activities related to the Homes of Executives were detected.
During the nights 6/7, 8/9, 10/11, 13/14 movements of the CarIDs 15, 16,
21, 24 round 3:30 were detected. The
normal procedure seems to be that everybody just spend have a night at the
point of interest. Under this condition the happening at Home 10 on the 6/7
where ID15 joins in and ID16 stays there for the whole night is “unusual”.
Img:
Night Activities at Home of Executives
No5: Recognized as special
places in the movement profile of Hennie Osvaldo (CarID: 21), a closer look
into the other movement profiles shows an overlap at this places with Inga
Ferro (ID13), Loreto Bodrogi (ID15), and Isia Vann (ID16). All of them are
working as Security (Site Control or Petrimeter Control) and the location might
be related to work tasks.
Img:
Special Locations of ID21
No6: Through our 3D
visualization of car movements we detected a meeting of Bodrogi Loreto
(CarID:15), Herrero Kanon (CarID:22), Nubarron Adra (CarID:25), and Vann Edvard
(CarID:34) at 18th Jan at a place, that matched into the tourist
maps was identified as Capitol. Already on the 11th ,
1 week earlier, Willem Vasco-Pais visited this place at 1:52pm
Img:
Meeting at Capitol
No7: Further we detected
in our 3D visualisation an interaction of CarID15 (Bodrogi Loreto),
CarID35 (Willem Vasco-Pais, executive), CarID10 (Ada Campo-Corrent, Executive),
CarID11 (Gustav Cazar, Engineer), and CarID30 (Resumir Felix, Security Manager) at Kalami Kalfenion on 18th
Jan.
Img:
Meeting at Kalami Kalfenion
No8: The highest payment
beside the truck drivers’ transfers belongs to Lucas Alcazar, CarID 1, and was
made on the 13th January. Checking the suspicious credit card
transaction of $10.000 at Frydos Autosupply we found that the gps locations of
Lucas Alcazar do not match with the credit card payments on this day. Also two
other transactions on this day are mismatches.
Against his habit
shown the week before to pay at least one meal per credit card, he has not used
his credit card on the following two days. This makes it even more suspicious
and we assume that he didn’t make the three payments by himself.
Img:
Credit Card Transactions per Location
Img:
Mismatches between Credit Card payment and location for CarID1
No9: Concerning the
locations where the car stays overnight CarID9 shows an uncommon pattern. While
most of the night locations are related to the respective homes, the movement
profile of CarID9 contains many star structures and has different locations for
a lot of the nights. By checking the detailed gps tracking path this strange
behaviour may understood as an effect caused by missing data.
Img:
“parking places” for CarID9
No10: Truck workload By drawing the
trajectory in 3D a high intensity for TruckID 107 is observable on the 15th
and on the 17th. Repeatedly he drives the same loop for several
times without stop. The three longest trips, we generated by concatenating
consecutive tracking points in a 1min window, take place on 16th .
The Trucks 104, 105, and 106 are involved and they have a duration of 2.5h (105
only 108min).
Img:
High Workload for TruckID:107 on 17th
No11: Truck overtime On the 16th
January, Mies Henk who drives TruckID:104 works until 21:06. This is much later
than normal; the other day he already finished around 4:30pm. Also the driver
Cecilia Morluniau who drives TruckID:107 works until 17:21 at 17th
Jan. Normally she only works during the morning.
Img:
“late work” for TruckID 104
MC2.3 – Like most datasets, the data you were provided is imperfect, with possible issues such as missing data, conflicting data, data of varying resolutions, outliers, or other kinds of confusing data. Considering MC2 data is primarily spatiotemporal, describe how you identified and addressed the uncertainties and conflicts inherent in this data to reach your conclusions in questions MC2.1 and MC2.2. Please limit your response to no more than five images and 300 words.
Answers:
The GPS data were condensed in two steps. First, we merged all tracking
points within a time span of 1 minute. Then, consecutive tracking units were
concatenated to complete trips. We assumed that as long
as the car doesn’t move no tracking entries are generated. The concatenation of
the tracking points lead to an overview of single trips with a start point and
end point, and departure time and arriving time. Checking this summarized data
shows that there are some noisy tracking records. If there is just one single
measurement within 2 minutes, this point seams not related to any car movement.
We verified this by calculating the distance to the predecessor and between
consecutive points. Then we removed these 1-point journeys.
Img:
Condensation to trips and Noise Reduction
We used some statistical measurements (average and standard deviation)
to detect credit card bookings having the same timestamp every day. The movement profiles link geographical
location, functions and the information of the credit card and loyalty card
transactions. Based on this, we were able to align the noisy signal of CarID28
to locations he visited.
Img:
Alignment of tracking path of CarID28
Jumps in the trajectory caused by inconsistent gps data or missing
tracking points were visualized as diagonals in the time-space diagram.
Img:
Jumps in trajectory
Mismatches in credit card transactions and loyalty card records were
detected by displaying the gps position (trajectory) in the maps and highlight
the places (coloured dots) where a transaction at the same time (day for
loyalty card) took place.
Img:
Credit Card transaction and trajectory
