This Answer Sheet should be used for your VAST Challenge 2014 Mini-Challenge 2 submission.  Please maintain the .htm format and make sure that all hyperlinks are relative to the answer form. 

Rename this form "index.htm" for your submission.  Remove these instructions and any other example text below that is highlighted in yellow. Please see the "Submission Instructions" at http://vacommunity.org/VAST+Challenge+2014 for more detailed instructions.    

Entry Name:  "GT-Stasko-MC2"

VAST Challenge 2014
Mini-Challenge 2

 

 

Team Members:

Hannah Kim, Georgia Institute of Technology, hannahkim@gatech.edu PRIMARY

Jaegul Choo, Georgia Institute of Technology, jaegul.choo@cc.gatech.edu

Francesco Poggi, Georgia Institute of Technology, fpoggi3@mail.gatech.edu

James Nugent, Georgia Institute of Technology, jnugent6@gatech.edu

Yi Han, Georgia Institute of Technology, yihan@gatech.edu

Mengdie Hu, Georgia Institute of Technology, mengdie.hu@gatech.edu

Haesun Park, Georgia Institute of Technology, hpark@cc.gatech.edu

John Stasko, Georgia Institute of Technology, john.stasko@cc.gatech.edu

Student Team:  

NO

 

Analytic Tools Used:

Custom web-based visualization using D3

Matlab

 

Approximately how many hours were spent working on this submission in total?

100 hours

 

May we post your submission in the Visual Analytics Benchmark Repository after VAST Challenge 2014 is complete?

YES

 

Video:

GT-Stasko-MC2-video.wmv

or http://eelst.cs.unibo.it/vast/GT-Stasko-MC2-video-highres.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.

We visualized each employee’s daily movement in a space-time cube view using tracking data (lines) and spending data (circles). Fig. 1-1 shows on weekdays, most employees buy coffee on their way to work around 8. They work till lunchtime and leave work around 5:30pm, as also confirmed by GAStech’s timeline view in Fig. 1-2. After work, some go to shops and others go home.

Fig. 1-1. Daily patterns (hour vs. location) from the space-time cube

Fig.1-2. Timeline view for GAStech

 

For further analysis, we developed a web-based visualization tool (available at <http://eelst.cs.unibo.it/vast/map/>) that provides a zoomable map of Abila, two sliders to filter days and hours, and a combo box to filter vehicles. Circles represent shops, whose color indicates shop types (e.g., brown for cafe). From weekdays’ patterns (Fig.1-3), we observe movements in the morning (top in Fig. 1-3) involve homes, coffee shops, and GAStech. They go to restaurants/bars/stores during lunch/after work. Interestingly, coffee shops people go to in the morning were not visited during lunch (middle in Fig.1-3) nor after work (bottom in Fig.1-3).

Fig. 1-3. Places frequently visited by employees

 

Employees not fitted to this pattern turned out to be truck drivers. We identified which truck driver used which truck in a specific date by matching their spending records and trucks’ GPS data. As a result, we discovered that they are associated with specific routes and shops (Fig. 1-4a), and using the tool available at http://eelst.cs.unibo.it/vast/heatmap/, we can see, for instance, that Cecilia Morluniau visited only two shops (Fig. 1-4b).

Fig.1-4. (a) Truck drivers’ patterns (date vs. location) from the space-time cube

(b) Heatmap and transaction summary for Cecilia Morluniau

 

During the weekend, not much activities were going on, but we found that executives play golf on Sunday (1/12 and 1/19), as shown in Fig. 1-5.

Fig. 1-5. Timeline view for Desafio Golf Course

 

 

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.

1. We first visualized the credit card transaction data as a heatmap (store by date) for each employee. When looking at Lucas Alcazar’s spending history (left in Fig. 2-1), one element stood out, corresponding to $10,000 at Frydo’s Autosupply n’ More. Upon clicking it, the timeline view for this shop is visualized (top-right in Fig. 2-1), revealing Lucas was not there when the spending occurred. The space-time cube for Lucas (bottom-right in Fig. 2-1) confirms this, and at this point of time, his spending was shown at Ouzeri Elian. The cube visualization additionally shows that his spending on U-Pump on 1/3 did not match his GPS location.

Thus, we checked the timeline view for this shop (left-side in Fig. 2-2) and found that Minke Mies was at this shop when this spending occurred. Additionally, as we checked another shop, Albert’s Fine Clothing (right-side in Fig. 2-2), Lucas and Minke were at this shop for a few hours together on 1/11. They might have used each other’s credit card due to being in a relationship.

Fig. 2-1. Abnormal spending ($10,000) by Lucas Alcazar at Frydo’s Autosupply n’ More on 1/13

Fig. 2-2. Potential connection between Minke Mies and Lucas Alcazar

 

2. Lucas shows other suspicious patterns in terms of when he stays at GAStech. Other than normal business hours, he re-visited GAStech late at night for a few hours on 1/6,1/8, 1/15, and 1/17 (red ellipses in Fig. 1-2).

 

3. By analyzing GPS data, we were able to identify the locations of four executives’ houses and visualized their GPS histories (Fig. 2-3). Not surprisingly, this visualization reveals that they usually stay in their houses during the night and the weekend. Interestingly, we found that two people stayed at each of their places at the night of a particular day till the morning, e.g., Isia and Loreto at Ada Campo-Corrente’s on 1/7-1/8, Minke and Loreto at Orhan Strum’s on 1/8-1/9, Minke and Loreto at Orhan Strum’s on 1/8-1/9, Hennie and Isia at Williem Vasco-Pais’ on 1/10-1/11, and Minke and Hennie at Ingrid Barranco’s on 1/13-1/14. Based on their staying pattern, we conjecture that these people had discussion with them or possibly kept watching them by turns.

Fig. 2-3. Timeline views for executives’ places

 

4. Our analysis on GPS data further revealed five additional locations where employees’ cars stayed more than one minute, excluding known places, e.g., stores and other places shown in the map, and employee’s houses. From the GPS records for these locations (Fig. 2-4), we found that Minke Mies, Hennie Osvaldo, Loreto Bodrogi, and Inga Ferro stayed briefly in all the five locations at around noon on different dates. Their presence sometimes occurred at the same time, e.g., Minke and Loreto (1/8) and Loreto and Inga (1/17) at Location1, Minke and Hennie (1/16) at Location2, Hennie and Inga (1/10) at Location3, and Hennie, Loreto, and Inga (1/15) at Location 4. They might have conspired something in these locations or exchanged something confidential among them.

Fig. 2-4. Timeline views for executives’ places

 

5. We checked the timeline view for Lidelse Dedos/Birgitta Frente, two of which seem to live together (top in Fig. 2-5a). This view shows that Hennie Osvaldo often visits them and sometimes spends the night together, e.g., on 1/8, 1/11-12, 1/15, and 1/18. We now checked Hennie’s house location, which is co-located with the houses of Isia Vann, Loreto Bodrogi, and Inga Ferro (bottom in Fig. 2-5a). A short absence (red ellipses in Fig. 2-5a) frequently happened in the evening among the other three, so we checked if they often have dinner together. To this end, we started from the matrix view of them,and clicked several spending records, e.g., Isia Vann’s spending on Guy’s Gyros on 1/19 (top in Fig. 2-5b). Now, the timeline view for this shop pops up, which shows many people had dinner at the same time on 1/19 (ellipse in Fig. 2-5b), including Isia, Loreto, Minke Mies, and Edvard Vann as well as Sten Sanjorge Jr. This might imply some suspicious meetings among them.

Fig. 2-5a. Timeline views for Dedos/BFrente and Osvaldo/IVann/Bodrogi/Ferro

Fig. 2-5b. Spending patterns for Osvaldo/IVann/Bodrogi/Ferro and the timeline view for Guy’s Gyros

 

6. We checked the timeline view for Frydos Autosupply n’ More (Fig. 2-6). First, we noticed that some people has regular spending records with no GPS records, e.g., Mat Bramar, Anda Ribera, Linda Lagos, Ruscella Mies Haber, Carla Forluniau, and Cornelia Lails, who had all  assistant titles. No GPS records was because assistants have no car assigned, their frequent spending patterns might be because they have to manage the cars of those they assist, e.g., CEO, CFO, COO, etc. The next noteworthy pattern is that some people, who did not mostly visit there before, has visiting/spending records on 1/18-1/19 (red ellipses), e.g., Axel Calzas, Lidelse Dedos, and Willem Vasco-Pais on 1/18 and Varja Lagos, Hennie Osvaldo, Isia Vann, and Edvard Vann on 1/19. We conjecture that these people might be those disappearing in the kidnapping accident and they might have known about it and prepared for it in advance. Furthermore, considering that Hennie visited Lidelse’s place, as we showed earlier, they might also have planned this kidnapping.

Fig. 2-6. Timeline view for Frydos Autosupply n’ More

 

7. We checked the timeline view for Chostus hotel (top in Fig. 2-7). Obviously, Sten Sanjorge Jr. spent the nights on 1/18-1/19 to play golf with other executives as described in our answer to the previous question. However, an interesting observation is that Isande Borrasca and Brand Tempestad visited there during the lunch time on 1/8, 1/10, 1/14, and 1/17 with spending records per visit per person. Clicking these events opens up the associated space-time cube view (bottom in Fig. 2-7). Isande and Brand are both drill technician in the same department, and both came from and went back to GAStech at almost the same time (red ellipses). Considering that they could have used a single car, which was not the case, they might be in a relationship and wanted to hide this from other employees.

Fig. 2-7. Timeline view for Chostus Hotel and the space-time cube for the associated visits

 

8. We checked the timeline view for Lars Azada. On 1/10 (Friday), many people gathered at his house at night. From our analysis on the email header data from Mini Challenge 1, we think that it is the event called Casino night (Fig. 2-8).  

Fig. 2-8. Timeline view for Lars Azada

 

9. As seen from our answers to the previous question, the truck GPS data has their own daily patterns. Now, we compared between the patterns of the first and the second weeks via the space-time cube view (Fig. 2-9). The two left and the two right views correspond to the first and the second weeks, respectively. This visualization reveals an interesting pattern that in the second week, the amount of the truck movements increased significantly (red ellipses). They kept driving back and forth between GAStech and the particular shops they normally visit, but without spending records for each visit, which used to be the often the case before. Although not shown here, we further found out from the space-time cube view of individual truck drivers that these dates correspond to each truck driver’s last work day in this week. For example, Henk Mies (orange-colored) normally works on Monday through Thursday, Cecilia Morluniau (cyan-colored) on Monday through Friday, Benito Hawelon (yellow-colored) on Tuesday and Thursday, and Albina Hafon (red-colored) only on Monday. These abnormal activities might be because the truck drivers might be aware of the kidnapping accident in advance and wanted to have alibis.

Fig. 2-9. Space-time cube view for trucks

 

10. We found interesting spending patterns between Kanon Herrero and Elsa Orilla. That is, they mainly use Kanon’s car to go somewhere, and Kanon uses his credit card while for the same transaction, Elsa uses her loyalty card. This finding is revealed by the timeline view (Fig. 2-10). In this view, we can see that Elsa does not have any GPS records but has spending history only reported in her loyalty card (red-colored spending amount text). On the same day, Kanon has both GPS records as well as the spending history only reported in his credit card (spending amount text with a non-white background color). We conjecture that they may be in a relationship.

Fig. 2-10. Timeline view for the places that Kanon Herrero and Elsa Orilla visited

 

11. From the timeline view for Kronos Capitol near Abila Park (Fig. 2-11), we found that Loreto Bodrogi, Edvard Vann, and Kanon Herrero were there on 1/18. I think this might be related to POK rally occurring on 1/23 from Mini Challenge 3 data.

Fig. 2-11. The timeline view for Kronos Capitol (near Abila Park)

 

 

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.

GPS data are not collected while the car (and GPS) is turned off, thus with no movement in between. If GPS locations change while no GPS data are reported, our space-time cube view highlights such location discrepancies as red lines. Axel Calzas’s car and car 107 (shared by truck drivers, Cecilia Morluniau and Irene Nant) contain such cases (Fig 3-1). It might be due to temporary power-off of GPS devices, but employees might have intentionally hidden their GPS records. Furthermore, car 107 was not tracked on its way to Carlyle Chemicals from Gastech right after the driver changed.

Fig 3-1. Space-time cube views for Axel Calzas and Truck 107

 

Additionally, the GPS device on Elsa Orilla’s car was malfunctioning. Elsa’s trajectories significantly jittered and were shifted to the upper-left direction (green arrows in Fig 3-2).

Fig 3-2 Map view for Elsa Orilla

 

Particular shops’ spending data did not match with GPS data. For the cafes visited in the morning, e.g., Bean There Done That, Brewed Awakenings, Jack's Magical Beans, and Coffee Shack, all their transactions were time-stamped exactly at noon (Fig 3-3). Additionally, its timeline view of Kronos Mart (left in Fig 3-4) revealed that most of its transactions were time-stamped 12 hour later than the actual GPS records, and after adjusting the timestamps of credit and loyalty card records, we could achieve the proper alignment of GPS and loyalty/credit card transaction data (right in Fig 3-4).

Fig 3-3. Timeline views with credit card transaction time-stamped at noon

Fig 3-4. Timeline view for Kronos Mart before and after adjusting data

 

217 pairs of credit/loyalty card records had the spending amount difference of $20, $40, $60, or $80 (credit > loyalty). These could be the cash withdrawal (cashback) in addition to their purchase when using the credit cards.