Entry Name: "SDU-Li-MC2"

VAST Challenge 2014
Mini-Challenge 2

Team Members:

Aonan Li, Shandong University, Akriox@hotmail.com PRIMARY
Zhaosong Huang, Shandong University, hzs147258@163.com

Student Team: YES

Analytic Tools Used:

Tableau

Excel

Echart, Based on Canvas, pure JavaScript chart library that provides intuitive, vivid, interactive, can be personalized custom data visualization charts

Sonux, developed by the Shandong University's VAST team, 2014

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

120

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

YES

Video:

SDU-MC2.mp4

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

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.

These five images below show the common daily routines for typical GAStech employees. We visualized vehicle tracking data by Sonux and Tableau, and then set opacity to find the real clustering trajectories. We considered all tracking data in workday except truck drivers’, because they got different daily routines from most employees. As we can see in the images and video, there are significant trends of vehicle movement.

Employees, who lived in the East and North, went to nearby cafe or restaurant for breakfast and then drove to the GAStech building in the South.	Fig1.1 6:00-9:00
Scarcely any employees drove out during the working time. There were just few employees buying coffee.	Fig1.2 9:00-11:00 15:00-17:00
Employees went to nearby cafe or restaurant for lunch and then drove back to the GAStech building.	Fig1.3 11:00-15:00
Most employees drove home while others went for supper or shopping.	Fig1.4 17:00-19:00

Fig1.5 7:00-8:00 Consumption pie chart

We visualized credit card data by Echart to confirm the trends we found in trajectory images above. As we can see in the pie chart, employees preferred to go to Brew’ve Been Served, Hallowed Grounds, and Coffee Cameleon for breakfast.

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.

To find unusual events or patterns, we analyzed vehicle tracking data, credit card and loyalty card data.

2.2.1

We analyzed GPRS tracking data with Sonux, an analytic tool we developed for this mini challenge. Firstly to make the moving trends more significant, we converted those tracking points into trajectories (ID, start, end, x, y, time, car ID), and removed trajectories which contained no more than 10 points. This method can simplified these data by removing outliers. Secondly we partitioned map into 10^6 blocks, and then, counted each points in each blocks. For those low frequency blocks we marked them out and got all the singular points in them. After that we found which trajectories these singular points belong to. All the singular trajectories were suspected to be related to unusual events. This method can only find out patterns unusual in space, which is useless in searching abnormal data in time. So we analyzed these trajectories again, partitioned time instead of map space, and similarly found more singular trajectories. For each car’s own tracking data, we also calculated the abnormal trajectories, and compared them with trajectories above. After that we got all the suspected trajectories. By counting specific trajectories’ start points and end points, we could easily get the coordinates of each employees’ home, and of some cafe or restaurants, which can contribute to the analysis of suspected trajectories.

Fig2.1 Singular trajectories in blue

2.2.2

We also analyzed consumption data by making an area graph below. Different colors refer to different employees, and each employee’s data showed some common pattern. One’s consumption record in the first week were similar to the next week. By searching the differences and outliers between records in two weeks, we could find some unusual events.

Fig2.2 Consumption area graph

1 1/6 22:11-22:15 Lucas Alcazar drove to GASTech building from home. 1/7 1:10-1:14 Lucas Alcazar drove home from GASTech building. 1/16 0:09-0:13 Lucas Alcazar drove home from GASTech building. 1/17 22:41-22:45 Lucas Alcazar drove home from GASTech building. Why significant: Time singular trajectories. Level of confidence: 80%. Midnight activities. Too early (midnight) to go to work, too late to get off work.	Fig2.3 Lucas Alcazar
2 1/7 3:20-3:25 7:19-8:26 Loreto Bodrogi drove to Ada Campo-Corrente’s home close to Speson Park in the midnight and then drove to GASTech building in the morning (breakfast in Jack’s Magical Beans). 1/6 23:01-23:09 1/7 7:30-7:58 Isis Vann drove to Ada Campo-Corrente’s home close to Speson Park at 23:00 and then drove to GASTech building in the morning. Why significant: Time singular trajectories. Level of confidence: 90%. Midnight activities. Stay overnight. Trajectories intersected.	Fig2.4 Lorento Bodrogi
3 1/9 3:20-3:32 7:23-8:18 Loreto Bodrogi drove to Orhan Strum’s home in the midnight and then drove to GASTech building in the morning (breakfast in Jack’s Magical Beans). 1/8 23:00-23:06 1/9 3:30-3:40 Minke Mies drove to Orhan Strum’s home at 23:00, and then drove home in the midnight (2minutes before Loreto Bodrogi arrived). Why significant: Time singular trajectories. Level of confidence: 90%. Midnight activities. Stay overnight. Trajectories intersected.	Fig2.5 Lorento Bodrogi Fig2.6 Minke Mies
4 1/14 3:20-3:31 7:47-8:31 Minke Mies drove to Ingrid Barranco’s home in the midnight and then drove to GASTech building in the morning. 1/13 23:00-23:08 1/14 3:20-3:43 Hennie Osvaldo drove to Ingrid Barranco’s home at 23:00, and then drove home in the midnight (11minutes before Minke Mies arrived). Why significant: Time singular trajectories. Level of confidence: 90%. Midnight activities. Stay overnight. Trajectories intersected.	Fig2.7 Minke Mies Fig2.8 Hennie Osvaldo
5 1/8 11:17-12:46 Minke Mies and Lorento Bodrogi first drove to somewhere close to Bean There Done That and then drove to Guy’s Gyros. Why significant: Space singular trajectories. Level of confidence: 80%. Unusual trajectories. Synchronized action.	Fig2.9 Lorento Bodrogi and Minke Mies
6 1/10 17:06 - 1/11 0:30 There were about 15 employees gathering, seem to have a party. Marin Onda and Felix Balas stayed longer than others. Why significant: Time singular trajectories. Level of confidence: 60%. Unusual trajectories. Synchronized action. Maybe just a party.	Fig2.10 Employees gathering
7 1/13 19:20 Lucas Alcazar bought something in Frydos Autosupply n’ more, which cost him $10000. Why significant: Unusual consumption. Level of confidence: 85%. Huge consumption, far more than the average.	Fig2.11 Abnormal consumption

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.

Fig3.1 We found part of the credit card data and loyalty card data garbled caused by coding error. We corrected that garbled data with “Katerina’s Café”.
Fig3.2 There are duplicate data and conflicting data among the credit card data and the loyalty card data, which can be selected easily. For those duplicate data, we counted each instance for only once. For those conflicting data, we counted the more reasonable one in (closest to the average).
Fig3.3 There are some outliers in the vehicle tracking data. These points existed for only 1-3 seconds, and did not move anywhere, which is significant on the map. We filtered tracking data and just deleted those outliers.
Fig3.4 Elsa Orilla’s car (id: 28) got an abnormal trajectory. The tracking point moved disorderly in a small range, but showed a clear trend in a large range. We simplified these points which made them regular.