Purdue University – MobileAnalymator

VAST 2011 Challenge
Mini-Challenge 1 - Characterization of an Epidemic Spread

Authors and Affiliations:

      Victor Yingjie Chen, Purdue University, chen489@purdue.edu  [PRIMARY contact]
      Cheryl Zhenyu Qian, Purdue University, qianz@purdue.edu
      Li Zhang, Purdue University, lzhang3@purdue.edu

Tool(s):

MobileAnalymator is a customized version of an ongoing research project in the Purdue University to visually analyze geospatial-temporal information. It is in Client-Server architecture. Its client is developed by Adobe Flash to display and animate data. The server side uses Java and MySQL to query, compute, and serve data. The user can run the client on a tablet computer and from a web browser to analyze.

From the client side, we focus on two perspectives:

• Be flexible and easy: use limited screen space to display large volume of data and run properly in a web browser, so it can be used on mobile tablet devices (e.g. Motorola Xoom with 3G or Wifi).
• Animation to simulate data changes: Because of the limited human attention and visual memory, expecting a user to notice changes among static and similar images is challenging. We use node-link graphs and animations to simulate changes in front of the eyes and expect the salient movements can reveal implicit stories.

The server side queries data by:

• Queuing records through their creation time and keywords. These blogs can be further marked if they contain certain highlight keywords. Client side displays these blogs on the map according to their geo-locations. Blogs containing keywords are displayed in red.
• Tracing movement of bloggers through querying his/her blogs within a given period.

MobileAnalymator was designed and developed by Victor Chen and Cheryl Qian from Purdue University (Interaction Design graduate program– http://www.interactiondesign.us). Li Zhang designed its color schema and abstracted the grey map.

Video:

Online Hi Resolution Video

ANSWERS:

MC 1.1 Origin and Epidemic Spread: Identify approximately where the outbreak started on the map (ground zero location). If possible, outline the affected area. Explain how you arrived at your conclusion.

The outbreak started from the west end of Downtown at 8:10am on May18th. The investigation area should start from the location (93.4240, 42.2217) (the intersection point of two white lines in Figure 2). The bridge of Highway 610 should be paid more attention.

Figure 1: Blogs that were created between 8am to 9am on May 18th. Red dots are blogs containing highlighted keywords

Figure 2. Filtered sickness blogs created during the first 6 outbreaking hours. Blogs outside the white boundary are mainly talking about friends' illness.

The affected areas include Downtown, Eastside, south of Uptown, south of Suburbia, and north of Lakeside (Enclosed by white line in Figure 2). There are three highly affected area in downtown - Doom (93.4017 - 94.3865, 42.2224 - 42.2179), City Hospital (93.3697, 42.2243 - 93.3553, 42.2156), and end of highway 278 (93.3519, 42.2296 - 93.3322:42.2153).

MC 1.2 Epidemic Spread: Present a hypothesis on how the infection is being transmitted. For example, is the method of transmission person-to-person, airborne, waterborne, or something else? Identify the trends that support your hypothesis. Is the outbreak contained? Is it necessary for emergency management personnel to deploy treatment resources outside the affected area? Explain your reasoning.

Our hypothesis is that the transmission method is airborne. Such a conclusion came from the following considerations:

Unlikely to be person-to-person: There are two reasons: Firstly, look at the statistic graph on the bottom, there is a small jump at 12am May 18th (mostly talking about their friends' illness, Figure 4), then a big jump at 8am. The peak is at 6pm of May 18th. If it is a person-to-person transmission, the curve should be growing smoothly and last for longer period. Secondly, we double checked the Person-to-Person method by checking blogs created at similar place and similar time (distance with in a given range and created within a certain period). For example, Figure 3 traced six sick bloggers' activities from 8am May 18th during the next 3 days. Analymator is able to record any other bloggers who sent blogs during the same time period and in the same location (within 40 meters and 2 hours). We assume that these bloggers had some person-to-person contacts with our original six bloggers. There is no clear evidence to show that these contacted bloggers got affected after the contacts.

Figure 3: Trace person-to-person contacts.

Figure 4. The trend of regular and sickness blogs. There is a sudden jump on May 18th.

Highly follow the wind direction: May 18th has strong west wind (9 miles/hour). Figure 2 also shows the affected area is expanded at the east side (unlike downtown, Eastside is widely distributed). Blogs outside the white boundaries are mostly talking about their friend's sickness (Figure 5). The bottom curve of Figure 4 shows that a lot of people started to get sick at almost the same time.

Figure 5. Detailed blogs created at 8am on May 18th in northern part of Uptown. They mainly talk about other people's sickness.

• The ground zero point: Using the build-in drawing tool, we can draw lines along boundaries of the affected area. We drew such two lines and got an intersection point. We suggest this point is the ground zero point that requires further investigation.
• The outbreak is stable. From the curve, it seems the number of sick people is stable during these three days except for a high peak at 6pm on the 18th. Blogs reporting sickness in latter days are mostly already affected in the first day. In order to confirm that, we use animation to check sick bloggers of the last day to trace their activities from in the past three days. We found that most of them had already got sick on the 18th. At night (starting from 6pm on May 18th), there are many sick blogs appears outside the affected area (especially Westside and Villa). We selected blogs in this area from 8pm to 11pm, and trace them using "by people" mode to check where they were in during the previous 12 hours (from 9am), we found they are almost all came from downtown and already affected (Figure 6).

Figure 6. Trace sick bloggers who blogged at night May 18th in Villa and Eastside (the blogger moved from the dark shade end to the bright yellow end).

To ensure the contained condition, we checked sick bloggers on May 20th. We selected sick people at 4pm on the 20th, used animation to check their activities during the three day period, and found that they had already got sick in previous days.

During the last day (20th), we found most sick bloggers went to the hospitals (last image in Figure 7).

Figure 7. Select sick bloggers who blogged on May 20th and trace their activities during the final three days.

There is no need to deploy treatment outside the affected area. As we saw from the pervious evidences, the affected area did not spread - there is no new affection coming out of the affected area in the later days. We should focus on the area of Downtown, south side of Uptown, Eastside, and the north side of Lakeside.

In order to search for how this outbreak took place, we checked blogs created close to our suspecting original point around 8am. There is one blog may worth checking: user ID 174633 at 18th 8:42am, location (93.4186, 42.2193), the content is: "Love is in the air. Well I'll put on a mask".

Another thing worth mentioning is that, at 9am on May 17th , there is a explosion happened at the northeast corner of Smogtown. The location is about (93.545, 42.182). We noticed this incident because there are suddenly a lot of blogs created after 9am. After examining the detailed content, we saw blogs at the northeast corner were talking about "hear", "fell", or "saw" by themselves. But in the southern part of Smogtown, people are talking about their friends "saw", "heard" the explosion.