Entry Name: "TJU-Li-MC3"

VAST Challenge 2019
Mini-Challenge 3

Team Members:

Zeyu Li, Tianjin University, lzytianda@tju.edu.cn PRIMARY

Jiaqi Wang, Tianjin University, qimelbourne@gmail.com

Shichao Jia, Tianjin University, jsc_se@tju.edu.cn

Jiawan Zhang, Tianjin University, jwzhang@tju.edu.cn

Student Team:

YES

Tools Used:

D3.js

Gephi

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

100

May we post your submission in the Visual Analytics Benchmark Repository after VAST Challenge 2019 is complete?

YES

Video

https://youtu.be/acVQqaw-KGY

Questions

The City has been using Y*INT to communicate with its citizens, even post-earthquake. However, City officials needs additional information to determine the best way to allocate emergency resources across all neighborhoods of St. Himark. Your task, using your visual analytics on the community Y*INT data, is to determine the types of problems that are occurring across the St. Himark. Then, advise the City on how to prioritize the distribution of resources. Keep in mind that not all sources on Y*INT are reliable, and that priorities may change over time as the state of neighborhoods also changes.

1 – Using visual analytics, characterize conditions across the city and recommend how resources should be allocated at 5 hours and 30 hours after the earthquake. Include evidence from the data to support these recommendations. Consider how to allocate resources such as road crews, sewer repair crews, power, and rescue teams. Limit your response to 1000 words and 12 images.

We summarized 10 topics to describe the condition of the city, such as power, water, and transportation. Each of them corresponds to a kind of resource, represented by a set of descriptors. For example, the descriptors of transportation topic including: bridge, routes, buses, etc. Messages that contains one of descriptors are regarded as belonging to corresponding topic.

By observing the changes in the number of messages contained in the earthquake topic (green), we identified three earthquakes:

Few citizens discussed these 10 topics at 5 hours and 30 hours after the first earthquake. Only transportation topic is observably mentioned at 5 hours after the third earthquake.

We mainly analyze the second earthquake which has a big impact on people's lives. The second quake struck at around 8:30 a.m. on April 8. Five hours later, citizens' discussion peaked, and 30 hours later a third earthquake struck.

Resource allocation recommendations at the fifth hour (the words in brackets are descriptors):

1. Water ("water") and sewer ("sewer", "pipe", "flood")

Within three to five hours of the quake, the city was experiencing major water problems. By looking at the word cloud and specific messages, we learn, concerned about the contamination of drinking water from nuclear leaks and broken sewers, citizens are buying bottled water or boiling it to use.

To distinguish between resource scarcity and recovery, we provide the ability to customize filter words to refine messages. After further applying filter words: "no;lack;need;thirsty", we identify three regions that most eager water: Downtown，Palace Hills, and Scenic Vista. And, plumbing and sewer workers should be sent to the following regions first：Southwest, Weston, and Southton.

2. Power ("power", "electric")

After applying filter words: "no;not;out;outage;need;electrician;faulty;flashlight;generator", we identify three regions that most eager electrical maintenance: Downtown, Palace Hills, and Southwest.

3. Medical rescue ("ambulance", "injur", " dying ", "help!", "hurt", "medic", "blood")

Few meagers mentioned about the above descriptors except the last word "blood" which is relatively hot. Checking the messages that contain "blood", we find that people were talking about donating blood, and some people have difficulty finding places to donate. Hence, Mobile blood donation site, like blood donation cars, should be decorated to Weston and Downtown.

4. Other rescues ("rescue", "police", "911", "crew", "damaged", "evacuat", "emergency", "collapse", "fire")

The four regions that need the most other rescues are Weston, Downtown, Northwest, and Scenic Vista. By comparing their word clouds, we find that the first three most need the fire brigade, Scenic Vista seems need evacuation assistants and rescue team.

Northwest and Weston happened more collapses.

Resource allocation recommendations at the thirtieth hour (Thu 09 3PM-6PM):

People are sending far fewer messages than they were in the fifth hour and only a portion of the resource requirements can be captured.

1. Water

After filtering by words: "no" and "out", we detect four regions that most need water: Weston, Downtown, Scenic vista and Cheddarford.

2. Other rescues

Oak Willow, palace Hills, Weston, and Southwest need firemen, Weston needs someone to help evacuating. Southwest needs people involved in collapse rescue.

3. Transportation

We find the messages that involved transportation around this moment seems like annunciates, talking about "please courteous use". Few accounts discussed traffic condition with words like "awful", "stuck", "worst", "bad", "gridlocked", "slow", "delay", "jam", "congestion", or block". It shows that there is no traffic jam, so there is no need for traffic police. And no clue indicates the need for road maintenance crews.

2 – Identify at least 3 times when conditions change in a way that warrants a re-allocation of city resources. What were the conditions before and after the inflection point? What locations were affected? Which resources are involved? Limit your response to 1000 words and 10 images.

In addition to the transportation and other rescues topic, the number of messages of other types of resources did not increase rapidly as soon as the earthquake occurred, but increased suddenly at the fifth hour after the earthquake and reached the maximum of the whole provided period.

We provide two perspectives to examine the dynamic changes in resource requirements.

From the perspective of one specific resource (topic).

1. Water

Water is the fastest growing of all resources at the fifth hour. But after an hour, it quickly loses its heat. Southton rocketed from sixth place at 1pm to first place at 2pm, it is the only region whose heat did not fade significantly, and it has been in the top 4 ever since. Therefore, more water-related resources should be allocated to Southton after 1pm.

Between 5pm and 6pm, there is a big change in requirements rankings. Regions at the top of the 5pm chart, such as Northwest, Cheddarford, and Weston, fell a lot, while regions at the bottom of the 5pm chart, such as Palace Hills, Downtown, and Old town, rose a lot after an hour.

2. Gas

Most resources reach maximum discussions at the 5th hour, while the gas category reaches maximum discussions at the 30th hour after the second earthquake. Therefore, it is necessary to focus on the dispatch of gas piping maintainers at the 30th hour.

3. Power

Palace Hills went from no. 11 at 1 PM to no. 1 at 9 PM and stayed there for the next 3 hours. While for most regions, demand for electricity resources fell at 2PM. After 5 PM, suddenly people in Palace Hills suddenly stopped talking about electricity.

4. Transportation

Let us focus on "bridge" which is mentioned most in transportation topic.

At Wed 08, these messages grew immediately and quickly after the earthquake. After checking the contents of the messages, we find some bridges were closed for inspection, and this aggravated

the morning rush hour traffic jam.

At Thu 09, transportation is the only category that involves a lot of messaging at 4 AM . However, the number of messages sent at 5 AM is much smaller than that at 4 AM, and the number of messages increased during 6 AM to 9 AM. We guess that the 4 AM is the time of the night shift or the morning shift, while 6 AM to 9AM is the time of the morning rush hour. After apply the filter words like "awful;stuck;worst;bad;gridlocked;slow;delay;jam;congestion;block", we identify a similar pattern. This means more traffic police should be sent to manage the traffic at 4 AM and 8 AM.

Other rescue

The region ranking of collapse topic has undergone a major transition from 7 AM to 8 AM on the 9th.The regions that needed more rescue was transformed from Broadview, Terrapin Springs, and safe Town to Palace Hills, Weston, and Oak Willow. If we change the interval to 4 hours, we can see a noticeable change in Downtown's ordering. The danger of fire mainly happened on the 8th, while the messages of rescue crew mainly happened on the 9th.

From the perspective of regions

We find:

1. For many regions on the 8th, the most needed resource changed from water to other rescue from 1pm to 2pm.

2. When earthquakes occur, the ranking of regions changes the most.

3. For most resources, the general evolution pattern is similar across regions, especially when earthquakes occur.

3 – Take the pulse of the community. How has the earthquake affected life in St. Himark? What is the community experiencing outside the realm of the first two questions? Show decision makers summary information and relevant/characteristic examples. Limit your response to 800 words and 8 images.

Besides changes in different resources, we build a social network base on users’ @ relationship. We can see several active users and the communities those active users make up, including Tonyeterson with RachelSimmons , ChemistryTeacherRichardNelson with DerekNolan and “Always Safe Power Company”.

Clicking a user on the network will display the timeline (middle) with the user as the analysis center. This timeline shows all messages sent by the user, and the node size indicates how many times it was reposted. The circle on the upper axis indicates the user who has a @ relationship with the current user, and the circle on the lower axis indicates the user who reposted the center user message.

The horizontal position of these users is determined based on the average time when the reposting or @ occurred. We can see that ChemistryTeacherRichardNelson and DerekNolan @ each other for many times(blue line means center user @ other users, red line is the opposite), which constitutes a dialogue.

By looking at the messages posted by DerekNolan, we judge him as a conspiracy theorist who spread rumors and create panic. He didn’t trust the government and alwaysSafePowerCompany, he thought the earthquake has caused very serious pollution but they concealed the news. ChemistryTeacherRichardNelson used his expertise to debate with him and refuted his absurd remarks.

Looking at the interaction between DerekNolan and alwaysSafePowerCompany, We found that the latter had two explanations for the former, but the former was not convinced, and then unilaterally repeated @ and questioned the latter.

The situation is similar between ChloeJohnson and alwaysSafePowerCompany, except that ChloeJohnson requested for help to alwaysSafePowerCompany for many times.

TonyPeterson and RachelSimmons are normally daily interactions.

4 –– The data for this challenge can be analyzed either as a static collection or as a dynamic stream of data, as it would occur in a real emergency. Describe how you analyzed the data - as a static collection or a stream. How do you think this choice affected your analysis? Limit your response to 200 words and 3 images.

Our system is not designed for dynamic stream of data, but with a little modification it can be used for it. Because the analysis methods and visualizations we designed are both time independent. Reflected in:

1. Analysis of data at a certain time or during a certain period of time does not depend on subsequent data;

2. We can analyze new dynamic data as increment without reprocessing the previous data. Obviously, analysis of dynamic stream data is needed in real-world application and it can help users focusing on the current problem they are facing. But it doesn't analyze data from a global view which may lead users into misunderstanding of the data and making wrong decisions. Static data analysis is more suitable for historical review from a global perspective and can give interpretation to some certain problem using subsequent data but it often miss the real-time of rescue.

In the analysis process provided by our system, we often choose a short time (1~3 hours) before the moment of interest as a reference for analysis, which is similar to the scene of dynamic analysis. Therefore, our current static designs won’t hinder us from simulating analysis of dynamic stream data.