Your organization ICM is investigating a criminal conspiracy. Investigators are very confident because they know several members of the cabal, but they hope to find other members and leaders before they are arrested. The mastermind and all the people who may be suspected of complicity work for the same company in a big office, and the relationship is complicated. This company has developed rapidly and is famous for developing and selling computer software suitable for banks and credit card companies. ICM recently received a message from a group of 82 workers, who thought it would help them find the most likely candidates for unknown conspirators and leaders in the company. Because information circulation involves all workers working in the company, it is likely that some (perhaps many) identified communicators in this information circulation do not actually involve conspiracy. In fact, they are sure that they know some people who are not involved in the conspiracy.

The goal of modeling work is to determine who is the most likely accomplice in this complex office.

The priority list is ideal because ICM can investigate, * *, and/or ask the most likely candidates based on it.

It will also be beneficial to draw a line between non-conspirators and conspirators, because people in each group can be clearly classified.

It would also be very helpful to the prosecutor's office if the leaders of the conspiracy could be nominated.

Before giving the current data to your crime modeling team, your boss gave you the following information (called investigation EZ), which was a case when she was working in another city a few years ago. She is very proud of her work in a simple case. She said this is a small and simple example, but it can help you understand your task.

Her data are as follows:

The ten people she thinks are accomplices are Anne #, Bob, Carol, Dave *, Allen, Fred, George *, Harry, inez and Jayne #. (* means known accomplice, # means known non-accomplice).

She numbered 28 news records according to the theme on which the analysis was based.

Annie said to Bob, Why are you late today? ( 1)

Bob said to Carol, This damn Annie always looks at me. I'm not late. ( 1)

Carol to Dave: Annie and Bob are arguing about Bob's lateness again. ( 1)

Dave said to Allen, I want to see you this morning. When can you come? Bring the budget document by the way. (2)

Dave said to Fred, I can see you anytime and anywhere today. Let me know when is better. Do I need to bring budget documents? (2)

Dave said to George, See you later-there's a lot to talk about. I hope everyone else is ready. Did you do it right? Very important. (3)

Harry said to George, you look nervous. What's going on here? Don't worry, our budget will be fine. (2)(4)

Inez said to George, I'm really tired today. What about you? Are you okay? (5)

Jaye to Inez: Not so good today (? )。 How about going to lunch today? (5)

Inez to Jaye: Fortunately, everything is calm. I'm too tired to cook lunch today. Excuse me! (5)

George said to Dave, come and see me now! (3)

Jay said to Annie, are you going to have lunch today? (5)

Dave said to George, I can't go. I'm going to see Fred now. (3)

George said to Dave, call me when you see him. (3)

Annie to Carol: Who will supervise Bob? He idles around all day. ( 1)

Carol said to Annie, let him go. He works well with George and Dave. ( 1)

George said to Dave, this is very important. Damn it, Fred. How's Allen? (3)

Allen to George: Have you talked to Dave? (3)

George to Allen: Not yet. What about you? (3)

Bob said to Annie, I'm not late. Besides, you know I work at lunch time. ( 1)

Bob said to Dave, tell them I'm not late. You know me. ( 1)

Allen to Carol: Contact Annie to arrange the budget meeting schedule for next week, and help me appease George. (2)

Harry to Dave: Have you noticed that George looks nervous again today? (4)

Dave said to George, Damn Harry thinks you're nervous. Don't worry him, lest he ask around. (4)

George to Harry: I just work too late and have some problems at home. Don't worry, I'm fine. (4)

Allen said to Harry, I forgot about today's meeting. What should I do? Fred will go. He knows the budget better than I do. (2)

Harry said to Fred, I think next year's budget will put some people under a lot of pressure. Maybe you should spend some time today to reassure everyone. (2)(4)

Fred said to Harry, I think our budget is normal. I don't think anyone will be under pressure. (2)

End of communication record.

Your boss pointed out that she only assigned five different information topics and numbered them:

Bob is late,

2) Budget,

3) Important unknown problems, which may be conspiracy,

4) George's pressure,

5) Social issues such as lunch.

As can be seen from the message coding, some messages have two themes according to the content.

Your boss analyzes the case according to the communication network constructed by communication contacts and message types. The following figure is a message network model, and the code of message type is indicated on the network diagram.

Your boss said that in addition to known accomplices George and Dave, according to her analysis, Allen and Carol were also considered accomplices. And it didn't take long for Bob to confess that he was really involved, hoping to get a reduced sentence. The charges against Carol were later dropped.

Your boss is still sure Inez was involved, but he never filed a lawsuit against her.

Your boss suggested that your team find out the guilty parties, so that people like Inez would not escape the net and people like Carol would not be framed, thus increasing ICM's credit and making people like Bob have no chance to get a reduced sentence.

Current case:

Your boss has built the status quo into a database similar to the network, with the same structure as above, but with a larger scope. Investigators have some clues that a conspiracy is misappropriating the company's funds and using online fraud to steal the credit card funds of customers doing business in the company.

She showed you a small example of a simple case, only 10 people (nodes), 27 sides (news), 5 topics, 1 suspicious/conspiracy topics, 2 confirmed criminals and 2 known innocents. So far, this new case has 83 nodes, 400 faces (some involving 1 multiple topics), 2 1000 words of message records, 15 topics (three of which have been regarded as suspicious), 7 known criminals and 8 known innocents. These data are given in the attached spreadsheet files: names.xls, Topics.xls, Messages.xls

Names.xls contains the employee names corresponding to the key nodes in the office.

Topics.xls contains the code name and short description of the 15 topic.

Due to security and privacy issues, your team will not have all the direct message records.

Messages.xls provides the node pairs for transmitting messages and the topics of the messages (there can be multiple topics, up to three topics).

In order to make the exchange of information more intuitive, Figure 2 provides a network model of employees and message links.

In this case, the subject of the message is no longer displayed, as shown in 5438+0 in Figure 65. Instead, the number of topics is given in the Messages.xls file and described in Topics.xls.

Requirements:

Claim 1: Up to now, it is known that Jean, Alex, Elsee, Paul, Yao and Harvey are criminals, while Darren, Tran, Jia, Erin, Garde, Chris, Peggy and Esther are not criminals. Yes, the message topics are 7 1 1 and 13. For more information about this topic, please visit Topics.xls

Establish a model and algorithm, sort according to the possibility that 83 nodes are conspirators, and explain your model and indicators. Jerome, Dolores and Gretchen are senior managers of this company. If any of the three of them are involved in the conspiracy,

This will be very beneficial.

Requirement 2: There will be mysterious changes in the priority list. What if new information tells us that the theme 1 is also related to conspiracy, and Chris is a conspiracy? (that is, two more clues)

Requirement 3: A powerful technology for acquiring and understanding text information like this message circulation network is called semantic network analysis; As a method of artificial intelligence and computational linguistics, it provides a structure that can be used for reasoning about knowledge or language. Another computational linguistics related to natural language processing is text analysis.

For our case solution, explain: if you can get the original news, how helpful and reinforcing will it be to help your team develop a better model and classify office personnel by analyzing the context and content of information flow semantically and textually?

Did you use these functions to improve your model according to the topic description in the Topics.xls file?

Requirement 4: Your complete report will eventually be submitted to the prosecutor's office, so be sure to explain your assumptions and methods in detail and clearly, but it cannot exceed 20 pages. You can include your programs as attachments in a separate file, so that your paper will not exceed the page limit, but it is not necessary to include these programs. Your boss hopes that ICM is the best organization in the world to solve white-collar and high-tech conspiracy crimes, and hopes that your method will help solve important cases around the world, especially those databases with very large message flow (there may be tens of thousands of messages and millions of words). In particular, she asks you to discuss in your paper how a deeper analysis of network, semantics and message text can help your model and suggestions.

As part of your report to her, please explain the network model technologies you use, and why and how to use them in any type of network database to identify, prioritize and classify similar nodes, not just criminal conspiracy and message data. For example, given various images or chemical data, which show the infection probability and some identified infection nodes, can your method be used to find infected or diseased cells in biological networks?