Course talk:CPSC522/Inactive Cookie Mapping via Trail Matching

Hi Dandan, Thank you for your informative page. I learned a lot regarding how ads are generated on user browsing trends. I have the following feedback which I believe would be helpful:

• A lot of the ideas on the page were new for me. Some external links regarding those would be of great help
• The Hypothesis and your contribution were not clearly stated, though it is understood once the page is read through. Explicit mention of your hypothesis and contributions would be a great help
• The results section could use a bit more addition. A analysis of how algorithm 2 does better than algorithm 1 would be particularly enlighting.
• I was a bit confused about the concept of noise avoidance. Are infrequent visits totally ignorable. Is there any particular reason why we believe they are just noise? Some insights regarding that would be appreciated.

Thank you for your hard work and such an informative page.

Hi Mehrdad Ghomi,

1. I added some more detailed background information at the beginning of the page. 2. Also highlighted the hypothesis. 3. I added a 10-round test to compare how algorithm2 performed better than 1. 4. Those noises are temporary visit sets, which can not be formed as part of the pattern because those are far less frequent than the real pattern trail.

Bests, Dandan

Hi Dandan,

It is quite a interesting page! It really gives me an insight into how ad providers can quickly presenting advertisements relating to what I have just browsed. You designed a very good experiment to compare two types of tracking algorithm.

Only few suggestions, It might be better to add one sentence summary to help readers quickly understand what this wiki page is going to talk about, so does abstract section. And it might be better to clearly state your hypothesis.

Best regards,

Jiahong Chen

Hi Jiahong,

Thank you for your advice. I have added more information at the beginning of the page and also highlighted my Hypothesis.

Bests,

Dandan

Hi Dandan,

Interesting work! Here is my feedback:

General

• One-sentence-summary, Abstract, Builds on, and Related pages are all stubs. Don't forget to add that content.
• There are grammatical errors throughout the page that make it much harder to read; proofreading is highly recommended.
• It can be useful to readers to highlight important terms (through bold text, for example).

Problem and Issues

• I'm not exactly sure what the problem is that you're trying to solve. There were indications in the previous section, but it should be stated explicitly in this section as well. The text needs to be more clear; proofreading should help with this.
• You state there are no existing algorithms to solve the problem, but that you will compare two algorithms. Did you come up with these algorithms yourself?
• You state "for each user ID t in T stores data in a table..." - who or what is storing the data? Also, user IDs are called t in one sentence and then l in the next. Is there some reason for this distinction? At present, these issues make the section less clear.
• Is your hypothesis that Trail Matching performs better than the other algorithm? It would be good to make it more clear in the organization of the page (eg. have a "Hypothesis" section).

Algorithm 1

• So it seems at this point that the problem to be solved is matching old users with new users; is this correct?
• How is the transformation from T^B to T^C and T^D done?
• You have a T^C_inactive. Should it be T^D_inactive?

Algorithm 2

• In the new tables C and D, you have IP addresses under the heading "IP pattern". Does "pattern" here indicate simply that there are repeated visits? If not, what part of the data indicates the pattern itself?
• I'm having trouble understanding your pseudocode:
  • there may be some indentation issues with the for loops
  • I'm not seeing arrayActive referenced anywhere inside the algorithm.
  • "item" is not very descriptive
  • "a" is being used as an element in multiple nested for loops
  • where do "itemA" and "itemB" come from, and why are you removing them?
  • what is the condition for the second break?

Conclusions and Future Research

• Can you elaborate on how the algorithms have probabilistic aspects?

Hi JordonJohnson,

Thank you very much! Your feedback is so precious for me, and I would definitely add some more details on my page to make it easier to understand and interesting on the next version. And I am still working on different experiments to compare the two algorithms.

Thank you so much again.

Bests, Dandan

Hi Dandan,

An interesting wiki page! And here are two suggestions: 1. I suggest you clean the page, delete some unnecessary part, for instance, the Abstract section. 2. I did not understand the hypothesis until I read the Algorithm 1. Therefore, I suggest that you should state the theory clearly at the beginning of a page, maybe right under your name.

Sincerely,

Junyuan Zheng

Hi Junyuan,

Thank you for your advice. I will add more detailed background info about what the problem is and give a general description about both the hypothesis and the theory at the beginning.

Bests, Dandan