search

w201rdada / portfolio-Michael-Diamond-Data

portfolio-Michael-Diamond-Data created by GitHub Classroom
0 stars 0 forks source link

Comment From Li #3

Closed aassdd654 closed 7 years ago

aassdd654 commented 7 years ago

Hi Michael,

I was surprised by the fact of 30% transferring rate and realized that there is a solid need from future students may face a tough decision process in a non-transparent world.

In the problem part: I feel that your idea is trying to solve 2 things: 1) how to make easier to gather school profile in multiple dimensions, something similar to "user review"

I feel If the school is the one to collect the data(in proposed solution part), then there will be the bias from the schools when presenting the data. Meanwhile, It seems that there is no unified place to let the existing student give reviews on their colleges in general, I feel the data collection is a major challenge here.

2) how to judge the student is fit or not fit to a certain school, something similar to "like/dislike" button.

From my experience around, the "Fit or Not Fit" feelings are combined with many factors, some of them may be not directly from the college. While new students are trying to fit into the college, they are also trying to fit in a bigger environment: the new city, room share, the new weather, leaving parents, leaving friends, all of these factors may add up to "the fitting feeling" measurement of the new college life. So maybe add some broader variables for these non-college factors?

Looking forward to your idea in the real world!

Thanks, Li

Michael-Diamond-Data commented 7 years ago

Li.

Thanks for these comments, and the direct conversation we had. I have been thinking about ways I can incorporate third party reviews into the algorithm and/or as you suggested let some third-party administer the testing and evaluation of existing college students - so there is no obvious bias in the results drawn by the colleges.

The broader variables is clearly where a lot of the unpredictable, but potentially dispositive, impact will be. Will have to think hard aobut how to get that in. I was hoping to capture that sort of thing twice: first, as a sort of brute force filter on top, ie one of the things you can select on is big city vs. small city vs. campus town vs. rural etc etc, ie a way to narrow down the search; but secondly I could see if these factors are actually predictive, somehow in determining the strength of the relationship by types of people with specific colleges.