Comments from Grace

[graceyqlin]

Hi Michael,

You talk about a very interesting topic: How to find the best fit for college applicants? It is an extremely important decision for both students and schools. By creating a match tool for students, they will be able to choose the schools which fit them academically and socially. It save people tons of time and efforts and can generate important information for schools to improve their student satisfactory as well.

You can keep the idea to get data from students' pyscho-social profiles and from schools satisfactory surveys. Create an algorithm to match students with their best fit schools.

Maybe you can explore and quantify more features, like the costs of the education, location, weather, and current student's ratings for engineering, social science, softball, swimming. In addition to conducting interviews with current students to get the information, you can use natural language processing techniques to scrape data from the schools' bbs or public media tools, like twitter or facebook, to analyze how they feel about their schools.

You can also create a recommendation tool to provide suggestions to students based on their historical academic and social performance. The data science knowledge you learn from Berkeley MIDS program hopefully can help you to customize this recommendation tool for this project.

Thanks for sharing!

Grace

[Michael-Diamond-Data]

Grace.

Thanks for your comments - yes there are some good additional thoughts in there about factors that I could bring in - cost of education, current student ratings etc. I'm thinking of at least two types of data, something like "stated" and "revealed" preferences. The hard factos or known-knowns like ratings, costs etc... are part of the stated facts.

The natural language processing is a fascinating additional idea, that i want to figure out how to incorporate. I don't know enough about it yet ... but sounds like could be very useful. Is this an area you have worked in at all. Are there key techniques I should start to explore. I would think data from the schools but also data from students at the schools that I could categorize as positive and negative and semantically parse, would be great.

The recommendation tool is a very practical idea --- I think this was implicit in my thinking, that ultimately this tool would be some kind of "College Finder" and would make recommendations.

Thanks so much