-
logistic algorithm result
Inspiration
The yelp database is provided by Yelp company and the link by HACKATHON. With more users starting to complete order in mobile app like Yelp, the comments and ratings in Yelp becomes the key to judge whether a restaurant is good or not. However, the recent research fount out that many customers in App like Yelp might give bias ratings, which could lead to negative effect to enterprise in Yelp.
What it does
We are able to build the norm of tendency of users' personal information and enterprise's information deciding final ratings. Then with this norm, we can predict the rating in the future.
How I built it
We analyzed the six data set Yelp provided, and used three of them to build a fundamental relationship among user, business, and review. Then we classify the categories and ratings in these attributes in order to get a more explicit relationship among these three entities we mentioned before. With these decorated and cleaned data sets, we are able to use logistic algorithm in ten folds cross validation on this data to find the patterns.
Challenges I ran into
The database they provided is too large and raw. We have to re-classify some categories and clean many unnecessary data to grab the useful information from database.
Accomplishments that I'm proud of
The correctness rate calculated by logistic algorithm is high, which prove ours hypothesis. And the pattern we found could be used in Yelp company to help them give a better and much fair rates, which benefit both business and users.
What I learned
We learned how to deal with commercial company's data, and get more familiar with the process of analyzing data and using machine learning to find norms. During this process, we all find the way to work with each other and realize the importance of collaboration.
What's next for Rating predictions
Because we have a good result, we can develop a good algorithm to calculate a better ratings for Yelp. Besides, we also want to use some other technologies like text mining to dig more information from other tables like TIP and Photo
Log in or sign up for Devpost to join the conversation.