What it does
It can predict the quality of a restaurant in four different aspects based on the reviews and make suggestions to improve the restaurant based on these scores. It can also suggest an area to open a new restaurant given it's characteristics based on the location on other similar restaurants in order to avoid competition while staying in a profitable area.
How we built it
We classified words according to their correlation with the rating of the reviews and the aspects of the restaurant they are related with. This allows us to construct a rating for each category based on both the ratings in which it is mentioned and what words are used when it is being commented about. Given a hypothetical restaurant that we wish to open somewhere in Barcelona and predictions about its characteristics we consider the location of restaurants of the same kind and, taking into account their similarity in the different aspects we measure, we locate the ideal spot to open this hypothetical restaurant.
Challenges we ran into
Our lack of knowledge about machine learning and other data science techniques limited our tools to process data to basic statistics and resources we could find on the internet.
Accomplishments that we're proud of
We managed to make everything we wanted work as intended, even those things that used techniques we did not fully understand or had worked with before.
What we learned
We have learnt to organize data large enough to not be readable and extract meaningful information and techniques that help in this tasks, the concept of a Word2Vec neural network was specially interesting.
Log in or sign up for Devpost to join the conversation.