What Restaurants in Barcelona would you recommend?

Inspiration

We got inspired from Reinforcement Learning in the term of Data Exploration and Exploration in recommender systems. In the term of like or dislike a restaurant we got the snippet, so we can use the sentiment analysis, the state of the art.

What it does

From a score we get the user score, as it is explicit data. If we don't have the score we evaluate it with implicit data with Sentiment Analysis and get implicit data.

How we built it

We built a KNN with Means model using surprise python library, and a review sentiment analysis using transformers.

Challenges we ran into

Almost every user in the dataset has only one rating, so it's quite difficult to do accurate prediction based on user similarity, so we had to switch from user-user filtering to item-item.

Accomplishments that we're proud of

Giving all we can.

What we learned

We've studied and tried different techniques for Recommendation Systems, and learned that real data is not as easy to deal with as competitions data as Kaggle.