An unsupervised approach to creating a restaurant recommendation system

Over the last few years, recommender systems have become increasingly popular due to the technological advancements occurring in the fields of data mining, predictive analysis, and machine learning. In this, we try to use an unsupervised learning approach to cluster restaurants that are similar and recommend restaurants to users according to the ones that they have ordered from in the past. The models used in this project gave us average results given that clustering techniques have their problems and are not the most novel architectures available. However, working on a real-life data set lets us take a deep dive into the nuances of mining data and making predictions in big projects taken up by companies. DBSCAN seemed to work better than K-Means given the latter is very susceptible to outliers. Motivation to use DBSCAN is, this method is able to represent clusters of arbitrary shape and better to handle noise. However, the models that we have built still need a lot of work such as hyperparameter tuning, cross-validation, etc. to increase their accuracy and to be used in real life.