O processo de Feature Engineering e aplicação de aprendizagem de máquina para sistemas de recomendação de cervejas artesanais

Abstract

This research presents the process of feature engineering to add new features to data sets to help in machine learning models construction, which can be used to build recommendation systems of beers. In the last few decades, there was a trend between Brazilian consumers, in part due to the increases in the buying power, for high quality products, among them beers. In parallel to that, social networks has gain popularity, including social networks towards beer consumers and enthusiasts. Using data generated by these social networks, it is possible to construct data sets of beer reviews which can be used with machine learning models in order to build tools that can help in the beer consuming process. Based on that, this research makes a feature engineering process to construct new features in these data sets and experiment with Ordinal Logistic Regression models in order to verify the viability to build a recommendation system using these public data sets. In the results, it was possible to observe that clustering and dimensionality reduction helps these models to gain a big performance. At the end of this research, it was concluded that it is possible to build a beer recommendation system using these public data sets using Ordinal Logistic Regression models.