Investigating Graph-Based Recommendations Systems and Graph Traversal Algorithms

Abstract

A recommendation system is used by most if the industries nowadays as it helps in increasing the sales in all. This is due to the personal touch a customer gets because of the recommendations. A recommendation system is a technique that helps in recommending products and other services to consumers using information filtration. Recommendation systems can be of many types a few of them being, collaborative filtering-based recommendation systems, content-based recommendation systems, utility-based recommendation systems and more. During this study we will be using collaborative filtering-based recommendation system along with graph-based learning techniques like Gensim Word2vec and GraphSage. For the collaborative filtering part we will take help of a correlation matrix and a pivot table to get the relationship between the users and products and eventually finding the relationship between different users. And in the graph learning models we will be using the word embeddings created using the graphs made using the gathered data and for the traversal of graphs we will be using deep walk and random walk algorithms, and then using this information we have recommended the products to the users. The assessment of each model is done on the basis of Personalization method.