The dataset used is available at Stanford site
The project main idea is explained in Project Proposal
The project mid-term report is Mid-Term Report
The project final report is Report
This project explores the potential of the local clustering coefficient as a feature of popularity for books and genres. The clustering coefficient gives a strong metric of correlation between books bought together, so it can be used to determine a possible order of the ele- ments, obtained by analysing how often a book appears in a triangle. Given the values of the local clustering coefficient, we investigated the possibil- ity of approximating the salesrank order for the books. We also compare the order for the different genres that can be obtained using the salesrank and the clustering coefficient of the different books within the genre. Then we try to classify each book into 4 different categories based on a new joint definition of popularity that we came up with, with the objective of capturing the nature of the books.