Predicting the popularity of books before publication using machine learning

• Main Authors: Sachdeva, Hansika, Puri, Ujjwal, Poornima, S.
• Format: Conference Proceeding
• Language: English
• Subjects: Datasets; Machine learning; Performance prediction
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0217437

The publishing industry is highly competitive, and predicting the popularity of books before their publication can significantly benefit publishers and authors alike. Publishers tend to put a lot of resources into all the books they choose to publish. If the book ends up being unsuccessful or does not perform as well as it was supposed to, they lose the time and resources spent on that book. There is no measure that can validate the editors’ decisions about books and predict their popularity. In our paper, we propose a novel approach for predicting the popularity of books before publication using the text of books as the primary data source. We investigate the connection between two independent data sets and a book’s popularity. The first one is the various textual features of book manuscripts, and the second is the metadata of the book, like the topic, the genre, and the author. To validate the proposed approach, we have used a large data set of published books and their corresponding ratings and rankings.