For this, the algorithm reads the popularity of the book on other sites.
Researchers from northwestern University, the research division of Microsoft in India and the Indian Institute of technology in Kharagpur have developed a model that can predict the success of a book on the biggest shopping on the Internet-Amazon, analyzing the behavior of readers on the Goodreads platform.
Machine learning methods are often used to predict any process. In essence, this class of artificial intelligence methods: the peculiarity of these algorithms is that they are trained in the process of solving the many problems. In the case of work to predict what book will become a bestseller, the algorithms use training by precedents, that is, the characteristics of the readers ‘ behavior.
The authors note that the popularity of the book depends on many factors and can be measured using several parameters. But in a particular study they focused on how the books readers most often prefer and how to read them. So the researchers took the necessary data from the platform Goodreads and tried to link them with sales books on Amazon.
First, the developers have analyzed the collective behavior of users on Goodreads. Then they identified the characteristic features of the works that have become bestsellers. The researchers noticed that the ratings and reviews of books on Goodreads is not as effective in prediction as compared to data on the status of reading the book each individual user. On the platform Goodreads this data tracking is especially easy, as there are readers to share information about how many pages of the works they have read, comment on the book and so on. After data collection and analysis the researchers developed a model to predict the success of books using machine learning methods.
The model achieved accuracy of 88.72%. This is 16.4 percent higher than the baseline methods, considering only traditional measures of popularity such as ratings of books or reviews.
© 2018, z-news.link. All rights reserved.