مهرناز تقي وند

With the wide availability to web contents and the huge amount of News produced, the need to explore this subject in its various dimensions and as a result to find knowledge in these types of data has also increased. In this context, one of the dimensions to study is to analyze and represent new approaches to predict the popularity of news content. Various approaches have been presented to solve the problem of predicting news popularity; that considering the time of data availability, the type of features under investigation, and also the type of the target variable; These approaches can be classified into different types. According to the category of features used and attention to the features related to the users' feedback regarding the news, it suggests two categories of prediction before the publication or after the publication of the news. Also, the expected output from popularity prediction can be expected as categorization or the goal of prediction can be expected as an estimation of the exact value of the target variable, which also divides this issue into two categories of classification and regression. This research targets the prediction of news popularity by using news features before its publication and also in a category with two popularity classes and seeks to find methods to increase the accuracy of news popularity prediction; This goal allows more optimal news content to be offered. The main motivation is rooted in the high volume of news content and increasing user demand for quick and easy access to information. It is also worth mentioning that solving this problem is beneficial for news agencies and beneficiary people in this business. Because knowing the popularity of each news item is effective in cases such as the position of the news on the site and advertisements. The purpose of this research is to review past works and compare various methods in order to find a method that has the ability to improve the accuracy of predicting news popularity compared to similar works. In this eva‎luation, several models of machine learning models are examined. A combined learning model consists of two feed-forward neural networks, the first feed-forward neural network is designed with the ability to memorize the model and the second neural network is designed with the ability to generalize in the model, which provides the highest prediction accuracy equal to 83%.

محبوبيت خبر , محتواي خبري آنلاين , پيش‌بيني محبوبيت

news popularity , online news content , Popularity Prediction

Mehrnaz Taghivand

Dr. Hassan Naderi