Collecting and analyzing news from newspaper posts in facebook using machine learning

Abstract

Many people use social networks to spend their free time. News, especially at the time of great world changes, began to gain considerable popularity. Washington Post, New York Times, Time, Reuters, Forbes are among the most famous global newspaper publications. An average analyst can spend up to 40 hours a week collecting information about competitors and researching the most popular posts. According to the conducted research, an average of 40 new posts with news per day. The data processing process can be automated using modern information tools to facilitate the routine work of analysts. To analyze the target audience and reach, it is worth considering the text of the message, the number of likes, comments and links. This information was obtained using the Selenium automated web page testing tool using the Java programming language. The time spent on collecting data in the described way from four newspaper editions amounts to approximately 12 hours. The Tensorflow library using the JavaScript programming language is applied to the collected information. Based on information about the number of shares, comments, likes, frequency of news posts, an analysis was carried out using machine learning algorithms. Based on the clustering data, we can observe such a tendency that posts with a large number of likes receive a large number of comments and vice versa. An analysis of the most active hours of users in the network based on news posts is performed. As a result, the highest activity is observed at least three times a day, namely: in the morning hours from 9:00 to 11:00, in the lunch time of the day from 12:00 to 15:00 and in the evening time period from 18:00 to 20:00. This trend is due to the work schedule of most employees during the working week. The resulting statistical information in the work can be used for other content or user behavior in social networks.