D by the data’s nonlinearity. Therefore, the efficiency in the MLP classifier significantly improved the accuracy on the predictive task. An exciting method focusing on the attributes is presented in [15]. The authors hypothesized that the title’s grammatical construction and also the abstract could emerge curiosity and attract GNE-371 Cell Cycle/DNA Damage readers’ consideration. A brand new attribute, known as Gramatical Score, was proposed to reflect the title’s capability to attract users’ focus. To segment and markup words, they relied on the open-source tool Jieba [58]. The Grammatical Score is computed followed the measures below: Each and every sentence was divided into words separated by spaces; Each word received a grammatical label; The quantity of each and every word was counted in all items; Finally, a table with words, labels, and the quantity of words was obtained; Each item receives a score with all the Equation (10), where gci represents the Grammatical Score on the ith item within the dataset and k represents the kth word inside the ith item. The n is the quantity of words within the title or summary. The weight would be the level of the kth word in all news articles, and count in this equation is the level of the kth word inside the ith item: gci =k =weight(k) count(k)n(10)Sensors 2021, 21,15 ofIn addition to this attribute, the authors employed a logarithmic transformation and normalization by constructing two new attributes: categoryscore and authorscore: categoryscore = n ln(sc ) n (11)The categoryscore is definitely the typical view for every category. The variable n in the Equation (11) represents the total number of news articles of every author. For every category, the data that belonged to this category were selected, and Equation (11) was utilised: authorscore = m ln(s a ) m (12)The authorscore is defined in Equation (12), where m represents the total number of news articles of every author. Prior to calculating the authorscore, data are grouped by author. For the prediction, the authors used the titles and abstracts’ length and temporal attributes moreover to the 3 pointed out attributes. The authors’ objective was to predict regardless of whether a news article could be well-known or not. For this, they utilized the freebuf [59] site as a information source. They collected the products from 2012 to 2016, and two classes were defined: popular and unpopular. As these classes are unbalanced and common articles will be the minority, the metric AUC was used, that is significantly less influenced by the distribution of unbalanced classes. Furthermore, the kappa coefficient was used, that is a statistical measure of agreement for nominal scales [60]. The authors chosen five ranking algorithms to observe the very best algorithm for predicting the popularity of news articles: Random Forest, Choice Tree J48, ADTree, Naive Bayes, and Bayes Net. We identified that the Guretolimod web ADTree algorithm has the most effective functionality with 0.837 AUC, along with the kappa coefficient equals 0.523. Jeon et al. [40] proposed a hybrid model for reputation prediction and applied it to a true video dataset from a Korean Streaming service. The proposed model divides videos into two categories, the initial category, called A, consisting of videos that have previously had associated operate, one example is, television series and weekly Television programs. The second category, referred to as B, is videos that are unrelated to earlier videos, as inside the case of movies. The model uses different qualities for each and every variety. For sort A, the authors use structured information from prior contents, including the number of views. For form B, they use unstruct.
