Kday is Tuesday Weekday is Wednesday Weekday is Thursday Weekday is Friday Number 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 Feature Weekday is Saturday Weekday is Sunday Is Weekend Title Polarity Title Subjectivity Tasisulam Protocol Description Polarity Description Subjectivity Rate of Negative Words in Description Rate of Good words inside the Description Price of Constructive Words amongst non-neutral inside the Description Price of Negative Words among non-neutral in the Description Average of Damaging Polarity amongst words inside the Description Maximum of Unfavorable Polarity among words in the Description Minimum Unfavorable Polarity among words in the Description Average of Optimistic Polarity amongst words in the Description Maximum of Constructive Polarity amongst words in the Description Minimum Optimistic Polarity amongst words in the Description -6.four. Word Embeddings Word embeddings are dense low-dimension real-valued vector representations for words that are discovered from information. Their aim should be to capture the semantics of words so that similar words have a comparable representation within a vector space. Utilizing word embeddings, a single can anticipate not to depend on the attribute engineering stage, which often demands study and prior expertise on the content material to be predicted. In addition, if there’s no knowledgeSensors 2021, 21,27 ofabout the texts to become analyzed, it’s attainable to receive important predictive capabilities. As a counterpoint, we have the disadvantage of losing the interpretability in the characteristics. To collect the word embeddings in the title and descriptions, we use Facebook’s fastText [94] library for Python, which already comes with a pre-trained model for the Portuguese language. Their algorithm is based around the work of Piotr et al. [20] and Joulin et al. [95]. For each title and description, we very first eliminate the quit words. Then, we run the fastText library and BSJ-01-175 Purity & Documentation acquire a vector of 300 dimensions for the texts. 6.5. Classification The reputation of content may be the connection in between an individual item along with the customers who consume it. Popularity is represented by a metric that defines the amount of users attracted by the content, reflecting the on the web community’s interest within this item [8]. Taking a look at the “most popular” videos or texts on the internet, the notion of recognition is intuitively understood. Having said that, it’s essential to define objective metrics to compare two products and define which 1 could be the most well-known. Quite a few measures point out which content material attracts one of the most consideration on the net: the number of users prepared to consume the item searched. Within this operate, we are going to use the quantity of views as a reputation metric. The decision of machine understanding models to conduct the classification activity took into account the work carried out by Fernandes et al. [10] that selected one of the most utilised models within the researched literature. In addition, we group ML models into distance-based models (KNN), probabilistic models (Naive Bayes), ensemble models (Random Forest, AdaBoost), and function-based models (SVM and MLP). Within this way, our option tried to cover all these categories for comparison. We use six classifiers to establish irrespective of whether a video will grow to be well-known or not just before its publication: KNN, Naive Bayes, SVM having a RBF, Random Forest, AdaBoost, and MLP. We performed 5 experiments to evaluate the effectiveness of these models. Inside the very first experiment, we used only the 35 attributes obtained from Attribute Engineering as presented in Section 6.three. Within the second, we made use of the vectors obtained with all the f.