A recent British study of people between the ages of 14 and 35 has shown that social media has a negative impact on mental health. The purpose of the paper is to detect people with mental disorders' behaviour in social media in order to help Twitter users in overcoming their mental health problems such as anxiety, phobia, depression, paranoia. The authors have adapted the recurrent neural network (RNN) in order to prevent the situations of threats, suicide, loneliness, or any other form of psychological problem through the analysis of tweets. The obtained results were validated by different experimental measures such as f-measure, recall, precision, entropy, accuracy. The RNN gives best results with 85% of accuracy compared to other techniques in literature such as social cockroaches, decision tree, and naïve Bayes.

A recent British study of people between the ages of 14 and 35 has shown that social media has a negative impact on mental health. The purpose of the paper is to detect people with mental disorders' behavior in social media in order to help Twitter users in overcoming their mental health problems such as anxiety, phobia, depression, paranoia, etc. For this, the author used text mining and machine learning algorithms (naïve Bayes, k-nearest neighbours) to analyse tweets. The obtained results were validated using different evaluation measures such as f-measure, recall, precision, entropy, etc.

No abstract

In today's digital world the email service has revolutionized the sphere of electronic communication. It has become a veritable social phenomenon in our daily life. Unfortunately, this technology has become incontestably the original source of malicious activities especially the plague called undesirable emails (SPAM) that has grown tremendously in the last few years. The battle against spam emails is extremely fierce. This paper deals with an intelligent spam filtering system called artificial heart-lungs system (AHLS) mimicked from the biological phenomenon of general circulation and oxygenation of blood. It is composed of different steps: Selection to stop automatically emails with undesirable identifier. Multilingual pre-processing to treat the problem of multilingual spam emails and vectoring them. Heart filter and lungs filter to classify unwelcome email in the spam folder and welcome email in the ham folder to present them to the recipient. The method uses an automatic updating of learning basis and black list, and a ranking step to order the spam mails according to their spam relevancy. For the authors' experimentation, they have constructed a new dataset M.SPAM composed of emails pre-classified as spam or ham with different language (English, Spanish, French, and melange) and using the validation measures (recall, precision, f-measure, entropy, accuracy and error, false positive rate and false negative rate, ROC and learning curve). The authors have optimized the sensitive parameters (text representation technique, lungs filters, and the size of initial leaning basis). The results are positive compared to the result of other bio-inspired techniques (artificial social bees, artificial social cockroaches), supervised algorithm (decision tree C4.5) and automatic algorithm (K-means). Finally, a visual result mining tool was developed in order to see the results in graphical form (3d cub and cobweb) with more realism using the functionality of zooming and rotation. The authors' aims are to eliminate a large proportion of unwelcome email, treated the multilingual emails, ensuring an automatic updating of their system and poses a minimal risk of eliminating ham email.

The popularization of computers, the number of electronic documents available online /offline and the explosion of electronic communication have deeply rocked the relationship between man and information. Nowadays, we are awash in a rising tide of information where the web has impacted on almost every aspect of our life. Merely, the development of automatic tools for an efficient access to this huge amount of digital information appears as a necessity. This paper deals on the unveiling of a new web information retrieval system using fireworks algorithm (FWA-IR). It is based on a random explosion of fireworks and a set of operators (displacement, mapping, mutation, and selection). Each explosion of firework is a potential solution for the need of user (query). It generates a set of sparks (documents) with two locations (relevant and irrelevant). The authors experiments were performed on the MEDLARS dataset and using the validation measures (recall, precision, f-measure, silence, noise and accuracy) by studying the sensitive parameters of this technique (initial location number, iteration number, mutation probability, fitness function, selection method, text representation, and distance measure), aimed to show the benefit derived from using such approach compared to the results of others methods existed in literature (taboo search, simulated annealing, and naïve method). Finally, a result-mining tool was achieved for the purpose to see the outcome in graphical form (3d cub and cobweb) with more realism using the functionalities of zooming and rotation.