Search

Content generated by users on commercial social networks about products and brands generates large volumes of data that can be transformed into relevant and useful recommendations for marketing decisions. Every day, consumers post their opinions online on social networks about products they have purchased and used, and companies are increasingly interested in tracking this information in real time for better decision making. The main problem is to extract key information from consumers' textual comments and use it automatically to measure the quality of products or brands. In this work, we propose a hybrid approach to automatically analyze these reviews, assigning a quantitative score to negative and positive user content. The analysis of online consumer sentiment has increased significantly in recent years, being crucial to determine the success of businesses in a wide range of sectors, tourism, hospitality and e-commerce. In the same context, this work proposes a framework for analyzing the sentiment of reviews posted on the Twitter network towards products and brands. The first step is the construction of a dataset by collecting a set of reviews posted online on Twitter, processing and cleaning the textual data for better accuracy, and then developing a hybrid approach for product evaluation and polarities creation using lexicon-based methods and machine learning-based analysis techniques.

Many floods had pounded several Moroccan regions and had claimed a heavy toll of human life and immense damages; the most terrible event is the one that destroyed the area of Ourika in the surrounding of Marrakech City in 1995. We aim through this study to propose an effective flood forecasting and warning system (FFWS) for better management of flooding situations based on hydrology sensors for data gathering and least square support vector regression algorithm (LSSVR) optimized with PSO (particle swarm optimization) for monthly flood prediction. Since we expect that the future river flow values are heavily controlled by the past ones, as well as no autocorrelation in the residuals is required by the least square method that we use in this study, the model accepts time series lagged river flow values in the input and predicts river flow time series in the output. The smart proposed system allows to collect and predict the hydraulic situation in Aghbalou catchment and then forecasts whether a flood should be expected. The obtained results disclose the great quality of prediction of the LSSVR-PSO model with 0.8707 of R-squared score using a specific combination of cross-validation dataset.

Social network sites (SNS) play a role of paramount importance in the political landscape and influence the elections' outcomes around the world in an enormous way. As cases in point as for these phenomena, the world has witnessed the 2016 US presidential elections, the Brexit referendum, the yellow vests movement in France and the boycott campaign which took place in Morocco two years ago. In this vein, the citizens of all countries, especially the young ones, have been using social media to share and to get political information as well as to express their opinions, trends and contestations relating to economic, social and political issues as the habitual militating's ways are nowadays unable to make the citizen's expectations heard. The contents and networks structure of parties' candidates and politically engaged users have been used by many researchers to extract political insights that could help to understand the political events developments and to predict their results. Our contribution aims to explore the multiple techniques applied by the different approaches, examine their results in terms of accuracy and political insights and suggest a standard approach that will rely on the advantages of these contributions and standardize the processes of extracting and analyzing the political discourse which could help to decipher the enormous democracy's challenges raised by social media.

At present, the modernization of public administrations, fostered by technological progress, is leading to a rapid growth in services and information exchange between public administrations (across borders and sectors). Therefore, the heterogeneity and the interoperability of e-government systems represent a real challenges in facilitating these information exchanges, it is then necessary to find a solution to overcome these challenges. The aim of this article is to propose a novel design of government data warehouse based on the ETL process and Service Oriented Architecture. We describe a framework that provides a comprehensive set of interoperability services to help organizations collaborate flexibly and deliver better quality of services.

International audience; Aim/Purpose: In this paper, we highlight the need to monitor and diagnose adaptive e-learning systems requirements at runtime to develop a better understanding of their behavior during learning activities and improve their design. Our focus is to reveal which learning requirements the adaptive system is satisfying while still evolving and provide specific recommendations regarding what actions should be taken and which relevant features are needed to help meet the specified learning requirements.Background: Adaptive e-learning systems research has long focused on user modeling and social learning to personalize each learner experience, while fewer instruments are reported to assess the quality of the solutions provided by such adaptive systems and to investigate their design problems. The design problems may emerge due to ever-evolving requirements being statically specified at design stages and to the changing environments that can be difficult to control and observe. The combination of some or all of these factors can lead to a definition of inconsistent or insufficient adaptation rules, which in turn may prevent these systems from providing appropriate resources to learners even if the needed ones have been accounted for within the knowledge space.Methodology: An empirical study has been performed to check and validate the behavior of a real-world adaptive e-learning system under four stated requirements. The study used a novel monitoring and diagnosing tool that reads the collected data from the system and checks its behavior against constraints that are derived automatically from the requirements specification.Contribution: The results provide statistical insights and highlight some issues related to requirements compliance at runtime; which helped us detect unforeseen instructional design issues.Recommendations for Practitioners: The study suggests that diagnosing requirements compliance at runtime can be an essential means to increase the confidence in their adaptive e-learning systems capabilities at runtime.Recommendation for Researchers: The study suggests that further research for developing specific indicators related to requirements compliance is needed in the field of adaptive e-learning systems.Future Research: Future work includes the study of possible improvement of our diagnostic tool using probabilistic reasoning.Runtime requirements; requirements compliance; adaptive e-learning system; learning analytics; goal modeling; adaptive e-learning system; evaluation.

Real-time tilt sensing in urban areas can provide crucial information regarding the instant evolution of sensitive assets of the city to better safeguard the sites sustainability and security of people. In this paper, we proposed a novel approach based on the accelerometer sensor to recognize tilt angle between the vertical plan (the light poles along the road, the transmission tower, crossing bridges, etc.) and the ground. Indeed, we used the tri-axial accelerometer embedded in a Wireless Sensor Network (WSN) platform manufactured by Libellium, named “Waspmote Starter Kit”. The collected data are sent from one sensor node to another until the gateway and then to the processing center. Our recognition method uses the calculation and mapping of the values sensed by an accelerometer to a corresponding angle. The z-component corresponding to the gravity allows recognizing the real-time tilt of the horizontal plane while the x and y-components indicate in which side the plan is bent. This approach is applied to the smart city to provide an effective solution for real-time electric transmission pole monitoring. A computer program processes the angle results of the mapping and plots the 3D scene on the monitoring screen reproducing the behaviour of electric pole remotely located. The experimental results of our approach are satisfactory because the average error that can be committed by comparing the calculated and measured angle is about 1.0.

Our research study is related to data Mining software and analyzing big Data especially in public administration. The purpose of this study is to compare softwares and algorithms used for statistical analysis. Following the comparative study of supply chain management (SCM) softwares, we chose the best to make a comparative study of descriptive and predictive algorithms using dataset collected from the new IT solution developed in our PHD research study.