Data warehouse building to support opinion analysis in social media

Social media have changed the world in the last decade. In fact, the rapid and vast adoption of these platforms had an impact on human behavior, in particular, and society in general. All these channels of media are indispensable for our daily life because they enable us to easily communicate, interact and share a variety of content, with one another. In addition, companies have been interested in this phenomenon to discuss and share products with their customers, improving their production from reviews posted on these platforms. However, analyzing and understanding the volume of generated data from social media requires the use of data warehouse methods and techniques to discover new knowledge to assist companies in the decision-making process. In this research paper, we propose an approach for data warehouse construction for opinion analysis, which can help decision-makers to make decisions by analyzing the users' opinions from the data warehouse. In this paper, we are interested more specifically, in the designing of a data mart from social media accounts, such as Facebook, Twitter, and YouTube. Second, we define a set of semantic relations between the multidimensional concepts of data marts to generate the global schema of the data warehouse by integrating three data mart schemes. Then, we proposed a method of opinion analysis from social media based on machine learning techniques to identify the polarity of the users comments posted on social media, and finally, we are focusing on the implementation of our data warehouse on the NoSQL database to make the analysis task easier for the decision-maker.