Mineração de dados em rede social para avaliação de tendências de consumo do queijo artesanal no Brasil

Abstract

The development of alternatives to traditional market research has been a majorchallenge in corporate environments. Such research has a scientific basis, is well validatedand brings good results, however, factors such as geographic limitations, long time ofapplication and analysis, high cost, in addition to having forms that can be extensive andoften with low adherence can make their use unfeasible. With so many people using themost diverse social networks, companies are increasingly attentive to what is commentedon in these virtual environments. With so many messages, posts, comments and audios, alarge volume of data is generated and, therefore, the term Big Data in this environmentis very present referring to the large volume, variety and speed with which the dataare generated. Knowing the consumer profile and consumption trends can help a lot inmaking strategic decisions in the most diverse sectors and commercial segments. Thatsaid, the present work refers to the development of a computational system capable ofcollecting, storing, processing and extracting information in data from the social networkTwitter, with the objective of evaluating consumption trends through the identification ofconsumption characteristics and habits of artisanal cheese in Brazil. For this, initially a setof keywords was developed referring to the names of the artisanal cheeses of interest andan algorithm capable of collecting this data on Twitter. Then, it was necessary to build adatabase model capable of efficiently storing the newly collected data and the results ofother analyzes. To identify the polarity of the feelings present in the tweets, a model wasdeveloped using the Voting Classifier Ensemble to perform the classification of publicationsin the negative, neutral and positive classes. It was necessary to create a specific trainingdata set for artisanal cheeses, since no literature that met these specifications was foundin the literature. Natural Language Processing techniques were used to treat textualdata and extract information from them. A set of keywords was developed based onthe opinion of researchers at EMBRAPA Dairy Cattle, representatives of the AssistanceTechnical and Territorial Extension Company (EMATER) and the analysis of the mostfrequent words in tweets about artisanal cheese, capable of identifying the characteristicsand consumption habits of the artisanal cheese. With the analysis of the opinion expressedby the consumer on Twitter’s social network about artisanal cheeses in Brazil, it wasfound that defining a general trend for all artisanal cheeses was a practical task due toissues such as regionalism, characteristics and consumption habits vary for each type ofcheese, a fact that could be observed by analyzing the consumption profile of Coalhocheese. However, taking into account the results obtained, the computer system developedproved to be capable of combining several techniques and methodologies in one place, making decision-making more efficient based on the analysis of real data, thus innovatingthe way of doing traditional market research.