Due to the short and simple way of expression on social media platforms such as Facebook and Twitter, millions of people share daily real-time opinions about everything in an informal way due to the use of short language (slang) and emoticons, which generates an increasing availability of unstructured and yet valuable information to data science researchers. Traditional approaches such as paper-based surveys are not the wisest path for collecting and studying consumer behavior because they are time-consuming which leads to considerable losses for companies around the world. In this paper, we develop a hybrid system to identify and classify sentiment represented in an electronic text from Twitter where users post real-time reactions about everything to improve the decision-making process for companies. To do so, we used tweepy to access Twitter's Streaming API, we combined natural language processing techniques with Naive Bayes to classify users data, we used the Python library Matplotlib to display the results. The purpose of this paper is to propose an efficient and accurate approach for predicting sentiment from raw unstructured data in order to extract opinions from the Internet and predict online customer's preferences, which could be valuable and crucial for economic and marketing researchers.