Airlines based Twitter Sentiment Analysis Using Deep Learning

The airlines industry has been a competitive marketplace that has grown rapidly over the last few decades. Mostly, customers like family members, businessman, sportsman and youngsters are traveling through Airlines. If people are participating, their feedback is extremely important. Customer direct feedback may be favorable or negative, but understanding their Tweets is critical for improvement. This serves as a conduit for open-source inactive communication between promoters and customers coincidentally exercising their time and duties on the same platform for a variety of reasons. Sentiment Analysis on the social networking sites like Twitter or Facebook that bridge the gap between information and real time feedback, has become an amazing method for finding out about a user's feelings and has a wide scope of utilizations. It focuses on polarity (positive, negative, and neutral), sentiments and emotions (urgent, not urgent), and even intents (interested not interested). In this paper, the idea is to analyses the tweets emerging from social site such as Twitter, necessarily focused around the airline industry, its customers and employees, current as well as imminent. So ultimately, the objective is to deploy the deep learning algorithms on dataset of 14641 total tweets regarding U.S airlines, collected from Kaggle repository. The similar data set is utilized for both training and testing because there is more possibility for errors, which raises the likelihood of inaccurate predictions. Therefore, train_test_split function of scikit-learn python library has been used. It allows to breaking a dataset with ease while pursuing an ideal model. To prevent over fitting associated with the co-adaptation of feature detectors, the dropout learning algorithm has been called on to remarkable results.