Sentiment analysis through tweet of airway companies with data mining techniques

As a result of the continuous development of the technology and internet, which are the inseparable parts of our lives, social media platforms have increased. There is vast amount of non-structural data in these social media platforms where people express their thoughts. Since this big data is non-structural and takes up a lot of space, it is impossible to analyze it, therefore; the notion of sentiment analsis has shown up. Sentiment analysis aims to negatively, positively or neutrally classify the emotions, thoughts and the feelings about the subject in data by analyzing them. Today, keeping monitoring customer satisfaction has become an obligation rather than being an advantage. Previously applied survey studies used for being able to evaluate the situation of customers, keeping up with advertisement campaigns, evaluating new products market values to get results requires a long process and not always accurate. For this reason, social media platforms that make customers express themselves sincere and clear gives better and more current results for large data and emotion analyses. At this thesis study, Turkish tweets written for two airline company is analyzed for certain dates at Twitter. During these processes, tweets were pre-treated and edited due to usual spelling errors. Edited tweets were classified as positive, negative, neutral; with classical classification methods; Artificial Nerve Network, Closest Neighbor, Bayes, Support Vector Machines and newest, the most popular method; Deep Learning. Deep Learning, which has gain popularity in the last years, is compared with classical classification methods in terms of analyzing results.