Classification of customer complaints of the telecommunications industry on social media messages

Social media has become a communication and broadcasting platform where emotions, thoughts and opinions are frequently expressed. Expressing emotions through social media requires great care as an important part of our lives. In addition to sharing ideas and thoughts, we also share good memories. Social media such as Facebook, Twitter and LinkedIn are recognized as rich sources of opinion-rich text data. Both organizations and individuals use social media platforms to analyze people's opinions and extract emotions and feelings. Most studies on social media analytics classify sentiments into positive, negative or neutral classes. The difficulty in sentiment analysis lies in the fact that people can express one or several emotions in a single phrase. One can recognize these different emotions well, but it is still not easy for a sentiment analysis system. Telecommunications refers to a field that uses communication technologies to transmit communication signals such as information, voice, images and data. These technologies cover a variety of communication tools such as voice telephones, mobile phones, the internet, radio, television, satellite communications and other communication systems. Telecommunications facilitates the sharing of information across distances between people. Communication signals can transmit in different formats, such as voice, images and data. For example, voice telephones allow people to talk to each other, while data such as text, photos and videos can be shared via the internet. Telecommunications are also of great importance in commercial and industrial areas. Businesses communicate with their customers over the telecommunications infrastructure, share data transfer and manage business processes. Telecommunications are also used in many sectors such as health, transportation, media, education and finance.With the developing technologies, there are continuous innovations in the field of telecommunications. For example, new technologies such as wireless communication technologies (Wi-Fi, Bluetooth), fiber optic infrastructure, 5G networks have made telecommunications faster, more reliable and more comprehensive. With this development, consumption increases and the situation of providing the best service among telecommunication sectors is evaluated. When consumers do not receive the required service and experience problems, they express this on social media platforms and create a data corpus on social media platforms. In this study, the dataset includes a corpus of complaints about the telecommunications sector collected from Twitter. Out of three million tweets, five thousand tweets were labeled. After the preprocessing and cleaning process of the labeled data was completed, classification was performed with GRU, LSTM, N-Gram GRULSTM models. The findings obtained as a result of classification were discussed and evaluated with tables and figures. In terms of error values, it was found that LSTM and N-gram LSTM models performed better than GRU-based models. This is attributed to the success of LSTM-based networks in capturing long-term dependencies.