Evaluation of customer satisfaction about telecom operators in turkey by analyzing sentiments of customer through twitter

Objective Considering that the use of social media is increasing today, it has become that people share their feelings and thoughts through these social media platforms. For this reason, emotion analysis has gained sensitivity through these platforms where individuals quickly share their experiences. Especially, these platforms, where individual comments are concentrated, make companies evaluable from the perspective of the customer. Estimating customer satisfaction is important for companies. The purpose of this study, traditional machine learning algorithms and enriched them with embedded models using conventional deep learning algorithms, Turkey, which is one of the three largest operators Turk Telekom, Vodafone Turkey, and Turkcell through customer satisfaction prediction of company we aim to create important resources. As we know, the traditional machine learning for the operator company in Turkey over Turkish text algorithms and customer satisfaction is the first study conducted on behalf of estimation using the deep recessed models alongside the traditional learning model. Materials and Methods In this study, conventional word recessed models, deep learning algorithms and new generation word is the largest operator with the customer volume in Turkey using recessed models Turk Telekom, Vodafone Turkey and comprehensive experiments to estimate the satisfaction Turkcell customers are made. All the stated accuracies are an evaluation criterion used in experiments to show the classification performance of each model and the contribution of our work. It was aimed to improve the classification performance of the proposed model with the application of pre-processing methods. The combination of Turkish texts containing user comments with traditional machine learning models, word embedding models, deep learning models hybrid models and deep learning models preprocessing methods will be an advantageous choice to determine the sensitivity of users in directing their shares and to achieve the best classification success. The accuracy values between 80% and 97% obtained from the collected Turkish data set were demonstrated. However, traditional deep learning algorithms produced better results than the new generation word embedding models in data sets with Turkish text content. Results vi In this study, conventional word recessed models, deep learning algorithms and new generation word is the largest operator with the customer volume in Turkey using recessed models Turk Telekom, Vodafone Turkey and comprehensive experiments to estimate the satisfaction Turkcell customers are made. All the stated accuracies are an evaluation criterion used in experiments to show the classification performance of each model and the contribution of our work. It was aimed to improve the classification performance of the proposed model with the application of pre-processing methods. The combination of Turkish texts containing user comments with traditional machine learning models, word embedding models, deep learning models hybrid models and deep learning models preprocessing methods will be an advantageous choice to determine the sensitivity of users in directing their shares and to achieve the best classification success. The accuracy values between 80% and 97% obtained from the collected Turkish data set were demonstrated. However, traditional deep learning algorithms produced better results than the new generation word embedding models in data sets with Turkish text content. Conclusions In this study, conventional word recessed models, deep learning algorithms, and having a large volume customers for customer satisfaction experiments using a new generation of models to predict the word flush out the operators in Turkey has made customer satisfaction. Each accuracy specified constitutes an evaluation used in experiments to show the classification performance of the models and the additive output of our work. It was aimed to improve the classification performance of the proposed model with the application of pre-processing methods. Accuracy between 80% and 97% obtained from the Turkish data set collected from Twitter, where the combination with pre-processing methods, which is a new generation word embedding model for classifying Turkish texts containing user comments, will be an advantageous choice to determine the estimation of customer satisfaction and to achieve the best classification success. It has been demonstrated with its value. However, traditional deep learning algorithms produced better results than new generation word embedding models in data sets with Turkish text content such as news and analysis.