Yazar "Parlar, Tuba" seçeneğine göre listele

Listeleniyor 1 - 1 / 1
  • [ X ]
    Öğe
    A Study on Twitter User Gender Classification using Feature Selection
    (Kırıkkale Üniversitesi, 2022) Parlar, Tuba
    In today's business models, institutions or organizations want to know users’ opinions to improve their decision-making processes. Millions of people all around the world express their daily comments and thoughts using text messages, videos, or photos via social network applications. The rapid growth of social networking applications such as Facebook, Instagram, Twitter, and YouTube provides an attractive field for researchers to investigate the content of big data shared here and analyze user behavior. This enormous amount of data from social networks is used for effective marketing, personalized recommendation systems, finding opinion leaders, the pharmaceutical industry, or political policy making. A big amount of data obtained through social network applications is analyzed by machine learning methods. In this study, feature selection method is used to improve the automatic gender classification performance of Twitter users. The performance of the feature selection method that is applied on three datasets: user descriptions, tweets and where both are used together is evaluated with naive bayes and logistic regression classifiers. The results of the experiments show that the classification success of the selected features using chi-square feature selection method is much better with logistic regression classifier.