Analyzing the impact of political tweets on exchange rates

---

Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Marmara Üniversitesi, Fen Bilimleri Enstitüsü, Türkiye

Tezin Onay Tarihi: 2021

Tezin Dili: İngilizce

Öğrenci: AKINER ALKAN

Asıl Danışman (Eş Danışmanlı Tezler İçin): Ali Fuat Alkaya

Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu

Özet:

Social media has been increasingly popular and valuable along with their mass data. At the same time, currency exchange rate forecast has been an important topic for researchers, analysts, and investors for a long time. In this study, we have combined exchange rate time series analysis and Twitter sentiment analysis to build machine learning models. We have built the models in three stages for a six-month period: (i) we have watched financial and political hashtags and applied sentiment analysis on the tweets retrieved from these hashtags, (ii) we have collected time series data on cross-currency exchange rates including cryptocurrencies, (iii) we have optimized the model to forecast USD / TRY with the data we have. We have experimented with several machine learning algorithms including linear regression, Bayesian ridge, support vector machines along multi-layer perceptron (MLP). It has been observed that in this novel approach, some regression algorithms performed better than MLP. Computational experiments showed that our approach gave 0,32% mean squared error performance at its best. Results suggested that sentiment analysis is a helping factor to forecast currency exchange rate and Twitter is a good data source due to its mass and interactivity. In conclusion, investors, analysts, and researchers can benefit from the usage of our proposed model and will be able to get strong and consistent results to forecast the USD / TRY currency exchange rate.