Döviz kuru tahmini için Türkçe tweetler ile twitter duygu analizi

In developing countries such as Turkey, foreign exchange rates are important in business and daily life decisions. It is important for decision-makers to predict exchange rates accurately and act within the scope of these predictions. Since exchange rates are complex and volatile, it is very difficult to accurately predict exchange rates, and many studies have been carried out in this context in the past. The newest trends in the literature are the research which are performed with machine learning methods. In this thesis study, it was aimed to determine whether more accurate predictions can be performed with the sentiment analysis method or not. Within the scope of the study, the value of Dollar, Sterling and Euro against Turkish Lira was predicted with sentiment analysis of more than 168,000 Tweets related to these exchange rates from Twitter. Sentiment analysis results and exchange rate values were analyzed by the support vector regression method. The time limits of the study were determined as 1 March 2022, starting from 1 November 2019, the beginning of the COVID-19 pandemic outbreak. In order to compare the results of the sentiment analysis method, the same data were predicted with support vector regression, naive forecasting and random walk models, and these data were compared with the values made by sentiment analysis. The results revealed that Twitter sentiment analysis increase the accuracy rate of exchange rate predictions and there is a positive and strong relationship between currency prices and negative Tweets, and there is a negative relationship between currency prices and positive tweets. Also, there is positive relation between number of Tweets posted and currency price.