Browsing by Author "Birim, SÖ"

Now showing 1 - 3 of 3
  • No Thumbnail Available
    Item
    Estimating Return Rate of Blockchain Financial Product by ANFIS-PSO Method
    Birim, SÖ; Sönmez, FE; Liman, YS
    Today, blockchain technology is developing rapidly and the volume of blockchain financial product trading is increasing rapidly as well. The aim of this study is to predict the return rates of cryptocurrencies with the help of artificial learning applications, considering the complex and unstable structure of the financial system. The rate of return is one of the important criteria used for investment decisions. Therefore, an efficient method for return rate prediction will help investors in preparing their portfolios. Ethereum, one of the top three most traded cryptocurrencies in the world, was chosen for empirical analysis. The adaptive neuro-fuzzy inference system approach (ANFIS) has emerged as a method that has been frequently used in recent years. ANFIS uses optimization algorithms to obtain the best prediction performance based on neural network modeling. The ANFIS approach has a multilayered structure consisting of many nodes inside and connections between the layers. ANFIS retains the properties of a fuzzy system while applying the principles of a neural network. Computations in the layers are conducted to learn and reproduce the information of the system. In this study, the particle swarm optimization (PSO) algorithm is used to train the ANFIS network. PSO aims to find the best-performing model in predicting the prices of three major cryptocurrencies that are Bitcoin, Ethereum, and Tether. The prediction accuracy of the proposed models was checked on the test set with performance indicators of root mean squared error (RMSE) and mean absolute percentage error (MAPE). The ANFIS-PSO approach gives strong results in cryptocurrency rate of return estimation.
  • No Thumbnail Available
    Item
    Social Sentiment Analysis for Prediction of Cryptocurrency Prices Using Neuro-Fuzzy Techniques
    Birim, SÖ; Sönmez, FE
    This study aims to provide an intelligent system that uses neuro-fuzzy techniques to predict daily prices of selected cryptocurrencies using a combination of twitter sentiment and google trend data. Although previously used to predict bitcoin prices, Neuro-fuzzy systems are used with this study for the first time with sentiment analysis to predict price trends of digital currencies. An adaptive neuro-fuzzy interface-based network was used to predict the prices of three selected cryptocurrencies, Ethereum, Ripple and Litecoin. The difference from the current study is that Twitter sentiment and Google trends have not been used as a predictor in a neuro-fuzzy network before. ANFIS has the advantage of combining the properties of fuzzy systems and neural networks. This advantage is manifested in producing lower error and higher accuracy in predictions. According to the findings, different results were obtained for different cryptocurrencies in the model in which the ANFIS estimation method was used. For ETH and LTC, the best forecast performance is obtained when twitter sentiment and google trends are used together. The Twitter sentiment model took second place by only a small margin. For XRP, only twitter sentiment shows the best forecast performance.
  • No Thumbnail Available
    Item
    Detecting fake reviews through topic modelling
    Birim, SÖ; Kazancoglu, I; Mangla, SK; Kahraman, A; Kumar, S; Kazancoglu, Y
    Against the uncertainty caused by the information overload in the online world, consumers can benefit greatly by reading online product reviews before making their online purchases. However, some of the reviews are written deceptively to manipulate purchasing decisions. The purpose of present study is to determine which feature combination is most effective in fake review detection among the features of sentiment scores, topic distributions, cluster distributions and bag of words. In this study, additional feature combinations to a sentiment analysis are searched to examine the critical problem of fake reviews made to influence the decision-making process using review from amazon.com dataset. Results of the study points that behavior-related features play an important role in fake review classifications when jointly used with text-related features. Verified purchase is the only behavior related feature used comparatively with other text-related features.