Loading...

Investigation and Analysis in Social Media Data and Cryptocurrency Bubble Fluctuations

Mirab Samiee, Zahra | 2019

398 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 52831 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Fatahi Valilai, Omid; Haji, Alireza; Beigi, Hamid
  7. Abstract:
  8. Behavioral economics demonstrates that the sentiment can affect the behaviors and financial decisions of people. Can this be generalized to societies? For instance, is it possible for societies to experience emotions that will change their ways of decision making? This thesis attempts to perform an experiment addressing these questions. Bitcoin has been facing a transient phase of price volatilities. Many articles believe that bitcoin does not have an inherent value and is only derived by factors like the perception and acceptance of it and the sentiments among the investors. Digital currencies provide the unique possibility of measuring socioeconomic signals using digital traces.The emergence and growth of sentiment analysis were concurrent with the growth of social media on the web basis. For instance, online reviews, discussions in online forums, blogs, microblogs, Twitter and other social networks. This phenomenon, absorbed much attention as this was the first time in the history of humankind, that a large volume of data was being stored. Since the early, sentiment analysis turned into one of the most essential fields in natural language processing. This field is also being used extensively in data mining, web mining and text mining. These studies have been extended out of the field of computer science, and emerged in management and social sciences, due to their ever-growing importance in society. In recent years, business activities have also been plenty in the field of sentiment analysis. Concretely, sentiment analysis systems have found their application in every social and business matter.This thesis investigates whether taking Twitter sentiment into account can increase the prediction models' ability to predict the direction in which the price of Bitcoin is changing every day. In order to achieve this, a dataset from tweets has been gathered and a novel approach to classify the sentiments regarding the Bitcoin has been proposed. The results of this classification have been used in predictive models and the levels of precision had been investigated and compared with the literature. Among the predictive models, multilayer perceptron (MLP), support vector machines (SVM), Random Forests (RF) and Long-Short Term Memories have been implemented. This comparison takes both the utilization of market and sentiment analysis effectiveness into account
  9. Keywords:
  10. Bitcoin ; Sentiment Analysis ; Machine Learning ; Electronic Cash ; Price Variation

 Digital Object List

 Bookmark

No TOC