Cryptocurrency Price Prediction Model And Application Using Sentiment Analysis

Nowadays,due to the rapid development of blockchain technology and wide acceptance of cryptocurrency,cryptocurrency has become a trendy investment opportunity.At the same time,market sentiment can have great impact on the price movement of cryptocurrency.Short texts platform is commonly used by amateur investors to share their ideas and express attitude and emotions.For the information in the market,people tend to pay attention to various mainstream market platforms as their investment basis.In this thesis,we take BTC(Bitcoin),the representative cryptocurrency,as an example to study the relationship between its historical price trend and investors' sentiment,and predict its future price.For BTC price prediction,the following work is done in this thesis: 1)The 3,000 investors with high influence are collected as data samples,supplemented with various technical indicators and daily price data of dimension,and stored in a database.2)Various sentiment classification methods are designed to determine the sentiment polarity of each text,including the modified VADER sentiment dictionary method and deep learning method.After labeling the data by selecting the best model,a dataset with sentiment labels is finally created.3)A price prediction model incorporating sentiment value is constructed.Firstly,the traditional time series forecasting method was used to predict the price of BTC by combining various features such as price,volume and capital rate,and then the sentiment value was added on this basis for further study.4)A set of visual dashboard was built,including various technical indicators trend,sentiment weathervane,price prediction and other functions.To verify the effectiveness of the method in this thesis,two and a half years of historical data were used as the training set for the experiment,and the last six months of data were used for testing.The experiment proves that the sentiment index can reflect the sentiment attitude of investors to a certain extent,and the prediction accuracy of the model is improved compared with the univariate time-series prediction model,and its prediction results can bring informative suggestions to investors.This thesis provides relevant researchers with an idea to correlate public opinion with sentiment,and also provides a new method for the calculation of sentiment value.