Abstract:

Abstract Bitcoin is steadily gaining popularity online and expanding its use in a wide range of transactions. The sentiments are the main driver of its pricing, which is very sensitive. Gains would increase as forecasting accuracy improves. Although there are many statistical methods for predicting prices, accuracy still needs improvement. This research aims to increase the forecast precision of Bitcoin price movements. The volume, polarity, and price variables of the data set have been added by including a new field called sentiment in order to achieve this goal. Twitter is used to determine sentiment, which is then included in the data collection. This work has suggested a dynamic sentiment analysis strategy that uses the long short-term memory (LSTM) to forecast prices with greater accuracy. Prior to and after the sentiment components were incorporated into the data set, prices were forecasted. The incorporation of mood has a beneficial effect on forecast accuracy, according to the findings. As a result, our approach can significantly reduce risk that is brought on by the high fluctuation of the price of bitcoin.

Keywords: