Twitter Attribute Classification With Q-Learning on Bitcoin Price Prediction

Abstract

Bitcoin price prediction based on people's opinions on Twitter usually requires millions of tweets, using different text mining techniques, and developing a machine learning model to perform the prediction. These attempts lead to the employment of a significant amount of computer power, central processing unit (CPU) utilization, random-access memory (RAM) usage, and time. To address this issue, in this paper, we consider a classification of tweet attributes that effects on price changes and computer resource usage levels while obtaining an accurate price prediction. To classify tweet attributes having a high effect on price movement, we collect all Bitcoin-related tweets posted in a certain period and divide them into four categories based on the following tweet attributes: (i) the number of followers of the tweet poster, (ii) the number of comments on the tweet, (iii) the number of likes, and (iv) the number of retweets. We separately train and test by using the Q-learning model with the above four categorized sets of tweets and find the best accurate prediction among them. We compare our approach with a classic approach where all Bitcoin-related tweets are used as input data for the model, by analyzing the CPU workloads, RAM usage, memory, time, and prediction accuracy. The results show that tweets posted by users with the most followers have the most influence on a future price, and their utilization leads to spending 80% less time, 88.8% less CPU consumption, and 12.5% more accurate predictions compared with the classic approach. 16 17 processing and application in the field of predicting future 49 Bitcoin prices. Processing a large amount of Bitcoin-related 50 tweets normally consumes a high level of computer resources 51 (CPU, RAM, memory) and time [31], [32], [33], [34]. Most of 52 the previous works is focused on how to reduce the resource, 53 so maximizing the prediction result at the same time is not 54 considered. However, tweets written by an expert, public fig-55 ure, or celebrity will become viral, with many replies, likes, 56 and retweets. Tweets with few replies, likes, or retweets are 57 unlikely to become viral because they are likely to circulate 58 mainly among close friends. Consequently, viral tweets are 59 expected to have a greater influence on price changes than 60 other tweets. If we can separate tweets with the highest 61 impact on future price changes from less important tweets, 62 it gives the possibility to employ less computer resources 63 usage while still obtaining accurate forecasts. 64 Hence, different from the previous approaches, in this 65 study, we analyze how Bitcoin-related information on Twitter 66 affects the actual Bitcoin price by considering four main 67 attributes: (i) the number of followers of the poster, (ii) the 68 number of comments on a tweet, (iii) the number of likes, 69 and (iv) the number of retweets. For this, we gather all 70 Bitcoin-related tweets within a particular period and divide 71 them into four groups based on their attributes. Since we 72 use th...

Multi-level deep Q-networks for Bitcoin trading strategies

“… Bitcoin historical data in strategy formulation Machine learning and technical analyzing algorithms 32 , time-series analytical mechanisms 33 , deep reinforcement learning 18 , LSTM 34 , ARIMA 35 , random forest, XGBoost, SVM, quadratic discriminant analysis, and LSTM 36 , Deep Q-network [This paper]. Twitter sentiment analysis for trading decisions Recurrent nets and CNN 37 , random forest 38 , linear discriminant analysis 39 , vector autoregression 40 , Q-learning 41 , logistic regression, Naive Bayes, and SVM 42 , Bullish tweet signals 43 , BERT and GRU 44 , FinBERT, CNN, and NLP 45 , Deep Q-Network [This paper]. …”

“…This subsection explores the role of Twitter sentiment analysis in developing effective trading strategies, highlighting how social media sentiment complements historical price data for a robust strategy. Numerous studies use sentiment data to predict near-future Bitcoin prices, suggesting that the public opinion expressed on social media significantly influences market trends and Bitcoin prices 37 – 41 . However, few studies have directly incorporated Twitter sentiment data into strategy development, which is a potential growth opportunity for cryptocurrency trading.…”

“…This study builds on the methodology presented in previous research, where Twitter sentiment data were employed to predict Bitcoin’s near-term price fluctuations 41 . The primary focus of this endeavor is to analyze the sentiments in tweets pertaining to Bitcoin, for a better understanding of community’s perceptions and potential market movements.…”

“…4 ). The comparative difference reward (CDR) function designed in our previous work 41 was adapted as the reward function. This is a unique reward function designed to provide more nuanced feedback to the model based on the accuracy of its predictions.…”

“…This is a unique reward function designed to provide more nuanced feedback to the model based on the accuracy of its predictions. The CDR function considers the rate of change in the actual Bitcoin price and introduces the concept of a zero-reward value, which was defined in 41 , as follows:…”