Sunday 09 March 2025
Researchers have made significant strides in developing a system that can accurately classify the sentiment of Persian tweets related to cryptocurrencies. The study used a combination of natural language processing techniques, machine learning algorithms, and deep learning methods to analyze over 4,000 tweets collected from Twitter.
The researchers began by pre-processing the tweets, removing punctuation marks, English words, and numbers. They then converted Arabic letters to Farsi using a Persian library. Stop words were also removed, as they do not carry much meaning in the sentence but are used only to comply with the structure of the sentence.
Next, the researchers applied three different methods for feature extraction: bag-of-words (BOW), FastText, and BERT. The BOW method uses a word frequency approach to represent text documents as vectors. FastText is an extension of Word2Vec that assigns separate vectors to each character in a word. BERT is a language model that has been pre-trained on a large corpus of text data.
The researchers then applied three machine learning algorithms – support vector machine (SVM), k-nearest neighbors (KNN), and AdaBoost – to classify the tweets into positive, negative, or neutral sentiment. They also used two deep learning models – LSTM and BERT – to analyze the tweets.
The results showed that the deep learning models performed better than the classical machine learning algorithms. The BERT model achieved an accuracy of 83.50%, followed closely by the FastText+LSTM model with an accuracy of 82.32%. The SVM, KNN, and AdaBoost models had lower accuracies ranging from 74.5% to 77.5%.
The study highlights the importance of using machine learning and deep learning techniques in analyzing sentiment analysis in Persian tweets related to cryptocurrencies. The results demonstrate that these methods can be effective in extracting features and classifying sentiments accurately.
The implications of this research are significant, particularly for investors and financial institutions that rely on social media data to make informed investment decisions. By analyzing the sentiment of tweets related to cryptocurrencies, researchers and analysts can gain valuable insights into market trends and public opinion.
The study also underscores the need for more research in natural language processing and machine learning in Persian language processing. The results show that even with pre-processing techniques, the accuracy of the models is still affected by the nuances of the Persian language.
In the future, researchers may explore other applications of this technology, such as analyzing sentiment in other languages or using it to predict market trends.
Cite this article: “Sentiment Analysis of Persian Tweets on Cryptocurrencies Using Machine Learning and Deep Learning Techniques”, The Science Archive, 2025.
Natural Language Processing, Machine Learning, Deep Learning, Sentiment Analysis, Persian Tweets, Cryptocurrencies, Twitter, Feature Extraction, Text Classification, Language Modeling.
