Fintech - Trading


Developing a Technical Indicator based on Twitter user Behaviour

Szymon Fonau, CTO @ OctoShrew


OctoShrew Technologies partnered with a prominent fintech company to address their need for sentiment analysis and price trend prediction using Twitter data. The collaboration aimed to develop a large-scale Twitter scraper capable of collecting tweets and retweets in real-time, conduct sentiment analysis, Named Entity Recognition (NER), Network Analysis, and integrate the results with cryptocurrency price changes for training a recurrent neural network (RNN) model. The RNN model would then serve as a technical indicator in the client's proprietary trading algorithm. The key novelty of the project was a new approach to identify high-impact users using Network Analysis based on a Graph Neural Network (GNN).

Problem Statement:

The fintech company faced challenges in collecting and analyzing a large volume of Twitter data for sentiment analysis, NER, and network analysis. They needed to correlate Twitter data with cryptocurrency price changes in real-time to train an RNN model for price trend prediction. Furthermore, the client aimed to integrate the RNN model as a technical indicator in their proprietary algorithm.

Solutions Provided by OctoShrew:

OctoShrew provided the following solutions to address said challenges:

Large-Scale Twitter Scraper:

OctoShrew developed a robust and scalable Twitter scraper that collected a high volume of tweets and retweets in real-time. The scraper utilized efficient algorithms and parallel processing techniques to ensure optimal performance.

Sentiment Analysis:

Advanced natural language processing techniques were implemented to analyze the sentiment of collected tweets and retweets. The sentiment analysis provided insights into public perception and sentiment towards cryptocurrencies, helping the client make informed decisions.

Named Entity Recognition (NER):

OctoShrew employed NER algorithms to extract relevant entities such as cryptocurrency names, prominent individuals, and companies mentioned in the tweets. The extracted entities facilitated further analysis and correlation with price changes.

Network Analysis:

Network analysis techniques were applied to identify influential users, communities, and patterns of information diffusion within the Twitter cryptocurrency community. This analysis provided additional insights for price trend prediction. This part is the key novelty in the solution found as prior sentiment-based algorithms are notoriously inaccurate as they are easily influenced by bots, users with limited following/financial understanding etc. Our network analysis allows us to identify users which function as consistent markers for change in public opinion, either due to their expertise or their popularity driving others' decision-making.

Real-Time Correlation with Price Changes:

OctoShrew developed mechanisms to capture real-time cryptocurrency price changes and effectively match them with the sentiment and NER results obtained from Twitter data. This correlation allowed for accurate analysis of the impact of public sentiment on cryptocurrency prices.

Recurrent Neural Network (RNN) for Price Trend Prediction:

Using the correlated Twitter data and cryptocurrency price changes, OctoShrew trained an RNN model. The model predicted price trends in real-time, serving as a valuable technical indicator for the client's proprietary algorithm, WHM Capital.

Results and Benefits:

The collaboration yielded the following results and benefits:

Accurate Sentiment Analysis:

The Twitter sentiment analysis conducted by OctoShrew provided the fintech company with accurate insights into public sentiment towards cryptocurrencies. This enhanced their decision-making process and improved their understanding of market dynamics.

Enhanced Market Analysis:

The integration of NER and network analysis allowed the fintech company to identify influential entities and track information diffusion within the cryptocurrency community. This enhanced their market analysis capabilities, enabling them to identify trends and make informed investment decisions.

Real-Time Price Trend Prediction:

The trained RNN model provided real-time predictions of cryptocurrency price trends. Integrating this technical indicator into the client's proprietary algorithm, allowed for more precise trading strategies and potential higher returns.


OctoShrew successfully collaborated with the fintech company to address their challenges in sentiment analysis and price trend prediction using Twitter data. The development of a large-scale Twitter scraper, coupled with advanced techniques such as NER and Network Analysis by GNNs, enabled the client to make accurate predictions and enhance their trading strategies. The integration of OctoShrew's solutions into the client's proprietary algorithm positioned them for success in the competitive cryptocurrency market. Future enhancements could further optimize the algorithm and expand the client's capabilities, solidifying their position as industry leaders in cryptocurrency trading.


Related blogs