1 The Problem
To develop a price prediction AI, that would go through Tweets and predicting future price movement based on a sentiment of a tweet. Besides Twitter, it should also be analysing news that are crypto-related, and predicting the price based on the sentiment in the article.
2 Data Sources
- www.twitter.com
- Require keywords to search tweets [Client input]
- To extract tweets
- News Websites
- Require websites [Client input]
- To collect news for bitcoins
- Bitcoin Historical Prices
- Require website [Client input]
- To collect historical bitcoin prices
3 Methodology
- Data Collection
- Collect data from all source
- Clean and transform data for analytics ready
- NLP
- Perform NLP to extract keywords from the tweets and news articles
- Extract key drivers / topics
- Compute and save NER, topics over time
- Prepare independent variables from both sources
- Hypothesis
- Independent variables are variables computed from tweets, sentiments, news articles, NER, and deep learning models
- Dependent variable is Historical Bitcoin Prices
- Predictive AI
- Train the models with independent variables and dependent variables
- Test the model
- Validate the model
- Predictive AI Tool
- Prepare functions to collect, clean and process the data to build the independent variables in real time
- Predict the bitcoin prices in Realtime
4 Deliverable
Predictive AI tool
5 Programming
Python