![Different machine learning algorithms are used to predict the stock market trading.
Use text from different sources and use Text and Data Mining (TDM) to extract pattern or
information or any hidden data of interest to predict the Ups and downs of the targeted
stocks.
Then
Data Mining Isn't a Good Bet For Stock-Market Predictions [2]
Aug. 8, 2009 - JASON ZWEIG , Wall Street Journal
Now
How Traders Are Using Text and Data Mining to Beat the Market [3]
Feb 12 2015 - Market Roy Kaufman , The Street
Applying Machine Learning to Stock Market Trading - Bryce Taylor [1]
Machine learning algorithm to read headlines from financial news magazines and
make predictions on the directional change of stock prices after a moderate-length
time interval
[Stanford Student project 2013, CS 229]
Introduction](https://image.slidesharecdn.com/stockmarketpredictionusingdatamining-shivakumar-soppanavar-151110225342-lva1-app6892/85/Stock-market-prediction-using-data-mining-2-320.jpg)
![Conclusion
Sophisticated model able to beat overall market trends by reading financial news
headlines cannot be easily found without fairly sophisticated human-like processing
of the headlines. –By Author
Examples:
Tweet on Credit card breach at Home Depot (HD) -> Stocks 2% down. (9/2/2014) [3]
Nate Silver's uncannily accurate predictions of the U.S. national elections. (2012) [3]
Yes, by using Text and Data Mining and superior algorithms in near future, we may be
able to predict the stock market with greater accuracy.](https://image.slidesharecdn.com/stockmarketpredictionusingdatamining-shivakumar-soppanavar-151110225342-lva1-app6892/85/Stock-market-prediction-using-data-mining-8-320.jpg)
Stock market prediction using data mining
- 1. Stock Market Prediction Using Data Mining By Shivakumar Soppannavar CMPE 239 Under the Guidance of Prof. Eirinaki Magdalini 11/10/2015
- 2. Different machine learning algorithms are used to predict the stock market trading. Use text from different sources and use Text and Data Mining (TDM) to extract pattern or information or any hidden data of interest to predict the Ups and downs of the targeted stocks. Then Data Mining Isn't a Good Bet For Stock-Market Predictions [2] Aug. 8, 2009 - JASON ZWEIG , Wall Street Journal Now How Traders Are Using Text and Data Mining to Beat the Market [3] Feb 12 2015 - Market Roy Kaufman , The Street Applying Machine Learning to Stock Market Trading - Bryce Taylor [1] Machine learning algorithm to read headlines from financial news magazines and make predictions on the directional change of stock prices after a moderate-length time interval [Stanford Student project 2013, CS 229] Introduction
- 3. Data Sources and Research question Twitter data to predict stock market changes Change in management, M&A Intermittent headlines to react to the first headlines up or down ??? Data sources: Headlines from financial analysts http://seekingalpha.com/ Historic stock prices http://www.nasdaq.com/ 7 targeted companies IBM, NFLX, GOOG, ANF, MCD, SHLD, AAPL Research Questions: “Given a headline released today about some company X, will the stock price of X rise by more than P percent over the next time period T?” T= 3 months
- 4. Bayesian Classifiers Bayesian Classifier Simple multinomial Bayesian classifier that analyze the headlines based on the presence of each token in the headline 51202 tokens -> Laplace smoothening -> 693 tokens -> Top 10 tokens Classification Error for Reduced features < 0.5 Precision/Recall Increase in P increases the Positive error and decrease in Negative error Support Vector Machines SVM (Polynomial, linear, etc) was used on reduced data set, didn’t beat the result obtained from Bayesian classifier
- 5. Naïve Baye’s Testing Error Table 1: Bayesian classifier result run for top 10 most indicative symbols
- 6. Few more ways of analysis! Natural Language Processing Stanford has a publicly available Natural Language Processing Toolkit that provides sentiment analysis to sentences with high accuracy (>80%) Use of NLP didn’t achieve high success Natural language processors would need to be specifically tailored to processing headline-like data to be able to make a meaningful contribution towards answering my research questions. Principal Component Analysis Principal component analysis are run on the data and then tested linear SVMs on several of the top principal components. Manual Key word Selection Keywords are selected manually
- 7. Few more ways of analysis, Results Principal Component Analysis Manual Key word Selection
- 8. Conclusion Sophisticated model able to beat overall market trends by reading financial news headlines cannot be easily found without fairly sophisticated human-like processing of the headlines. –By Author Examples: Tweet on Credit card breach at Home Depot (HD) -> Stocks 2% down. (9/2/2014) [3] Nate Silver's uncannily accurate predictions of the U.S. national elections. (2012) [3] Yes, by using Text and Data Mining and superior algorithms in near future, we may be able to predict the stock market with greater accuracy.
- 9. Thank you
- 10. References 1. B. Taylor. (2013). “Applying Machine Learning to Stock Market Trading”. Retrieved from Stanford CS229 project lists 2013. http://cs229.stanford.edu/proj2013/Taylor- Applying%20Machine%20Learning%20to%20Stock%20Market%20Trading.pdf 2. JASON ZWEIG , (Aug. 8, 2009). Retrieved from Wall Street Journal website http://www.wsj.com/articles/SB124967937642715417 3. M. R. Kaufman,(Feb 12 2015). Retrieved from The Street website http://www.thestreet.com/story/13044694/2/how-traders-are-using-text-and-data- mining-to-beat-the-market.html 4. http://cs229.stanford.edu/projects2013.html
Editor's Notes
- #3: Text mining is the data analysis of natural language works (articles, books, etc.), using text as a form of data. It is often joined with data mining, the numeric analysis of data works (like filings and reports), and referred to as "text and data mining" or, simply, "TDM.“ [3]
- #5: https://en.wikipedia.org/wiki/Laplacian_smoothing Support vector machines (SVMs) are supervised learning models with associated learning algorithms that analyze data and recognize patterns, used for classification and regression analysis.