Prev
Predicting stock price movements (over 60 minutes) - INFORMS Data Mining Contest Results
Recently, INFORMS Data Mining Section held
a Data Mining Contest
which required participants to develop a predictive analysis solution that predicts is a stock price will increasing or decreasing in 60 minutes according to the data we have now (60 minutes before).
Eg: Will Apple stock price increase or decrease in 60 minutes (ex.:at 10:30) according to the data we have now (60 minutes before, ex.: at 9:30).
Competitors were provided with intraday trading data (609 explanatory variables) showing:
- Stocks price values (at five minute intervals)
- Sectoral data (at five minute intervals)
- Economic data (at five minute intervals)
- Experts' predictions (at one week intervals)
- Indices (at five minute intervals)
Traders, analysts, investors and hedge funds are always looking for techniques to better predict stock price movements. Being able to better predict short-term stock price movements is a boon in high-frequency context, so the methods developed in this contest have a big impact on the finance industry. Knowing whether a stock will increase or decrease allows traders to make better investment decisions, better understand what drives stock prices (supporting better risk management).
This is the overview of the contest:
- 28,496 visits on the competition website (kaggle.com/informs2010)
- 894 participants
- 147 submitted predictive analysis solutions
- 27 countries were represented (In order of most participants to fewest: United States, Colombia, India, Australia, United Kingdom, France, Thailand, Canada, Germany, Argentina, Japan, Afghanistan, Albania, Austria, Belgium, Chile, China, Croatia, Ecuador, Finland, Greece, Hong Kong, Iran, Poland, Portugal, Slovak Republic and Venezuela)
Overall ranking:
1) Cole Harris from DejaVu Team
2) Christopher Hefele from Swedish Chef Team
3) Nan Zhou from Nan Zhou Team
"Not using future information ranking":
1) Anuja Kokrady, Sweta Agrawal, Merin Varghese and Mahesh Kumar Tambi from Ams2009 Team
2) Jong-Seok Lee from Jumper Team
3) Piaomiao from Piaomiao Team
See: kaggle.com/informs2010?viewtype=results for full results.
Cole Harris (#1), Christopher Hefele (#2) and Nan Zhou (#3) presented the methods they used at INFORMS Annual Meeting.