guohong630 / aml_project Goto Github PK

#AML FINAL PROJECT @hong & peng

Team name and logo wanted ...

• stock price prediction

##Feb 8

###work done:

• Topic: financial prediction
• routine: improve based on reference projects

###task in comming week: read reference projects and find out:

• what data?
  • price or RoE?
  • size?
  • accessible?
• what model?
  • liner-regression
  • svm
  • ann ...
• what trick?
  • deal with time-correlation? compared to image processing (receptive field?)
  • using any tech index?
• how to present result, any benchmark?

###issues to discuss with TA or instructor:

• language
• external packages
• grading criteria
• due date

###Work Done:

Papers Review I:

• Goal: to study and apply as many Machine Learning Algorithms as possible on a dataset for stock market

• Assumption: EMH

• Methods: Fundamental vs Technical

• what model:

• Support Vector Machines

• Linear Regression

• Decision Stumps

• Expert Weighting (Assume a best expert?? or WMA??)

• Online Learning (SVM online? )

• Boosting??

• what data?

• prices on EMA

• only two companies

• downloaded from the yahoo finance section.

• what feature?

• date + price

• what trick?

• deal with time-correlation? use EMA

• using any tech index? EMA

• Boosting + SVM

• how to present result, any benchmark?

• EMA as comparision? binary

Papers Review II:

• Assumption: EMH
• Methods: Fundamental vs Technical
• Goal:
  • The project will concentrate on building an automated trading platform which would con- sume web services and have a database back-end
  • research AI techniques involving ANNs and genetic algorithms to forecast financial time series data.
  • investigate the applicability of different neural networks architectures to forecast currency exchange rates movements on Forex market.

• Read paper II and compare with paper I for implemenations
• schedual meeting with Instructor on Thursday