Artificial Intelligence Research

[cliffconway]

Hi All,

I am current involved in some research into the use of AI systems for the selection of stocks.

Currently I have itemised the following data/information to enable me to build comparative models:-

1) End of Day prices
2) FTSE 30 company fundamentals
3) Main Sector cycle information

Item one is obviously the easy one and I have already source an adequate supply of 7 years of data. However, items 2 and 3 are the areas that I am struggle to find appropriate sources of historic data (the more the merrier).

Can any one suggest possible sources.

Many thanks in advance,

Cliff

 

[Pat494]

Hi All,

I am current involved in some research into the use of AI systems for the selection of stocks.

Currently I have itemised the following data/information to enable me to build comparative models:-

1) End of Day prices
2) FTSE 30 company fundamentals
3) Main Sector cycle information

Item one is obviously the easy one and I have already source an adequate supply of 7 years of data. However, items 2 and 3 are the areas that I am struggle to find appropriate sources of historic data (the more the merrier).

Can any one suggest possible sources.

Many thanks in advance,

Cliff

Obviously not knowing the ins and outs of your research project but as the FTSE is largely imho governed by the US market I would have suggested the S&P 500 or Dow Jones. A lot of company info is freely available at Yahoo.
Are you thinking of using neural networks or adaptive algorithms ?
best of luck

 

[Grey1]

Hi All,

I am current involved in some research into the use of AI systems for the selection of stocks.

Are you looking into application of expert systems or neural net technology ?

Grey 1

 

[Trader333]

Personally I have not found any AI approach that has been consistently of use for trading purposes.


Paul

 

[cliffconway]

Obviously not knowing the ins and outs of your research project but as the FTSE is largely imho governed by the US market I would have suggested the S&P 500 or Dow Jones. A lot of company info is freely available at Yahoo.
Are you thinking of using neural networks or adaptive algorithms ?
best of luck

Thanks for the replies guys,

I am looking at building various types of neural networks (MLPs, Kohonen and RBF types) and knowledge based systems (with and without fuzzy reasoning) and i am hoping to do a detailed comparative study.

 

[new_trader]

Thanks for the replies guys,

I am looking at building various types of neural networks (MLPs, Kohonen and RBF types) and knowledge based systems (with and without fuzzy reasoning) and i am hoping to do a detailed comparative study.

This sounds like an incredibly interesting study. But I doubt that "AI" will be better than real intelligence when choosing stocks for a long term investment portfolio (Assuming that is the purpose of the study.)

 

[cliffconway]

This sounds like an incredibly interesting study. But I doubt that "AI" will be better than real intelligence when choosing stocks for a long term investment portfolio (Assuming that is the purpose of the study.)

My instincts are probably telling me the same thing as your doubts. However, i do hope to improve my knowledge/intelligence along the journey.

 

[cliffconway]

Are you looking into application of expert systems or neural net technology ?

Grey 1

I am looking at a comparative study of various types of neural net and expert systems.

 

[cliffconway]

Personally I have not found any AI approach that has been consistently of use for trading purposes.


Paul

Its interesting to hear that you have travelled this road before. Do you have any info you would be willing to share on :-
1) what types AI techniques have you used?
2) have you developed/trained the models yourself?
3) even though you have not found any consistency have you found any avenues that would benefit from further research?
4) Do you have any links for sites with useful informaton on this topic?

Aplogies for the barrage of questions!!

 

[blackcab]

I am looking at building various types of neural networks (MLPs, Kohonen and RBF types) and knowledge based systems (with and without fuzzy reasoning) and i am hoping to do a detailed comparative study.

This is potentially a mammoth, multi-dimensional exercise. What do you want to achieve with this? Even just typing a list of possibilities is daunting!

 

[charliechan]

google Static Order Book Imbalance (SOBI)

this is the capstone of many computer based trading engines (both ai and non)

 

[Tuffty]

I have been experimenting with a neural net. Don't know if it's my human neural net that is at fault but I just seem to use it like an advanced correlation program looking for dependencies. The problem with an NN is that you end up with a black box ..... I think thats why I'm just looking for dependencies. Any thoughts.

 

[cliffconway]

This is potentially a mammoth, multi-dimensional exercise. What do you want to achieve with this? Even just typing a list of possibilities is daunting!

Hi,

This is part of a research project related to an honours degree (I have about a year to complete). So initially I am hoping to get a good grade for my project work. However, I am also hoping that I will learn a bit more about the application of AI techniques and also about different strategies that people follow in their pursuit of trading success.

If you or anyone else has a favourite strategy that they want to put through the works let me know the precise details of the strategy and I will attempt to model and test it some time over the coming year.

 

[Tuffty]

cliffconway,
This paper may be of interest www.aimlearning.com/Per.Anton.Thesis.pdf

 

[cliffconway]

I have been experimenting with a neural net. Don't know if it's my human neural net that is at fault but I just seem to use it like an advanced correlation program looking for dependencies. The problem with an NN is that you end up with a black box ..... I think thats why I'm just looking for dependencies. Any thoughts.

Yes this is one of the disavantages of using NNs - you end up with a black box (although sometimes it is possible to analyse the box to discover which inputs are not having an effect (or a lesser impact)on the outcome and remove them from the equation).

However, i am sure that if someone produced a black box, that improved the chances of success in trading, we would not object to using it on the basis that it was a black box.

 

[Tuffty]

However, i am sure that if someone produced a black box, that improved the chances of success in trading, we would not object to using it on the basis that it was a black box.

I couldn't use it as I need to understand what it is doing for me to have confidence in it....so for me at least the way I'm using an NN seems to be the only way to go. The link I posted above for genetic programming I found interesting as it gives you the 'black box' formula so you know what the equation is for the answer. Also genetic programming (or at least that program above) only gives you a TRUE of FALSE output which for me, at least is fine for what I'm doing.

 

[Pat494]

Hi,

This is part of a research project related to an honours degree (I have about a year to complete). So initially I am hoping to get a good grade for my project work. However, I am also hoping that I will learn a bit more about the application of AI techniques and also about different strategies that people follow in their pursuit of trading success.

If you or anyone else has a favourite strategy that they want to put through the works let me know the precise details of the strategy and I will attempt to model and test it some time over the coming year.

I have just bought an AI programme which uses adaptive swarm technology and invites customers to write their own swarm bots with Visual Basic. There is some help within the programme. The 2 buts are
1. learning Visual Basic
2. buying the programme - US $575 Biocomp.com ( Dakota )
It comes with some bots already in place too.