Saturday, March 28, 2009

MSIAD - Stock Market with Data Mining (BI Thesis)

I was a little absent this month in preparation for my thesis ... I promise to write more often now! Yesterday was my first battle on my dissertation, which I think it was well reached. Naturally, any research in this complex world of stock markets, in this case of applying data mining techniques, requires first understanding the financial world... I tried to demystify the complex stock market world... His movement, the factors that influence decision making, forecasting and the investor that feeds this whole game markets... and it was presented yesterday

After understanding the stock market as I did, I‘ll now focus on the application of Data Mining techniques (initially only with neural networks) and the implementation of a strategy that will be based from all rules made public by some researchers. That is, I will create a decision support system that in addition to track my portfolio of shares, it’ll generate the final output of the day that will be a buy, sell or hold signal.My strategy will assume the loss and also fix a limit on earnings. I do not need to be monitoring the stock market during the day, because I will use Stop Loss functionality in all investments that fire automatically an order of sale if the price of some share will excess my defined limit (applied only if I’m on the market)

In this matter, I'll then begin to examine in more detail the research of Dr. Chong from Chinese University of Hong Kong, that he’s the only economist, investor and researcher I know at the moment and who kindly offered me help and sent me all Articles that he has done. Dr. Chong is in the photo at right. Check his website at :

I already have done ETL processes to import historical data from a group of stock in my folder and a Datamart that stores the data in the form of information. It’s time to deepen into practice and return in May for a short presentation to be evaluated again. I assume the risk of this project that it’s very ambitious and attractive to me.

With nothing to lose! Regards!

Pedro

13 comments:

Vitor Silva said...

muito interessante.
quando podemos ler alguma coisa do teu trabalho?

Pedro said...

Oi grande vitor!
Estou em divida contigo em relação ao exemplo das redes neuronais... nao me esqueci! :-)

Em relação à minha tese, fiz já o capitulo 1 e 2, e estou já a começar o capitulo 3 que é mais prático, sendo a aplicação das técnicas de data mining no mercado de acções. É um tema ambicioso, tenho lido muito sobre o mercado de acções (movimento, factores que influenciam a tomada de decisão, previsão ou estratégia...) de modo a nao ser descredibilizado. Tenho também lido bastante sobre a aplicação das tecnicas de data mining nos mercados...e por isso sinto estar bem encaminhado...

Nas pessoas em quem confio, que és um caso, enviarei-te assim que possivel.

Um abraçao Vitor, e que esteja a correu tudo bem contigo ai na primavera!!!
Pedro

Pedro said...

Sorry... I forgot and wrote the previous comment in portuguese... sorry!
pedro

Anonymous said...

Congratulations on the successful defense of your thesis.
Your topic sounds so deep and rich - while I still take time to grasp how many sides a cube has!

Good work!
Your Canadian friend,
Rob

Pedro said...

dear friend Rob!
Thanks for your words... I'm trying to do my best on this dissertation... last week was only the first round... I have another one in may and the final round in the end of this year.

My big advantage, is the motivation and the pleasure I have at working on this!

I hope to hear moe news from you!
Cheers!
Pedro

The portuguese friend! :-)

Unknown said...

Hi Pedro:

Firstly, I congratulate you for taking an initiative on developing stock performance estimation system. I would like to know your mechanism for generation information and then doing OLAP on it.

Please write to me at dave.shar@gtf.net.au

Peace,

David Shar
First Global Securities Ltd

Ella Maschiach said...

Hi Pedro,

Any chance you'll post the ETL package here in the future? I'm sure we're all very interested as your thesis sounds fascinating.
Moreover, can we expect a post about SQL Bits 4? How was it?

All the best,
Ella

Pedro said...

HI David,
I sent you an email... i will try to keep in touch!
Thanks!

Hi Ella,
Of course I will post the ETL process!
Unfortunaty I wasnt't at Manchester to see SQL Bits IV... I tried on-line but they had some network problems...:-(

Cheers!
Pedro

Pedro said...

Message from David Shar
"
Pedro:

Thanks for your email and information. I was doing some business oriented research over BI module based on probabilities.

I understand that the algorithms can be used in order to predict activities. The vacillating nature of the stock market requires the use of data mining techniques like clustering for stock market analysis and prediction. Genetic algorithms and neural networks have the ability to handle complex data and are capable to tender authentic probabilities. In my opinion one can provide a platform based on genetic algorithms. The point that I wish to make is what logic could possibly enhance the learning rate of convergence.

In my opinion it would be appropriate if you choose any open source platform for the convenience it offers. I would like to discuss this with you at a length, please write back.

Best,

David Shar
"

Pedro said...

Hi David,

I know the risk I'll face choosing stock market as the subject of my dissertation, and for that reason in 3 months I read 7 stock market books. Now I'm currently applying all the books and articles that I read, into practice... I already created some ETL process and Data Mart for apply data mining techniques and analise the evolution about my strategy comparing to a passive strategy of a stock market magazine (Portuguese magazine). I really dont like the word "Prediction" in stock market and in other activity, and only because the humanity is not prepared to predict the future, because are only focused I study normal things and not the abnormal things, that are rare but with a high impact. The abnormal things that statistics call outliers and they ignore in a normal distribution. And for this reason, I prefer the use of the word strategy in spite of use prediction... because I will apply the output signal of a neural network (buy/sell/hold) with few but vey important rules investigated from persons like Dr. Chong and some stock market funcionalities like stop loss to assume the losses and set the profit goals.

Now, relating to tecnologies to apply, I selected Microsoft tools... and I give you some simple and important reasons:

1. If you have a SQL Server license (that is normal in any medium and large size companies) you have the BI tools like SSIS, SSAS AND SSRS for free.

2. Is portable and not so complex to learn. (I have a virtual macine in my laptop with all BI tools refered in previous point, plus PerfomancePoint Server and Sharepoint to work as front end)

3. As my friend Ella refered in other post of this blog, Microsoft is more intuitive and more easy to implement all the Kimball concepts (Kimball is one of the most experts in BI subjects)

4. Learning Open Source BI tools, could be a waste of time, because you need lot of time and the future of this products are strong unpredictable wrigth now... and nothing is free... but in this point, I will be waiting for a research that is being made by a master BI Coleague... (Filipe Nunes)

Thanks for your feedback!
Regards,
Pedro

Pratheep Reddiar said...

hi pedro,
i think i found your blog somehow, when i was actually searching for some SSIS stuff, Iam pradeep from India, iam a software engineer specializing in BI/DW, i work with SQL SERVER 2005/SSIS and BOXI R 3.1

When i read your blog it was amazing, found a lot of coincidence i was a day trader in India's Stock market(NIFTY) before taking up this Software Career both BI and stocks are really fascinating and intersting stuff

Iam looking forward to work closely with you as iam crazy bout both BI and Stocks, please tell me if there are any options :)

Thanks,
Pradeep

Pedro said...

Hi Pradeep,
Thanks for your comments!
I dont work in Stock Markets.. what I'm currently doing is a research in a scope of my master BI degree about Stock market!
I can put you in touch abot it, if you want!
regards,
Pedro

Pedro said...

Hi sharetipsinfo,
thanks a lot for your comments.
You almost convinced me to use indian tickets...
Could you suggest one? Give me the ticker quote... I have almost the process automated...
Best regards,
Pedro

LinkWithin

Related Posts Plugin for WordPress, Blogger...