CODEWITHSUNDEEP
data-miningdata-warehousebusiness-intelligence
invinc4u
invinc4u

Reputation: 1185

Data mining in finance domain ideas

I work for a financial firm and maintain the security(stocks) and trade data warehouse . I was wondering if someone could point out what kind of interesting information could be mined and extracted out of these warehouses ? I have recently been reading about Data Mining and Information Extraction and wanted to apply those algorithms to mine something that might be useful to a financial firm in general . Any ideas would be highly appreciated.

Upvotes: 0

Views: 881

Answers (2)

doug
doug

Reputation: 70038

About one year ago, i helped a friend who works for an ibank with a side project directed to exploratory mining of a large data mart comprised of financial data.

I suggested using an Association Rules algorithm (aka Market Basket Analysis) to perform exploratory mining of this data mart. He told me that he and those in his group were impressed by the results and intended to install this technique in their periodic mining workflow.

I chose Association Rules for that project because:

  • it runs nearly out-of-the-box (in particular, very little pre-processing of the data is required)

  • it is simple and fast to configure and begin using--in essence, you identify the data source and a minimum threshold for the association strength (i.e., *don't return any rule sets with an association below _*)

  • excellent open-source implementations are available--the two that i knew about are Orange (written in C++/python, scripting interface in Python), and R.

In R, Association Rules mining is available through several third-party packages, the one used most often is arules, available on RForge.

For Orange, the module you want is actually included (orngAssoc).

I don't know exactly how common it is to use this technique on Financial data; however, i do know there's certainly plenty of precedent for it, and it is currently applied in this context with success. (See e.g., Forecasting changes in Korea Composite Stock Price Index (KOSPI) using association rules.)

Upvotes: 1

valmo
valmo

Reputation: 1174

best bet would be to have a friendly chat to your financial analysts and ask them what tickles their fancy. i think anything that would look good on a graph and you could show weekly would be good.

Upvotes: 0

Related Questions