User Login

Error
Home Courses Course Calendar Teaching Team About Us Resources Contact
Close

Webcast: Applying Data-Mining in Finance

Speaker: Dr Jan De Spiegeleer

To watch this webcast now, please fill-in your email below:

Please enable JavaScript. Otherwise you cannot view this content.

*


A course on this topic is available in New York, London, Singapore and Sydney

Webcast Agenda


  • General concept of applying machine learning to financial data
    • Cross validation
    • Supervised vs. unsupervised learning
    • Classification vs. regression
    • Data visualization
  • Toolkit of the Data Scientist: main programming languages used in data science
  • Big data and big error: how a well-known classification model (Naive Bayes) can fail to achieve a correct classification on a simple dataset
  • Decision Trees
  • Case studies: K-Means Clustering & Ridge Regression
     

Q&A


1. Could you name other techniques used to find the relation between equity and credit?
A. The relationship as I have put forward on the webcast is used a lot. If you want to study more on this topic, I suggest the paper of Andersen-Boffum (2002). This can be downloaded from the following site (SSRN): http://papers.ssrn.com/sol3/papers.cfm?abstract_id=355308

2. Any reason you do not recommend Matlab as a language for data mining?
A. In fact, I am a big user of MatLab. However, what I like a lot in Python is its open source character and the power of the Pandas package. Matlab has a time series package that can be used as well in data-analysis, but in my opinion it is less powerful compared to what Pandas (Python) is offering.

3. Could you explain what pruning exactly is?
A. Pruning a tree is removing those nodes in the tree that do not add a lot of value (impurity improvement).

4. In the example with credit calibration to equity, does the k-mean clustering approach improve the overall quality of fit?
A. It only improves the local fit. The clustering allows you to deal with the fact that the link between equity and credit depends on the level of the stock price. For high stock prices, a change in the equity level hardly impacts the instantaneous default risk.

5. Any sites that allow data crunching grid that take our python scripts and send back results?
A. Yes! A lot of initiatives have been taken. I would suggest starting with Microsoft Azure and Amazon web services.

6. Current courses focus on specific areas such as Business Intelligence, Statistics and Finances. What you indicate to start learning in data mining applied to finances?
A. Come to our ‘Data Mining in Finance’ programme at LFS - https://www.londonfs.com/programmes/Mining-Big-Data-in-Finance/Overview/

7. Where do you see the future of analytics in financial markets?
A. Algorithmic trading, improving technical analysis, equity analysis (automating balance sheet studies), credit scoring, risk management (kernel densities for VaR calculations), etc.

8. In your opinion, what is the single most important area of financial markets to benefit from a strong analytics program?
A. Risk Management, in my opinion.

9. How successful can a machine learning algorithm be at predicting the stock market given efficient markets theory? Are there parts of the market or execution styles that are more suitable for this type of approach, like for example high frequency trading?
A. The fact that some hedge funds (e.g. Renaissance, BlueCrest) are successful in what they do indicates that there is value in applying statistics into a larger portfolio management context. In my opinion, this is not related to a particular execution style. A long-only portfolio manager might, for example, use a data-mining technique such as a logistic regression to study a balance sheet and construct - in a fully automated way - a probabilistic ranking of shares. The same method can be put at work on a high frequency desk. Here the input variables will be different and an investor will rely more on technical data.

10. Are people using neural nets in predicting the stock market? Where can I find more details? Also please recommend some books that we can use to build on your excellent session...
A. Yes, the example I covered in the course was classification trees. This is the same domain where neural nets could be put at work. If you want to read more on neural nets, a very good entry point is the book of Tibshirani, Hastie and Friedmand. This can be downloaded from the website of Stanford: https://web.stanford.edu/~hastie/local.ftp/Springer/OLD/ESLII_print4.pdf

Thank you to those attendees who submitted their questions.


LFS offers the 2-day 'Data Mining in Finance' programme with Dr Jan De Spiegeleer in London, New York and Singapore.

To find out more, click on the location links above or contact us at advisor@londonfs.com

Are you interested in running our public courses in-house? Contact our in-house team to discuss further.


Why travel? Many clients are already attending our courses from the convenience of their home or office with LFS's state-of-the-art remote learning platform: LFS Live