applications of machine learning / data mining

What are some areas of application of machine learning and data mining algorithms, which is directly used to generate revenue for the business?

For example, I know that a lot of data companies simply see machine learning and data mining scientists as must-haves, even though their role is not a major contribution of wealth to the business.

However, are there types of business which directly use machine learning to generate revenue? Or maybe other statistical algorithms.

As another example, a lot of banks employ quants for regulatory purposes, but those roles are simply used to please regulators. In other words, most of the bank's direct revenue comes from other departments, from traders in front office, etc. But most quants work support roles, which are not even directly relevant to the business itself.

Quote:

---

What are some areas of application of machine learning and data mining algorithms, which is directly used to generate revenue for the business?

---

How google ranks their results, how amazon creates recommendations, how facebook grows their network, how some use advertising (and where they may choose to place it based upon their mining results) are all examples - these are all core technologies of these companies and while each is different, each leads to sales in one form or another (direct sales of products, advertising, etc...). Another example: in the biomedical industry data mining can be used to identify biomarkers - markers which can be used to diagnose and study diseases, which in turn leads to products (eg sales) in the form or kits for diagnosis, drugs for treatment, or even better: cures.

Quote:

---

Originally Posted by copeg

How google ranks their results, how amazon creates recommendations, how facebook grows their network, how some use advertising (and where they may choose to place it based upon their mining results) are all examples - these are all core technologies of these companies and while each is different, each leads to sales in one form or another (direct sales of products, advertising, etc...). Another example: in the biomedical industry data mining can be used to identify biomarkers - markers which can be used to diagnose and study diseases, which in turn leads to products (eg sales) in the form or kits for diagnosis, drugs for treatment, or even better: cures.

---

I guess what I also want to know is how profitable are some of these jobs. For example, someone writing ML algos for stock market prediction is probably being paid a lot more than a bio data miner, because bio data miner is further from the money and the lion share of profits probably goes to the sales team.

Are there many ML jobs which have a closer involvement to the financial aspect of things?

I can tell you that the finance market highly recruits anyone with a strong mathematical background. Personally I find this annoying because I get a rather large quantity of email and physical mail asking me to pursue an MBA or some other finance degree, or get involved with some math/CS portion of a financial system, etc.

I can't comment on how lucrative such positions are, but there must be a good financial incentive since the demand is fairly large. Somehow I kind of doubt you'd be paid significantly better than you would at a bio firm or a large corporation like Google or Amazon. These are all multi-billion/trillion dollar markets and companies understand the significance of being able to manage lots of data effectively.

Quote:

---

Originally Posted by prokyon

I guess what I also want to know is how profitable are some of these jobs. For example, someone writing ML algos for stock market prediction is probably being paid a lot more than a bio data miner, because bio data miner is further from the money and the lion share of profits probably goes to the sales team.

Are there many ML jobs which have a closer involvement to the financial aspect of things?

---

While I'm not in the finance sector professionally, my understanding is that in the data mining field the efficient-market hypothesis actually detours folks from data mining (the best example I've heard of is an example in which a neural network was training with the mood on Twitter, it was reported that the model could accurately predicted the Dow Jones (I believe the next day closing) - whether this could be used to make a profit is an different story, my guess is no). I'd agree with helloworld that math and statistical aspects of someone's experience would be valuable in this sector.

Quote:

---

Are there many ML jobs which have a closer involvement to the financial aspect of things?

---

Are you talking specifically about the financial sector, or just the profit of a company, because I'd say the examples above are pretty closely tied to profit. Here's two more
- Supermaket chains. Do an experiment, collect the data, and evaluate it using mining approaches for higher profits. For instance, in some stores place items lower on shelves, or move certain items further away from each other, color or place signs differently, place certain items on sale while others are not, customize coupons, whatever...the key is you have a lot of data of different conditions and the result from those changes (profit). You can mine this data to look for patterns and correlations that help increase profit.
- Credit card companies - have to pay for fraudulent charges. They loose money when this happens, so its in their best interest to detect fraud as early as possible. Thus, they data mine cardholder's purchases, categorize the purchases as legitimate or fraud, which helps then detect fraud extremely quickly (with less monetary loss == higher profit)

I'm actually working as a quant for a major bank, where I get to build statistical models for market risk. And I'd like to switch away from finance. I went into more details here, as to why:
Nuclear Phynance

In theory, data mining and machine learning type jobs are supposed to be paying a lot of money, but in reality that's only if you end up working for a top company, where the hiring process is fairly stringent.

I'm convinced that I want to switch away from finance and more into machine learning / IT type of a role, but I'm not sure how to make the swithch at this stage. As posted on Nuclear Phynance, I'm pretty much sold at switching to IT, but the question is how do I get into more ML-type roles. I have pretty basic ML background. I can review C++, data structures, statistics and ML (hidden Markov models, mixture of Gaussians, logistic regression, Monte Carlo methods (rich background in MC methods), Naive Bayes).

From what I've seen so far though, most ML positions already require knowledge of collaborative filtering, for example. Some roles already require Hadoop/MapReduce as well. Any ideas? I need to find more basic ML roles to start.