What is the difference between beta and correlation coefficient?
Beta shows how strongly one stock (or portfolio) responds to systemic volatility of the entire market. A beta of 1 means that the stock responds to market volatility in tandem with the market, on average. A larger beta means that the stock is more susceptible to market risk while a beta less than 1 means that the stock is less responsive to market risk. Beta values are not bounded like the correlation coefficient. Correlation coefficient, on the other hand, must be between -1 and 1, where -1 means that the stock and the market move opposite of each other, 0 means that the stock and the market movements don't have a relationship, and 1 means that the stock moves with the market. Because of their different value meanings and bounds, the formulas are different. From Wikipedia, the correlation coefficient iswhere X is the stock return and Y is the market return.Beta is defined aswhich is equivalent towhere r_a is the stock return and r_b is the market return.In practice we use beta because it is volatility of a security relative to the benchmark. Correlation does not distinguish which we are studying with respect to the other.
Where can I find the beta of Indian stocks, and what is a link to it?
You'll get it on Reuters's Indian website Here's the link to the Reuters's page of Reliance Industries: Reliance Industries Ltd stock quote, Reliance Industries Ltd company overviewAnd here is a screenshot. The beta value is highlighted in green.
How can I get started applying machine learning to algorithmic trading? Are there any interesting papers to read?
First, you would focus on gathering as much data as possible and putting into a single large table form. This would be historical price data. Maybe augmentthis with newspaper articles, blog posts, sec filings turned into word count vectors etc. (using natural language processing techiques). You would then train a supervised algorithm for buy/sell decision. Suitable algorithms arelogistic regression (fastest) and random forests (most accurate usually). There are others, such as support vector machines, boosted decision trees,3-layer neural networks, but these don't offer as good accuracy as random forests (and often slower as well) or as much speed as logistic regression. In my opinion, the best choice would simply belogistic regression, and the best implementation is vowpal wabbit - extremely fast, can handle hugeamounts of data - 1 terabyte an hour on one machine, faster still in cluster - and open source. This also allows you to clearly see which indicators (columns in the table) are predictive.You could also add newspaper articles etc. directly as text to this by using the hashing trick. This is also implemented in vowpal wabbit as well, and so youcan handle very large free-form text articles in a single row as well.Having done the above to the point you can no longer collect more sources of data, it's time to move to the feature engineering stage. You already did someof this when you did nlp stuff (but you were using standard techniques - ie. did not think up/invent your own methods).There are two choices at this point - manual and automatic.Usually, people at this point invent their own features. This is what traders spend most of their time doing - the so-called "strategies" or "rules". These are tested against the data - called backtesting.Another, newer, authomatic method has also recently become available - unsupervised deep learning. Unsupervised learning existed before,but it was of the "shallow" variety and did not work well in practice. Deep learning neural networks using autoencoders is a new method (invented just 6 years ago) that works really well. This paper is a demonstration of it.http://research.google.com/pubs/...Essentially, by throwing lots of computers at the problem, it's possible to automatically form strategies. The larger the neural network, the better it does - but consequently more computers are required. This (in my opinion) though, is better/cheaper than trying to hire lots of highly creative/analytical/hard-working people.
How can you insert the alpha symbol in Microsoft Word?
Go to Insert on the ribbon, select ‘symbol’ and then it will depend on what fonts you have available. Symbol font does have a lowercase Greek set that includes the alpha. It is ‘ 97.’Thanks for the A2A.