In short, the main purpose of Blue Sphere ML module is to predict stocks prices based on companies quarterly financial reports.
The idea behind this is fundamental analysis. What fundamental analysis in the stock market is trying to achieve, is finding out the true value of a stock, which then can be compared with the value it is being traded on stock markets. And if the true value is above a traded one the next move is to buy and hold stock until it reaches its true value price.
Finding out the true value can be done by various methods with basically the same principle. The principle is that a company is worth all of its future profits added together and then discounted to their present value. One of the methods to find future profits of a company is to evaluate a company’s past and current performance as well as the credibility of its accounts.
Usually, such an analysis is being performed by financial analysts. And our endeavor is to substitute those financial analysts with ML algorythm to perform their job.
To do so this ML need some data inputs. Then ML Algorythm will process those inputs trough it’s deep neural network and will create an output. And that output will be a predicted stock price.
Seems to be simple but not clear. Let’s go deep into details then.
What are those inputs?
They are data from quarterly financial reports of a company. We use the EDGAR system provided by the SEC to get those financial reports. Then we extract data from reports and prepare it to be loaded into Blue Sphere ML.
On the next step, data is being processed in the hidden layers of Blue Sphere neural network. There are 3 hidden layers with 200 000 parameters in total.
And finally, we get the desired expected stock price figure from the output.