Analysis: Which method is used to build a stock quantitative trading database?

Quantitative finance requires a lot of data, and once you start processing data for smaller than daily time periods, the data becomes even larger.

Another approach widely used in the industry is to build pure Python solutions.Actually, my personal feelings about Python are mixed.

Python has the advantage of being very easy to code and counting using many ML libraries, Jupyter notebooks can simplify structural maintenance and facilitate self-documentation of the process.

This may not be the best solution, but in a mixed environment, I think it's very handy in terms of operation.I've always thought that Python has no advantage in performance or speed, but is very good at writing high-level maintainable software layers.The benefits of shorter development time and easier team collaboration outweigh performance barriers.

Experienced Python developers can also use alternative methods that can speed up Python.This approach has been used very successfully at an established small hedge fund firm.

C++/Java method

Another approach involves using C++ or Java as the environment and developing a custom framework to process the data.This is the method used by intraday high frequency operating companies with large amounts of data.

For scale data, this approach may be required at least in some areas of the infrastructure.Regarding the specific language to use, C++ is the first choice for flagship institutional companies, but Java can operate at a lower cost and there are very successful small companies that use Java to provide everything.The reality is that we can get almost the same performance as C++, but with a much easier language and development and debugging environment to manage than C++.

Some aspects of the trading software can be better expressed in the C/C++ language.For example, defining data structures for OHLC and quote data is more natural in C than in Java.

Using Java in transactions always involves coding in a way that is not very natural for a Java programmer, since we need to always consider low-level programming and performance, so a C background is beneficial.