When performing large amount of calculations, Python is in general ~100 times slower than C++. Still, it is de-facto standard in Data Science. How?
Reasons:
- Quick development — Python is a high-level language, that is easy to learn and to use
- Many Python libs (like numpy) are wrappers for C++ code, thus making performance issues less severe
- Enormously large and developed ecosystem
- In a data processing cycle (find and prepare, develop, IO, calculate, analize, etc.) coding time worth more than calculation time
- Python is interpreted language rather than compiled. Compilation would need recalculating large amounts of already processed data (when project is of that nature)