we analyze Habr, Habr Career and Stepik
It turned out to be not so easy to mark up Stepik courses. After visually evaluating the model’s predictions, I came to the conclusion that it performed poorly here. Only about 200 courses out of 1.5 thousand were marked up correctly, so I had to go back to my ancestors, namely mining fire labeling data manually. At some point, I got tired (after all, this is not how I imagined the life of a programmer!), So having marked a third of the data, I trained the neural network I had already written on them, and for the rest I got predictions.
Analysis
As a result, I had thousands of observations from different sites ready for analysis in the period from 2016 to 2021. For all datasets in each direction, I built trend lines, and in addition to the number of observations, I also investigated the interaction of people with these areas, that is, for example, the number of comments and views. It is worth noting that I considered percentage ratios, not absolute values - this way you can avoid the influence of the growing popularity of one of the three selected sites. In addition, we managed to collect interesting statistics on the most popular professions from the career.habr website and analyze trends in them.
What do we have?
Perhaps, I will not talk in detail about the trends in all the selected areas. I will focus only on “technical” ones (and if you are interested, I can tell you about the rest in the comments).
The strongest growth is observed in the direction “Iron and Electronics”: the number of vacancies is steadily growing, as is the interest of people:
The situation is somewhat sadder with web development: the number of vacancies and articles is decreasing, while user activity (comments, saves, views) is growing:
This is also noticeable when analyzing the demand for Frontend developers and PHP developers – the number of vacancies is falling:
The reverse situation with machine learning: demand in the labor market and in the educational sector is gaining momentum (albeit not so much), while the activity of users on this topic begins to fade:
The same conclusion can be drawn about administration and Devops: this direction is becoming more and more popular with employers, although the interest of people is slightly declining:
A similar situation with information security: The interest of employers in this area is growing, which is reflected in the increasing demand for courses, but among blogs, the popularity of this topic is greatly reduced.
Despite the decreasing number of articles in the direction design and testing, in general, interest both from blog readers and course participants, and from the labor market, is slightly increasing.
You can argue for a long time why certain results were obtained. But I think it deserves a separate study. What do you think?