Why Becoming A Data Scientist Is Very Difficult?

By Author: tarun
Total Articles: 9
Comment this article

Why becoming a data scientist is very difficult?
I was doing my research on data science and came across this article that says, “Becoming a data scientist is easier than you think”. And I was very confused after going through that article. That article also said “You can take the ML course on Coursera and you're magically a data scientist” and I completely disagree with this statement. According to my being a data scientist requires a larger skill set than understanding a few basic algorithms. I took a machine learning course from coursera and here is the list of the things that I didn’t learn in the course.
Programming languages and other important technologies:
Most companies that employ data scientists are not using Matlab and octave. Their backend web services are written in java, python, Scala, or ruby. These languages are not covered in the course I took up online. Python has libraries like Scipy, Numpy which are very great for solving numerical problems. R is used by most of the statisticians which are again not covered for me. When it comes to integrating an algorithm into a pre-existing web service, and you only know Matlab, that is going to be a huge problem. You have to be familiar with these languages to understand large, pre-existing codebases.
Big data software:
Even if you don’t know Java, you still need to know the importance of Hadoop.
Learning algorithms:
The course I took from coursera skipped over Bayesian learning. A lot of systems use this in production, but you would obviously know nothing about it.

Feature Extraction:
You can use most of the algorithms from machine learning courses to solve a real-world problem. Extracting features usually requires a deep understanding of the problem, then distributing the data or knowing the familiarity of how data is being generated.
Data cleaning:
Coursera even write all the scripts to load the data. That really doesn’t help you in the real world. You must know regular UNIX commands like Sed, Sort, map to clean these data setup.
Statistics and Probability:
These online courses touch only on some of the topics, but real-world problems usually require a deeper understanding when it comes to solving them.
For example,
1) What is F-test?
2) What are the standard errors?
3) What is a ROC curve?
4) What is hypothesis testing and when you can use it?
Database issues:
For the purpose of the machine learning course, the data is being stored in flat files but in the practical world, the data is being stored in MySQL, Casandra, or on the HDFS.
Visualization:
If you think Matlab 2D plots are awesome, Check out D3.js. You need to know functional programming, javascript and the learning curve for the D3.js APIs.
Debugging:
This topic is entirely covered in a machine learning course. You actually need to know about Conjugate Gradients, Partial Differential Equations and potentially a lot of other subjects. The machine learning course is only 8 weeks and they are not able to cover all the topics related to machine learning in course.
Beyond what I mentioned in the above article, being a data scientist takes years of experience. You should know more than how an algorithm works. No online course will be able to teach you the depth part of machine learning.
You really need a personal mentor to become a professional in machine learning course, where the need of Innomatics comes which is the best data science training institute in hyderabad
Innomatics is well known for data science course in Hyderabad