Avoid these Misconceptions About Data-Science

"Failure is not the opposite of success it is part of the success" - By Debbie Gregory

Data Science jobs are the sexiest jobs of the 21st century". Does this statement ring a bell in your mind?. It sure did for me and it is true to some extent. But believe me, to acquire this knowledge of Data Science is definitely not that sexy.

My journey with Data Science started a couple of years back in a training at my office. Solving business problems using statistics and machine learning models was the training agenda. This caught my attention and I got very passionate about Data Science topics. I was amazed at the idea of solving real problems using Data Science. This motivated me to immediately enroll in some Bootcamp courses and my self-learning journey started from there.

There is a saying in INDIA which I am going to translate here, "Mountain looks smooth from a distance but you will learn it is not as you approach closer".

This is exactly the same when it comes to learning Data Science. You will hit so many snags which you never expected before. There are subjects you have to learn even if you are bad at it and really hate in your colleges days. You really got to start liking Statistics, Programing and you have to improve your storytelling capabilities if you are already not good at it.

In this article, I want to share some pitfalls to avoid before starting the data science journey.

Some misconceptions about Data Science

• Predicting customer churn is data science.

Learning Logistic regression doesn't mean you are doing a Data-Science job. Logistic regression is just one technique within the data science spectrum. Of course, if not all business problems, most of the business problems can be solved using Logistic regression. It is one of the predictive analyses based on classification rules. However, there are many other models which can solve problems related to classification. It is better to understand the algorithm of other machine learning techniques so that one can compare outputs of multiple models and choose the one with better accuracy scores.

• You will surely find success by just following how to learn data science in 6 months timetable

I am sure It may have worked for a few people (if not who will ever share a timetable to become a data scientist in 6 months), but after researching the number of topics you need to learn in order to gain a strong knowledge in data science is huge.

Honestly, 6 months is definitely not enough to master the subject. To get a fair understanding 6 months may be enough, but the key is continuous learning. I cannot emphasize this more. A clear strategy to learn data science and being able to work on that plan on a regular basis is the key. Especially if your current job is not challenging you with data science-related tasks, then it becomes even more difficult to keep the momentum going. There are many other distractions on the way but if you are able to tackle the bigger ones, the smaller ones will fall in place.

• A strong desire to learn Data-Science is just enough

Just desire is not enough, you need a true commitment and passion towards the subject

Being amazed about Data Science and taking some courses will really not help. You need to have a clear roadmap and it should be time-bound.

If you say I will learn statistics, it's just not enough. You need to clearly define what is that you are going to learn and by when. By when is the most important question you need to ask. If you are not asking this question, then you will never be successful in this journey.

• Solving Problems in Kaggle is all you need to gain experience

Of course, problems solved in Kaggle will give a sense of how data science can solve real-world problems. But some random problem in Kaggle may not be the problem at hand and you can only gain the knowledge by doing things on real data.

One of the biggest challenges is going to be data. In a structured course the data you are going to get is already treated and prepared exactly how it needs to be fit into the model. But this is not going to be so easy in the real world.

You have to learn where and how to source the data and data quality is important. Getting relevant data is 50% of the problem. Fitting the model is another 25% of the problem. And the real problem is the remaining 25% which is all about being able to sell to the people who don't necessarily understand statistical concepts. I think the last part is the most difficult in this journey

Thanks for reading through and I hope you will avoid these pitfalls in your data science journey.

There are many other pitfalls and I have just jotted down a few which I have experienced so far. Do let me know what are some of the pitfalls you have experienced and how did you overcome those.