5 mistakes I made in my first year of machine learning and what I learned from them

                             5 mistakes of a ml beginner

Introduction:

This is a lightweight non-technical post. I have been practicing the craft named machine learning for the last 2.5 years now. I recently have started working with a few data science aspirants, and found them making a lot of similar mistakes like I would do. In this article, I will share 5 mistakes I did, and a lot of machine learning and data science beginners do.

First mistake: not reading your data:

Data science is exciting and machine learning starts with learning a lot of models and algorithms. Hence often when we start learning machine learning and data science, we don't learn the most important step. The most important step is to read your data. Now, I used to find it stupid to read the data, because I didn't know what does reading data mean. 

Reading data means understanding different patterns of the data manually, errors and inconsistencies in the data. Let's say you have a text data. Now reading this data means, you will have to see what are the types of sentences, tokens and paragraphs. You will try to find out what are the special types of phrases which are occurring, if the data makes sense in all the places and others. If you ever get a chance to work with me, we will go through ways of data reading together.

Second mistake: sticking with few algorithms:

During my first few months, I learnt random forest, linear regression, xgboost and a few other algorithms. As my first few projects involved non-linear relationship data, I got good results with random forest and xgboost. Soon, I was applying random forest and xgboost to most problems; and ran into problems when the application and type of data changed. The learning here is, never get attached with any algorithm or subset of algorithms. There is no algorithm which fits all; and therefore if you get yourself applying same 2-3 algorithms everywhere, time to check again.You need to consider atleast 4-5 algorithms for each problem, after considering the problem thoroughly.

Third mistake: not working on soft-skills:

For a data scientist, often we get so engrossed with technical works, that we don't give enough value to soft skills. For a data scientist in industry, the soft skills most needed are presentation and story telling. And after reading this, if you think that you are good with these skills from start only, think again. The story telling and presentation need a intense practice. Part of it comes from writing blogs, and creating beautiful side projects related to data science. The other one, comes from a change of habits. I had to put active efforts of writing blogs, presenting to my peers for getting feedbacks to improve on; as well as sought detailed guidance from seniors who give good presentation and tells great business stories. So think again and start working on these.

Fourth mistake: not joining kaggle earlier:

I didn't join kaggle when I started out my data science career. Later, when I joined kaggle, I realized that the kaggle community is a thriving machine learning community, specially important for beginners and amateurs like me. We can learn all the three main important points I mentioned before this; together from kaggle. So, if you are a data aspirant and not yet active in kaggle, do signup today.

Fifth mistake: not learning enough:

It takes a lot of effort to get to the first data science job. We learn, hustle and try so many things to get that first job. But once I got the job, my learning decreased from that peak for the next 6 months. Almost for a month or two, I became so busy and involved with office work and leisure, that I didn't almost learn anything new outside my work. This was a mistake. 

Whether you have a job or not, if you are a data science aspirant; you have to be reading all the time. There are 3 types of things to keep reading and a balanced diet of these are needed to keep your profile healthy and alive. These three topics are:

(1) new frameworks and algorithms (2) research papers (3) new business practices in ml and data science

So once I restarted my learning, my performance at work as well as in open source projects also started to pick up. As of today also, I am still improving my learning a lot. So it's a upward journey always.

Conclusion:

So in this article, I shared with you the five mistakes I made when I started out, and what are the things I did from there to remedy it. Hope this article helps you to grow too. If you have any suggestions or criticism; please comment and let me know. Also comment your mistakes as a data scientist/ aspirant which you have identified and worked on.

Thanks for reading! subscribe to my blog for more such awesome contents!