DATA SCIENCE AND MACHINE LEARNING; ONLY FOR THE UNDAUNTED.
Most time people think good stuff come by chance or luck. While that might be true for certain life situations, same may not apply to other kinds of situations – duality. Doing data science (DS) and Machine Learning(ML) projects unarguably is not the former. With all the media attention DS/ML is attracting, as being the future and the interminable training offered by diverse individual and organization in this space, an entrant may be quick to think doing a project is as easy as portrayed.
Who am I? Yea sure, no expert. Why the heck do I think I can say a think about DS/ML let alone make such a bold statement that DS/ML is only for the undaunted. Well just sharing experiences and lessons learned while journeying the path of becoming a data scientist so you can validate, correct and give feedback. I am just some young chap who has been given the invaluable opportunity to learn, explore and “test drive” the domain doing practical things. From Data Science and Machine Learning to DevOps and Cloud computing – AWS and GCP, and even Web Development(Django).
Learning and mastering a technology and technical concepts and applying them to a project in a short time is a very challenging venture. And to bring it home to our subject, DS/ML is no bread and butter. Most advise come in the form of; learn Python or R and its associated library to become a data ninja. The reality is far from simply learning these, understanding how to use them and applying them to a data set. More often those who have gotten a grasp of the language and library are mistaken in the thought that it is a plug and play task. That machine learning gives you an answer once you input the acquired values. Well sort of but not entirely so. The core of doing data science and machine learning and coming up with relevant answers is the practitioner’s thought process, imagination, and intuition. We say machine learning. Yes, it is, but man giving the machine a congruent data to learn from. That is why two sound accomplished DS/ML practitioner would come up with relevant answers with a different degree of correctness and unmatched potential for some adverse effect.
What this means is that practitioners profile data, get a glimpse of intrinsic representation of the data, apply some domain understanding and intuition of some expected factors, prune the excesses, select likely features they believe to tend towards relevant answers, mash them into the proper formats their chosen algorithms have been designed for, and finally hand this data to the machine. Now, this is the definition. The practitioner did their homework, profiled the data, observed the statistical meaning of the data – all things reduce to number, then brings intuition, and domain knowledge in.
As Google best expressed it: Do machine learning like the great engineer you are, not like the great machine learning expert you aren’t.
