Machine Learning 101
An almost non-technical explanation for anyone new to machine learning!
Before I get into this post, I want to thank all those that took the time to read my first post entitled ”Transition from Chemical Engineer to Data Science Enthusiast” then clapped, commented, shared. I hope everyone is having a great start to their new year
I know many of you asked me how to go about “doing” data science and if there were any courses available.
There are many free courses available online through sites like Coursera, EdX, DataCamp etc., on various topics in data science/business analytics/STATISTICS(this is super important), however, I would suggest you use those as an introduction and then work on your own project. It really is the best way to learn — LEARN by DOING :)
So back to the task at hand, what is Artificial Intelligence(AI)?
You could probably google the definition and end up with 340 000 000 results in less than 0.66 seconds!
However, I’ll define it simply in my own words: AI is the study of attempting to teach a computer to think or make decisions intelligently just as a human would, with room for calculated error. I could include jargon in the definition but let's keep it simple!
The branch/subset of AI that is applicable to data science is known as Machine Learning(ML) and how would we define ML? I think Wikipedia defines this simply:
“ Machine learning (ML) is the study of algorithms and statistical models that computer systems use to progressively improve their performance on a specific task.”- Wiki
ML algorithms learn from past experiences without being explicitly programmed to do so. There are 2 main types of machine learning algorithms: Supervised and Unsupervised.
Supervised learning may be defined as a data set that is labeled. For example:
The table above shows a sample of loan data (fictitious data that I made up). The last column “Loan_Status” is referred to as the label(outcome) column. This column indicates using Y/N (yes/no) if the loan has been granted.
An algorithm that uses this labeled data is referred to as a supervised learning algorithm. So I’m sure you can see that one particular application of such an algorithm will be to determine who gets a loan based on previous data of who was granted/denied a loan.
Realistically speaking, we would add another column that indicates of those granted who has kept to the loan repayments and who has defaulted because that would make more sense!
Ideally, you want to grant loans only to people who are going to pay back. There are many algorithms that fall into the category of supervised learning but we won’t tackle that in this post.
Now let’s talk about unsupervised learning. This is (naturally) an unlabeled data set. For example:
We feed the algorithm random images(input data) and thereafter the algorithm sees a pattern between the images and groups together or clusters similar images. In this examples 3 different types of fruit are fed and the algorithm recognizes this.
Therefore the output shows 3 distinct groups. A real-life example of cluster analysis would be customer segmentation for marketing analysis purposes. Specific clusters (or groups of customers) can be identified within a population.
Analysis of these groups can then determine how likely a population cluster is to purchase particular products or services. A marketing team can then target each cluster with tailored, targeted communication.
So that concludes our basic overview. I didn’t want these posts to be tutorials(there are plenty of those available on the net). I wanted to use these posts to expose a new kind of thinking to solve problems.
You don’t have to be a data scientist to apply data science. I’m in a pure chemical engineering field(for now until I decide where to from here…) and still finding it applicable to my work. However, be prepared for hours of cleaning up data(most times) and bad data which can lead to no/inconclusive results.
In the next post, I’ll be introducing a specific type of supervised machine learning algorithm known as Decision Trees. They have widespread applications and are super easy to understand (no technical background needed). It’s also one of the algorithms used in the loan prediction problem!
I hope this post was quick in providing you with a basic overview and if you enjoyed it please do clap, comment and share!