Difference Between Data Science ,Data Analytics And AI
“The breeze feels lovely today”, said Atul as he sprawled on the couch. He was on one of his weekend visits to my place where we generally hang out over coffee and discuss cool stuff. Atul and I go way back to our college days when he joined year junior to me in Statistics discipline. The senior junior camaraderie grew into a strong thought partnership over the years we knew each other.
“The weather is just right for a great evening on the balcony”, I quipped handing him over his cup of coffee.“Ah the coffee is amazing. To add to it, I have got a few questions to pick your brain”, Atul said as he fished out his notebook from his technical training sessions at office.
“Fire away. I’ll do my best. I believe they are teaching your batch the recently developed on Data Science course curriculum. How is it going?”, I said intrigued by what comes next.
“Raj, I must say, the course is amazing but there are areas that needs a bit of unclogging. For example, what is the difference between ‘Data Science’ and Data Analytics’? Are these interchangeable or one is a subset of another?” said Atul sipping coffee with the gentle spring breeze blowing into our faces?
“Good question Atul. Let’s see. Data Analytics is a branch within Data Science. Think of it as slicing and dicing of data to generate answers to specific questions related to business portfolios. Data
Analytics is what helps an analyst generate insights and create dashboards related to various aspects of customer lifecycle”, I said.“Let’s take an example. If I try find out what are top revenue generating marketing channels for Product X and Y over the last 12 months, I would be performing Analytics on the data to get to the answer. Therefore, any analysis that leads to insights related to comparison and trend over time would fall under data analytics! Isn’t it? “, said Atul
“That is right Atul, Data Analytics helps you look at past data and requires you to perform Descriptive Analysis on it to find out insights. But remember, that insights must be actionable otherwise they are just ‘Good to Know’ facts. To get past that problem, one must always link the business objective to the overall approach.
“Thanks Raj, that helps. Then there is Machine Learning and AI! How different are these concepts
from each other? Are they related to Data Analytics in some way? I mean as a part of Data Science course, would I automatically learn AI or is it supposed be separate branch of study”, Atul fires a volley of questions.
“Slow down dear friend. Sip in the coffee and let it sink in. It is going to take some time to answer all that!”, I said smiling.
?“Based on my experience Atul, Machine Learning can be described as a meeting place of Art and Science. We already discussed that analysing past data can reveal insights that are actionable.However, these insights are often disparate and needs weaving to structure a data story. Let me explain with a Customer Attrition use case. Say, you are interested to know if chances of customer attrition increase with tenure or not, you’d be performing analytics on customer data. You can look at the customer data in multiple ways and study the relationship between Attrition and individual behaviour attributes. All of this will lead to individual yet powerful behaviour insights”, I said taking a pause to finish the remaining coffee.
Placing it on table I lean over the balcony railing and say,” Machine Learning is much beyond that.
Like I said earlier it is meeting place of Art and Science where a Data Scientist uses technical and creative skills to create a quantitative framework for not only describing past events but also enable future prediction”
“Raj, that sounds good and perplexing at the same time. Help me out here! Would you mind explaining with an example?”, said a visibly confused Atul.
“Not at all. In the attrition use case, say we want to link up various observed customer behaviour attributes and compute probability of attrition in the next 6 months. If we embark on such a problem to solve, then we would be stepping into the domain of Machine Learning. We usually ask ourselves two fundamental questions at this point: A – Is my framework supposed to predict the likelihood of a certain event or B – Is the framework supposed to calculate a future value of a metric. A lead to a Classification Model and B leads to a Regression Model. The Attrition model is a Classification Problem. Now, among the host of techniques available in the Machine Learning arsenal, there is no one technique that works best all the time. You may have to try out a few algorithms before finalizing the model”,
“When we build Machine Learning model, we link various independent attributes to an outcome in an algorithmic manner. In our customer attrition case, when we train the model on voluminous data, the chosen algorithm picks up specific characteristics that maximises the likelihood of differentiating customers who are likely attritors vs those who are less likely to attrite.”
“Makes sense Raj”, said Atul,” I am beginning to piece this puzzle together. Machine Learning essentially creates rules to define the Decision Boundaries. Nevertheless, so far it feels like Machine Learning and AI are not different from each other. Are they interchangeable terms?”
“Well Atul. You can think of ML as a precursor to AI. What it means it that when ML is deployed at a scale which enables algorithms to self-adjust to underlying data to optimize decision boundaries, we achieve AI in our decision systems. AI tries to mimic the way humans learn themselves. If AI is enabled, then models would follow and adaptive methodology and readjust its predictive weights as per the changes in the underlying data in real time. It then becomes an autonomous learning system and would not require manual intervention in re-training of the model. AI enablement requires many training iterations and takes learnings from each step into account. Sounds good does it not?”
I said looking at the sunset over skyscraper filled city horizon as flock of homeward bound birds tweeted goodbye to us.
“Raj, thank you for explaining the differences to me and now I think I am much clearer about how best to place each of these items in a relation to one another. Here let me show you”, Atul said grabbing a pen and drawing the following diagram in his notebook to summarize our conversation.
“Excellent Atul! You’ve been following closely indeed. I’d like to add another layer to it if you don’t mind”, I said adding another circle within the AI universe and the following diagram emerged.