Importance Of Generalized Statistics!

I know you might be thinking, what is this new term called Generalized Statistics ?

Let me ask you a simple question to make it more simple for you, for example you want to take your family for a get-together dinner that includes 8 members in total.

Ask yourself whether:

1. You will look directly for the restaurant
2. Or first you will find out what kind of food family members want to eat for example some of them may like to eat veg,non-veg,chinese or italian food as well.

Above two statements are two different scenarios to understand Generalized Statistics in an easy way.

Don’t you think if you are going to choose the first option,it may lead to a bad choice because without asking from all members what kind of food they are looking for tends not to choose the right restaurant in simple words, you didn’t generalize the situation.

Next question that pops up is How to generalize the situation?

Let’s have a look step by step:

1.) Two important factors here are food & family members primarily , not which restaurant you want to go first.

2.) At first you will figure out what kind of food all members like to eat, some may like chinese,veg or non-veg.

3.) After you figure it out, maybe you are looking to visit a restaurant within 15km or 20km from your place.

4.) Next step would be checking the ratings of the restaurant with all categories of food, family members looking to eat(veg,non-veg & chinese).

5.) At the end check the cost for two people then whichever fits in your budget choosing the right one for the perfect family dinner or lunch.

From the above 5 key points, what do you understand about generalization?

The very first thing we can conclude is:

. It’s not a technique to master, instead it’s an art.

. Assumption building key skills while commonsense comes into a role at the same time.

Let’s connect with real time examples with respect to Data Science now but before that let’s discuss one of the important aspects to draw analytics or build any model.

When we look at one of the use cases that every learner would have worked on while learning this particular field is to build “Machine Learning Model whether an individual will get a loan or not”.

Most of the learners will focus on when an individual can get a loan & build a model without generalizing the situation.

For every Machine Learning or Analytics problem we have two scenario for any business related problem:

1. The first one is true or yes statement i.e. when person can get a loan

2. The second one is false or no statement i.e. when person can’t get a loan

You know what most of you will focus only on the very first statement to draw some analytics or build a model while focusing on when a person will get a loan which is somehow easy to do but to make a model? or draw some analytics when a person can’t get a loan is a difficult one.

But if we focus on “NO Statement” which in turn helps us to draw some insightful analyses.

In short, if you focus on “NO” statement at first & research for the same with respect to the problem statement will automatically help to draw insights when a person can get a loan but it’s not vice versa.

Instead figure out what are the conditions when an individual can’t get a loan will help to put insights with respect to the problem statement.

Key points to Generalized the same:

1. First step is to understand the business perspective instead of focusing on customer base data i.e. figure out how different banking sectors behave while disbursing loans to the customers.

2. Now understand how banking sectors behave while disbursing loans to different customers based on self-employed, salaried & retired etc.

3. For example figure out the important features like cibil score & research as per banking sectors affecting the same.

4. In short you don’t have to focus on customer behavior towards the bank, instead research how banking sectors behave with respect to different customers which will surely help you out to deal with your dataset.?

As per my experience this is the core fundamental to deal with any problem statement at first which will help clear your confusion about what to do with your dataset as it will help to develop deep understanding while applying commonsense at the same time.

This article is the very first with some short intro to generalized statistics & every wednesday you will find the next one while connecting with real time use cases.

Do Connect with me to join 1 to 1 mentorship session for Generalized Statistics which will help you in different ways to understand this domain , building assumption key skills & cracking the interviews as well.