Machine Learning - Real Life Application : Credit Card Approval

We are going to discuss one real-life application that is credit card approval

We have selected this very real-life application so that you can easily understand what is

the case study and how to implement it using our machine learning

So credit card company received thousands of applications for new credit cards and for each

and every application it contains information about the applicant like your age, marital

status, salary, outstanding debts and the credit rating, etc. So, problem here is that decide

whether credit card application should be approved and that is classify applications onto the two

different categories. One is the approved and the one is the not approved and rejected

So two classes are there. So in the outcome we are having two classes. One is approved and

the one is not approved. Data. a set of sample data records, example, the cases, case instances, training data set

about applicants, and described by key attributes or features, that is A1, a2, dot, AK

So, obviously, you should have a dataset. The data set will be gathered from the applications made by the applicants, and here this

particular data set will be having key attributes of features, that is A1,82, dot, AK

Classes. each sample is labeled with a predefined class and the available data set D is divided into two

disjoint subsets. So there is no overlapping their exclusive. So the training set that is a detrained for learning of the model and the test set that is

detest for testing the model also called the hand holdout set So one is the holdout set which will be used for the testing purpose and another one is the trend set which will be used for the testing purpose Another one is the train set which will be used for the training purpose and represented by

detrain. Goal, to learn a classification model that is from detrain that can be used to predict

the classes of detest or any future cases or instances. So now here we're having this

detrain on which the model will be trained and then. it will be asked to do the forecasting or prediction on the D test and from there we can

also evaluate the performance of the model. So learn there is a train, train the model

So learn a model using training data and computer system learns from data D to perform

tax T if after learning the system's performance on T improves by the measure M

So on this particular training set, the model will learn, the model will be able to learn

from the training set and now later the model will be used for the prediction

So test model accuracy, test the model using the unseen test data to assess the model accuracy

So assumptions, distribution of training, that is a sample data, is identical distribution

to the test data and future unknown data. Otherwise, the classification accuracy will be poor

That means the distribution of the training data set and the distribution of the tested

data set they are alike. So that's why we are going to, so we can select this particular tested data set very randomly

from the mother data set so that there should not be any biasness

If there is some biasness, then obviously the prediction accuracy will be too poor and

obviously the model will not be reliable and will not be usable So accuracy is equal to number of correct classifications by total number of test cases So depending upon the total number of test cases how many test cases have got correctly predicted

So the ratio that is the total number of correct classifications and total number of test cases

the ratio will give us the accuracy of the model. So training data, then learning algorithm

then the model will be formed so this is our step one also known as training

and then model will be formed and then test data step two that is a testing

then the accuracy can be measured so let us consider this particular sample

data set and it is containing ID age has job has own house credit rating and

class so here having multiple attributes are there and we are having multiple data sets are there

Now, from this particular data set we are supposed to form one classification model

So here is one model which has been suggested using our decision tree

So age young has job true or false in case of true we are having two by two in case of

false we're having three by three. What do they mean actually? So, age young and has job true

Just find it here. So age young, that means this part, and has job true, means this two cases only

So young age and has job true. So here you are finding this yes

So here I have been finding for the both cases we are finding yes. So how many times this combination has occurred two times and for the both cases I'm finding

yes So that why it is 2 by 2 In case of false that means has job falls age young and has job falls it is 3 by 3 Why 3 by 3 And here we are going for this no and 3 by 3

Why? Because you see, here we can do in this way. So age young, so this much, this much, and has job false, we are having three cases

And for the both cases, we're having no there. So that's why 3 by 3 has been written there

Similarly, let me go for the another one. So age middle and own house true, here we have written three by three

Why this three by three? So, let me go for the cross check. So own house is true and age middle

So age middle means we are having these five cases and own house is true we are having

here three cases and in both the cases we are having the yes, that is the class is yes, yes

the credit cut can be approved. So, as a result of that, we're having this three by three for years

So in this way, if you go for the cross check, you can make the video pause here, and

then also you can make the video pause here. You can go for the respective checking

Now what is the query which has been given to us? What is the respective new data set given to us for the prediction

So age young, has job false, own house is false, credit rating is good, then what is

the respective class. So, remember, age, young and has job is false. So, just remember

this one. So now here you find that age is equal to young and has job is equal to false

So the credit application, the credit card application should not be accepted, should be

rejected. So no is there. So in this way, we can do the prediction for that new test data

So, in this way, in this particular case, we have shown you one real-life application on

credit card approval. Thanks for watching this video