Project 5. Loan Status Prediction using Machine Learning with Python | Machine Learning Project

Hi! I will be conducting one-on-one discussion with all channel members. Checkout the perks and Join membership if interested: / @siddhardhan Check membership Perks: / @siddhardhan
. This video is about building a Loan Prediction system using Machine Learning with Python. Machine Learning Project with Python.
All presentation files for the Machine Learning course as PDF for as low as ₹200 (INR): Drop a mail to [e-mail]
Enroll at One Neuron to learn from 100 courses in one subscription with 5% discount: courses.ineuron.ai/neurons/Te...
Hi guys! I am Siddhardhan. I work in the field of Data Science and Machine Learning. It all started with my curiosity to learn about Artificial Intelligence and the ability of AI to solve several Real Life Problems. I worked on several Machine Learning & Deep Learning projects involving Computer Vision.
I am on this journey to empower as many students & working professionals as possible with the knowledge of Machine Learning and Artificial Intelligence.
Hello everyone! I am setting up a donation campaign for my KZbin Channel. If you like my videos and wish to support me financially, you can donate through the following means:
From India 👉 UPI ID : siddhardhselvam2317@oksbi
Outside of India? 👉 Paypal id: [e-mail]
(No donation is small. Every penny counts)
Thanks in advance!
Let's build a Community of Machine Learning experts! Kindly Subscribe here👉 tinyurl.com/md0gjbis
I am making a "Hands-on Machine Learning Course with Python" in KZbin. I'll be posting 3 videos per week. 2 videos on Machine Learning basics (Monday & Wednesday Evening). 1 video on a Machine Learning project (Friday Evening).
Dataset File: www.kaggle.com/ninzaami/loan-...
Colab file: colab.research.google.com/dri...
Download the Course Curriculum File from here: drive.google.com/file/d/17i0c...
LinkedIn: / siddhardhan-s-741652207
Telegram Group: t.me/siddhardhan
Facebook group: groups/49085... Instagram: / siddhardhan23

[@Siddhardhan]

Hi! You can join this Telegram group for regular updates about my videos: t.me/siddhardhan Thank you!

[@uppariamericavlogsjustforf3395]

Can I have data set of this video?

[@Siddhardhan]

You can find it in the description of the video

[@sanjayyv6888]

can you share the ppt

[@Siddhardhan]

I am sorry... I cannot share the ppt

[@piyushtale0001]

Thankyou Mr. Siddhardhan I would like to tell all that we can use logistic regression for the same as it gives same accuracy like SVM and logistic is faster in computing

[@taijosephinedanielle2962]

God bless you for distinct explanation... I had a total understanding of every step Thanks brother

[@baneledludlu7983]

Amen

[@shilpa2627]

Your videos are very helpful.

[@nileshchaudhary1643]

# This is the solution of exercise given by sir, # hope it helps # sample input from gender to Property_Area = 1,1,4,1,0,3036,2504.0,158.0,360.0,0.0,1 Input_data = (1,1,4,1,0,3036,2504.0,158.0,360.0,0.0,1) Input_data_as_numpy_array = np.asarray(Input_data) # reshaping the data as we are pridicting for one instance input_reshaping = Input_data_as_numpy_array.reshape(1,-1) x = classifier.predict(input_reshaping) print(x)

[@raj4624]

Good

[@Siddhardhan]

That's really great, Nilesh! All the best 👍

[@srikanthreddymanneti7116]

hi is it working fine for u?

[@fahaamshawl9335]

@@srikanthreddymanneti7116 bro did you find any solutions?

[@srikanthreddymanneti7116]

@@fahaamshawl9335 nope have u found?

[@shlokkumar6257]

sir, while making model as a input when i am feeding data to predict its having textual data, so i just want to know that do we have to import libraries like Tfidfvectorizer or something?

[@kirantodekar3481]

on what basis you have selected the SVR model ? please specify

[@abhishekmishra1313]

can you tell me why logistic regression model wasn't used in this project?😕

[@AmitTiwari-td4zj]

Hi @siddhardhan, nice explanation., But it will be great if you do more EDA part like krish naik do by taking some difficult data set ,like how t to find outlier, how to remove, box plot, null value treatment , most of the project your approach is same , after practicing also from your video ,when we go to solve real world problem we are facing difficulties

[@mvs69]

when replacing dependent values, it's assigned to new variable, 28:45 but before when replacing loan status no assigning,, why ?

[@ComputerScienceSimplified]

Awesome video, keep up the incredible work! :)

[@Siddhardhan]

Thanks 😇

[@puneetkorjani6648]

do we need to standardize the data before prediction? pls answer

[@chidichukwumezie8438]

Thank you for the video. Great lesson. However, I do have a question. Why was feature scaling not done on the dataset considering that some features like ApplicantIncome, CoapplicantIncome, LoanAmount, etc have higher values (weight) than the rest? I also know that the performance of the SVM model is affected by feature scaling.

[@felipemaldonado1057]

Scaling may make the algorithms works better. The vid contains just the most important or essencial things in order to make the models works. It is an educational vid and scaling is an advance thing but obviously recommended when necessary.

[@premalathas623]

Clear explanation.. keep up the good work!..

[@Siddhardhan]

thank you so much 😇

[@nishitagupta1349]

Sir can you help us with some doubts in making predictive system for loan status prediction using ML ?

[@jawehers]

Hi Siddhardhan, why didn't we use Logistic Regression here pls ?

[@LoneWolf-rj1px]

Thank you for also showing the data visualizations in this ML Project. However, I have 1 question, do we need to do these Visualizations every time while we need to do the ML, or is it just an optional thing just to show the relationship of the data?

[@rohanshah8129]

Data Analysis helps you in preparing a good Data Preprocessing Pipeline.

[@growingfire]

Thanks a lot !

[@sahilsvachhani]

Why did we separated labels with data?

[@gauravthorave1136]

Sir why we use linear regression ,why we don't use logistics regression can you plz tell or guide

[@mc_abah]

Sir pls I need a quick answer. Am having a problem with the prediction system. Am not having an accurate results it's seem the model is not learning from the data. The model is predicting '1' for all input data even when it's supposed to be '0' ie for loan rejected

[@rohanshah8129]

Some issue with your procedure

[@evavashisth9103]

My code is working absolutely fine but I’m unable to understand the reasoning behind the “Train Test split” why exactly are we doing that? And why did we do those four different variables, xtrain test, train, test? Could you please elaborate why we do that?

[@sandipansarkar9211]

finished watching

[@sapawar007]

Hello Sir the predictive system isn't working since we have the attributes as text,cat as well.Please help with the predictive system for this example

[@rohanshah8129]

Convert the text manually to the encoded values after you get input from the user. Like if user specifies male, change the gender variable to 1.

[@totosy]

can you send the link for the application test data

[@ahmedabid6799]

big thanks

[@mysteriovvn]

Hi, one question: The semi_urban showed high chances of getting loan approved than other two, so was it more appropriate to give it the weightage with value=2 when you replaced the category. Thanks in advance.

[@Siddhardhan]

hi! that's a really good insight. we can do that.

[@snehacookie4138]

Bro is dng this project as final project is good or not bro can u pls say

[@sumankundu3232]

in this project could we use logistic regression ???

[@LoneWolf-rj1px]

Can we import Label Encoder from Sklearn and do the encoding? How can we know what is labeled what by the label encoder?

[@rohanshah8129]

Use the value_counts() to see the unique categories after encoding is done

[@ruchibhadauria4977]

if we encode categorical variables with numbers like 1, 2 wont the model find pattern in it?

[@Siddhardhan]

hi! if we 2 or 3 categories, then it is not a problem. if we have more categories like 6 or more than that, we can use "One Hote Encoding".

[@kuldeepsinghdudi2679]

Great work 👏

[@Siddhardhan]

Thank you so much 😇

[@karthikeyatsvr3838]

how do we approach if we were given with traindata and test data separately, instead of the one dataset??

[@rohanshah8129]

Split would not be required in that case and the same Feature Engineering Steps must be applied on the test dataset too before the prediction.

[@rpml7939]

Hey, when you are converting the Y & N to 1,0 you will get the type error (Cannot compare types 'ndarray(dtype=int64)' and 'str'). To fix this issue you need to change the type of Loan_Status column to string or object.

[@kumarishikha8490]

Can you please provide the code for it ? How do we change the type?

[@felipemaldonado1057]

@@kumarishikha8490 you can use this as model. df['column_name'] = df['column_name'].apply(lambda x: '1' if x == 'Y or N up to you' else '0')

[@kumarishikha8490]

@@felipemaldonado1057 Thanks

[@HarshitSingh-tg9yv]

There are 614 data points in this example. Dropping all rows with NaN values leaves only 480 data points. Thus, a significant proportion of data points is dropped. Shouldn't we try another method like filling mean values in numerical columns, etc.?

[@dondata718]

I tried this and it didn’t change the accuracy score that much

[@rohanshah8129]

You can treat the Credit_History NaN values as third category. Use ordinal encoding here. 0 for bad (0.0), 1 for nan, 2 for good (1.0) NaN values here indicates that the applicant is applying for the loan for very first time. For other features, all of them have like about 5% missing data. So that won't really affect the model as much as you might think.

[@stephenmascarenhas3595]

Sir ur teaching is awesome But why can we do like this to get accuracy classifier.fit(x_train,y_train) classifier.score(x_test,y_test)

[@Siddhardhan]

Hi! you can do this as well.

[@madhumithan6653]

Hi sir, Thanks for the great effort.Can u plz explain which model (linear or logisticregression or svm or random forest )has to be chosen at what criteria for the data,as i m confused to chose which model.

[@symbiontacademy9232]

It depends on the data, I guess there isn't any hard and fast rule which works everywhere, as this is a classification problem so it rules out linear regression. You will need to use several models and then you will observe which model is performing better than the others. Keep watching other projects, every video will hit a new concept and by the end of the playlist, you will not have any such doubts.

[@madhumithan6653]

@@symbiontacademy9232 Thanks for the reply

[@rohanshah8129]

Logistic should be pretty fine to work with as its a small dataset + binary classification. If this doesn't work... then you can train multiple classification algorithms and compare which one is best and then fine tune it.

[@shalinilokeshgowda]

Hi Siddhardhan, this is great. May i know how to change the code if i need to read a pdf file and excel file and then load into dataset?

[@rohanshah8129]

You can use pd.read_excel(excel_path) For pdf, it will be a bit long to explain here.

[@funtime12345]

Hello sir, I have the same dataset with more columns and around 30000 rows and need to predict the loan amount where I am creating a Regression model. But the problem is handling the missing values. I am a bit confused about whether to replace it with mean, mode, or median in place of missing values. Sir can you help me with this as stuck with this problem.

[@Siddhardhan]

hi! watch my video on handling Missing Values

[@funtime12345]

@@Siddhardhan sir I cannot find exact video for it on your playlist. Can you give me the link please.

[@rohanshah8129]

kzbin.info/www/bejne/fau9npmbZZhjhrc

[@samuelpaul5923]

I guess you should have used StandardScaler(), coz some of the columns values are 1, 0 and some of columns contains above 3000 or 4000 values. Let me know if I'm wrong..

[@vanshdoshi3283]

same thought... @siddhardhan pls tell its good to use standardscalar here or not?

[@rohanshah8129]

You guys are right! @samuelpaul5923 and @vanshdoshi3283, you can surely try to use the StandardScaler. You basically check the spread of every feature and scale the one with pretty high spread. Here, as much as I could see, there is not a LOT of spread in the numeric data. Hence, the scaling won't contribute a lot as you may think. Still give it a try and let us know.

[@sandeepthukral3018]

@@rohanshah8129 I tried scaling column that is income but the accuracy decreased

[@ram9208]

1) why are we not doing feature scaling in this dataset given features are of different measures. 2) in case we have another column here lets say profession which contains more than 20 different professions, how would we convert that into numerical? Thanks for the video though Siddhardhan

[@sejaljamwal6773]

for the second one we'll do one hot encoding i guess

[@ashusingh8280]

Hello Why didn't you use another classification algorithm? Can we use any other classification algorithms as well?

[@rohanshah8129]

Surely. Try Logistic Regression for example, it shall work as good as what you have seen here as it is a binary classification problem and dataset is also small.

[@movieskuchhatkey3817]

i am not able to make predictive system pls help to create

[@samikshagodghate910]

i can say this person is one of the besttttttttttttttttttt person everyone needs to get to learn data science!!!!!!!!!!!

[@LoneWolf-rj1px]

How do we understand this Loan project will work best on SVM and not Logistic Regression?

[@rohanshah8129]

He just took that as an example. Logistic Regression will work as good!

[@amoldusane9851]

Sir I faced problem when I fit the data.... there is a value error: could not convert string to float: 'LP001431' PLEASE RESOLVE THE QUERY

[@rohanshah8129]

Drop this column of LOAN ID

[@techwithsufi5674]

Which software you are using for making the slides ??

[@Siddhardhan]

I am just using MS PowerPoint.

[@techwithsufi5674]

@@Siddhardhan all right thanks and your English is quite good I also speak English but hesitates a little bit

[@Siddhardhan]

Thanks for your appreciation 😇 practice more... Eventually you will be good at it. All the best!

[@techwithsufi5674]

@@Siddhardhan yes sure

[@namithashri]

Hello sir. This video is very much helpful for me as I'm doing my final project based on the same concept.. But the problem is, my faculty asked me that in the verification process cibil score is checked and why are these required.. can you help me answer that.. expecting your reply...

[@PraveenKumar-ob7gd]

Hellooo am doing same project help me

[@rohanshah8129]

credit history has a direct impact on cibil score better the credit history better the cibil score so the credit history feature shall do that task for you.

[@namithashri]

@@rohanshah8129 thank you

[@arsavarthiniamu2242]

Y u r using some model can use decision tree random forest any particular reason

[@Siddhardhan]

hi! each model has its pros & cons. refer my video on model selection for more information.

[@_poodsiepie_258]

Hey can you please give an example of input data for predictive system. Any one example.. plz

[@Siddhardhan]

hi! please refer diabetes prediction and other project videos. it's the same procedure.

[@aazimkhan1753]

In input_data , we should not add loan id and loan status right?? BTW love your video :)

[@Siddhardhan]

Yes, correct. and thanks 😇

[@sujeetsharma8791]

giving error please help

[@kishanpoojary1761]

Sir I got error that couldn't convert string to float: 'semiurban' What is the problem

[@kishanpoojary1761]

I followed all the steps in ur vedio sir

[@jhonceenaskt]

where is the dataset?

[@manjumankoji1448]

is it possible to SVM algorithm plot into the graph..

[@Siddhardhan]

hi! go through this documentation

[@amalkuttu8274]

why every feature is selected for building the model..is there a process called feature selection..I am a student doing a mini project in ML. so please sombody answer this.

[@Siddhardhan]

Hi! I wasn't focusing much on data analysis and feature engineering part as this is a machine learning course. Those topics will be taught in the data science course. The idea here is to show the implementation of ML models. But I would advise u to do all the feature engineering part if u r working on an end to end project.

[@amalkuttu8274]

@@Siddhardhan thanks fot the reply

[@mix_dj5580]

Good video

[@Siddhardhan]

thanks 😊

[@boddunageswari1945]

Sir please tell us the code for making a predictive system for this case 😢

[@rosan5657]

Bro if u r from Itahari International College then plz skip this project bcz i have already copied it ;))

[@twilightlifestylepriya]

Hi can we use LogisticRegression model for thie same

[@rohanshah8129]

yes

[@tusharkhan3133]

Brother can you tell me what I have done wrong here? My program always shows that , the person won't receive any kind of loan no matter whichever parameter I feed it. input_data = (1,1,1,1,0,4583,1508,128,360,1,0) # changing the input_data to numpy array input_data_as_numpy_array = np.asarray(input_data) # reshape the array as we are predicting for one instance input_data_reshaped = input_data_as_numpy_array.reshape(1,-1) # standardize the input data std_data = scaler.transform(input_data_reshaped) print(std_data) prediction = classifier.predict(std_data) print(prediction) if (prediction[0] == 0): print('The person will receive loan') else: print('The person will not receive loan')

[@Siddhardhan]

hi! you shouldn't use standard scaler as we have lot of categorical columns in our data. do the code without standardizing the data.

[@samarapires7631]

Were u able to get this ? I am a bit confused

[@emmanuelnyangweso9389]

In my case, i have not used standard scaler but the code is till having some problems since input_data = (1,1,1,1,0,4583,1508,128,360,1,0) shows person won't receive loan

[@fahaamshawl9335]

Did you find any solutions for input data? Having issues with the inputs to predict

[@rohanshah8129]

use sclaer only on numerical features

[@pradeeprayapati8667]

Sir i want pdf file sir, to learn..

[@incomegrowth18]

Sir can you plz share ppt on it

[@sivakumar.s2559]

Tell how to make the predictive system for this model the last part

[@sivakumar.s2559]

Pls post in soon are send any link

[@vishvapatel6876]

hyyy anyone please help with making predictive system provide the code

[@ishwarya22]

Can you please provide PPT of this

[@rishipatwa6823]

predictive system input_data=(1,1,4,1,0,3036,2504,158,360,0,1) #changing th array to numpy array input_data_as_numpy_array=np.asarray(input_data) prediction=classifier.predict(input_data_as_numpy_array.reshape(1,-1)) print(prediction) if (prediction[0]==0): print('not approved') else: print('approved')

[@aqsasaif8847]

any one here can share the code of the predictive system as well?

[@saikirana9629]

Did you get the code? @aquasaif8847

[@incomegrowth18]

Whu u not fill the missing value

[@aravindvempati6113]

i have a doubt, why feature scaling not done

[@rohanshah8129]

the spread of data is not huge, hence scaling won't really help our model here. However, if you try it there should not be any issues.

[@yallanurunaveenkumar3422]

Sir please provide how to implement logistic regression for loan approval prediction

[@Siddhardhan]

hi! it's the same procedure. watch my other videos to learn how to implement logistic regression

[@rohanshah8129]

Just change the SVM with LogisticRegression() and it should work just fine.

[@dolutarakesh2849]

sir can i get ppt you made so i can understand plz sir

[@Siddhardhan]

hi! I am sorry. I cannot share the ppt.

[@ryanranjith672]

Did anyone found out how to make predictive system?

[@lahuhavmore6803]

Why used svm here

[@pavanporika4135]

Sir u said I will do videos on interview questions in loan prediction but u didn't started yes sir makeee I have no time

[@Siddhardhan]

hi! I cannot make interview questions specifically for this project. today evening, I'll upload a video on interview questions for Machine Learning. it will be a general topic.

[@pavanporika4135]

@@Siddhardhan k sir can u tell me if interviewer ask me out of logistics regression and support vector machine which is better and you didn't tryed with decision tree... And other algorithms?? How to answer

[@rishabhmohata9268]

in spite of using replace we could have simply gone for get_dummies ...isnt it ?

[@Siddhardhan]

yes, you can try it.

[@rishabhmohata9268]

@@Siddhardhan I just finished up this project and accuracy certainly dropped to 71.27 to yours 80 ...

[@shubhamkhandagale1366]

Need help for same project

[@Siddhardhan]

hi! mail to [e-mail]

[@rohit-qx8jn]

is this project complete

[@Siddhardhan]

complete in the sense?

[@radhikataldar]

Need your help for such projectd

[@Siddhardhan]

Hi! You can contact me in social medias.

[@syedazharmohiuddin3007]

X_new = X_test[0] # Reshape X_new into a 2D array X_new = X_new.reshape(1, -1) prediction = classifier.predict(X_new) print(prediction) if (prediction[0]==1): print('Loan Approved') else: print('Loan Denied') prediction system for Loan Status

[@sandipansarkar9211]

finished coding

[@boddunageswari1945]

😢😢

[@abhishekmishra1313]

gender can"t be segregated into 0's and 1's 😂 what if the person is transgender lol

[@sahilsvachhani]

haha then assign 2 for transgender

[@vikrantspeaks]

I wrote a prediction system: #making a predictive system #Convert the dataframe to an array X_new = X_test.iloc[39].to_numpy() #reshape array X_predict_reshaped=X_new.reshape(1,-1) prediction=classifier.predict(X_predict_reshaped) print(prediction) if (prediction[0]==0): print('Not eligible for loan') else: print('Eligible for loan') #check the actual value Y_new = Y_test.to_numpy() print(Y_new[39])

[@ryanranjith672]

Thanks