Handling Imbalanced Dataset Using Cost Sensitive Neural Networks- Credit Card Fraud Detection
Рет қаралды 14,674
3 жыл бұрынgithub: github.com/krishnaik06/Imbala...
Unacademy is back with a yet another exciting series for all the ISRO aspirants!
⭐Join the FREE special classes, build your concepts and then take ALL INDIA ISRO MOCK TEST for FREE. Not just this, the top winners will be awarded exciting prizes. The test is curated to test your technical knowledge and will consist of 80 technical questions needed to be solved in 90 mins.
Tik-Tok the clock is ticking, don't let this slip away from your hands!
ENROLL NOW: bit.ly/2GUsDOJ
FREE ISRO PYQ Classes link
Mechanical: bit.ly/3pnXBjp
Civil: bit.ly/3kqY1SB
Electrical: bit.ly/35pLdYc
CS/IT: bit.ly/36vOMvo
ECE: bit.ly/36siB02
All Playlist In My channel
FASTAPI playlist: • FastAPI Deployment Tut...
Interview Playlist: • Machine Learning Inter...
Complete DL Playlist: • Complete Road Map To P...
Julia Playlist: • Tutorial 1- Introducti...
Complete ML Playlist : • Complete Machine Learn...
Complete NLP Playlist: • Natural Language Proce...
Docker End To End Implementation: • Docker End to End Impl...
Live stream Playlist: • Pytorch
Machine Learning Pipelines: • Docker End to End Impl...
Pytorch Playlist: • Pytorch
Feature Engineering : • Feature Engineering
Live Projects : • Live Projects
Kaggle competition : • Kaggle Competitions
Mongodb with Python : • MongoDb with Python
MySQL With Python : • MYSQL Database With Py...
Deployment Architectures: • Deployment Architectur...
Amazon sagemaker : • Amazon SageMaker
Please donate if you want to support the channel through GPay UPID,
Gpay: krishnaik06@okicici
Telegram link: t.me/joinchat/N77M7xRvYUd403D...
Please join as a member in my channel to get additional benefits like materials in Data Science, live streaming for Members and many more
/ @krishnaik06
Please do subscribe my other channel too
/ @krishnaikhindi
Connect with me here:
Twitter: / krishnaik06
Facebook: / krishnaik06
instagram: / krishnaik06
#HandlingImbalancedDataset
@RedShipsofSpainAgain 3 жыл бұрын
Hi Krish, good video on class weights. One thing to I'd suggest is to use the `stratify` parameter in the train_test_split() function: scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html This will ensure that your dataset is stratified keeping the same proportion of class_0 and class_1 in both your training set and your test set when you do the split. This also ensures you keep the same class_weight that you calculated with your calculator from the df['Class'].value_counts() , which showed 577:1 ratio of classes, versus when you calculated the proportion of y_train.value_counts() when you get a different ratio of 559:1. So, TL;DR, just use the `stratify` param in your split to maintain the same class proportions in your train set and your test set.
@suneel8480 3 жыл бұрын
So sir plz make video on how we can use F1 as metrics in model.compile! Keras do not provide it by default. Using F1 score we can make sure both fn and fp are less
@hakanbozcuk7761 Жыл бұрын
Just amazing work, thank you, sir!
@thetruereality2 2 жыл бұрын
A classification report would have explained exactly how the model performs in classifying the classes. Maybe that would've helped
@arjyabasu1311 3 жыл бұрын
Awesome video sir!!!!
@pusprajkumar1058 2 жыл бұрын
Thank you 💞 for this video
@suneel8480 3 жыл бұрын
Sir when using roc_auc it seems to be high even when we have too much false positives .Using F1 score can help to see if both fp,fn are less or more . So sir plz make video on how we can use F1 as metrics in model.compile! Keras do not provide it by default.
@meeradevi4432 3 жыл бұрын
Use tfa.metrics.F1Score or make custom function for f1score
@abc76485 3 жыл бұрын
Hi krish it would be awesome if ineuron launches an affordable course on A-Z of data structures and competitive prog. With all the mathematics included.
@nilakantas5153 3 жыл бұрын
Thank you sir
@pubgkiller2903 3 жыл бұрын
What is different between over sampling and Weight handling technique ? And which is best fit for which conditions ?
@deaglegaming3119 3 жыл бұрын
When he said titan rtx i was so exited which one graphics you have?
@mukarramali2958 3 жыл бұрын
hey, love the way you explain. I’m interested in buying the new MacBook with M1 chip, What would you suggest me to go for? The MacBook Air (8Gb ram 256ssd) or MacBook Pro (8Gb ram 256ssd)?
@vishaldas6346 3 жыл бұрын
Hi Krish, I have a doubt. I am facing an issue where my system is not calculating batch size, it uses the entire training set in model fitting. Suppose in model fitting it shows "Train on 199364 samples" and the epoch should have been like 199364/32 ~= 6231, but its still training on 199364 samples. By default our batch_size = 32, if not provided in model.fit.
@maryamzeinolabedini1515 2 жыл бұрын
Hi, Thank you. How can we apply cost sensitive on bayesian network for prediction of imbalanced data?
@Asma-cx8uc 2 жыл бұрын
Hello Sir ! Could you please describe how SMOTE technique can be used to balance data images
@pranjalsingh1287 Жыл бұрын
Knish. Can you tell the ideal % of 0&1 classes to say them as imbalanced
@yashbohra2593 3 жыл бұрын
SIR , how to apply the the same for multi-label classification ?
@Maryashahere 2 жыл бұрын
Sir for dataset size, 44 unbalanced, how much have to be the test rain split. For that if we apply DL, is there problem?
@leamon9024 Жыл бұрын
Can I use sample_weight instead of class_weight in your example?
@tirthadatta7368 2 жыл бұрын
Can we use same basic for multi-class classification problem??
@leenavig3441 2 жыл бұрын
if we run the ANN second time ...the scores are different ..if we print classification report the f1 scores are quite different..canu explain how to deal with this problem
Unfold Data Science
Рет қаралды 17 М.
Krish Naik
Рет қаралды 21 М.
Dr. Data Science
Рет қаралды 8 М.
Krish Naik
Рет қаралды 42 М.