GridSearchCV- Select the best hyperparameter for any Classification Model

[@akfaisel]

YT is suggesting this guys videos and they are very simple and understandable

[@SamratKar]

Thank you Krish! This is very detailed, and explains the GridSearchCV pretty clearly. It helped me a lot. Thank you again for your time and efforts in putting this together!

[@prachiarora7823]

Krish it's a very crisp n clear explanation of SVM. Really helpful and these 18 minutes are worth it. Understood the concept. Thanks

[@ayushpalak]

Such a neat explanation! Keep posting . God bless.

[@aiz_i564]

Game-changer! This is the best explanation! Thanks, Sir! God bless you!

[@harikrishna-harrypth]

Krish Naik, you are a Legendary teacher !!! Thanks much for your videos blud!

[@kingolafff7739]

All other youtube channels are a waste of time! what a well explained video ;) thanks millions of times :*******************

[@ibrahimmondal9104]

thank u sir.....now I understand how to apply best model under the specifics algo.

[@AshishBangwal]

bruh you are the Top G !!! respect

[@ahmedbouchou6893]

Thank you Krish ! Where can I find a simplified description of each model parameters. Sometimes the python documentation is very hard to understand.

[@sagaryadav3473]

Cool ! One of the best example I have seen, the way you explain is just wow :)

[@michaelschlitzer8742]

You are a good teacher! You have answered a question for me very succinctly. Thank you so much,

[@balamurugansanjeevirayar]

Very Good explanation of grid search. Clean and neat.

[@Trouble.drouble]

Superb explanation sir, how to use grid search CV for deep learning models and when to use random search CV

[@mayurkhandeshe4813]

very well explained by krish sir .....easy to understand

[@jitenkumarsahoo667]

Thanks sir......its properly explained.... couldn't find it in Google or anywhere...

[@suryameda6215]

Very neat and elegant explanation. Thank yo

[@himanshuinca]

Wow man. Great example. !! Well Explained with the example and code !

[@subodhagrawal4087]

oh..after seeing the 20 videos, I understand from your explanation.

[@adeyinkasotunde6870]

Bravo........ God bless you real good. You really imparted me with this great techniques. Well done sir. Nice one. wow.... cudos....

[@adeyinkasotunde6870]

I will love to see you teaching us on how to use XGBRegressor for example ( say House Sale price) just like the one on Kaggle.com. Second, I will love to see how to remove outliers and lastly how to normalize or standardize the data set. Thanks. Hope we will see you do something on that very soon. Thanks a lot Sir. More power to your elbow. God bless 🙏.

[@AkshaykumarPatilAkki]

Fantastik Explanation Anna... Thank you very much for the Knowledge which you are sharing with us.

[@usaikiran96]

Please make a separate video on running gridsearchcv on Random Forest algorithm.

[@saisumantabehera3891]

nice explanation Krish, how can we use grid search for multi-label classification problem

[@KieranBathgate]

Krish you're an amazing teacher

[@VamsiKrishna-vg6vd]

Very cleared explained. Thank u so much.. Keep posting more videos.

[@sudeshnadutta5702]

Krish can you please explain the difference between cross validation and gridsearch cross validation? As in how do we use cv or gridsearchcv to select among different models?

[@kirangeorge7659]

That was a really clear explanation. Thank you!

[@Mustistics]

Thanks for the video. I see you didn't take into account class imbalance, which makes accuracy not very reliable.

[@rambaldotra2221]

Mind Blowing Sir.

[@nagandranathvemishetti9247]

Hi krish sir can u make a video on applying LDA and perform hyper parameter tuning.

[@the_imposter_analyst]

this was so helpful. Been having great difficulty in parameter tuning, this has made it so much better, thank you sir

[@text-book-pages]

Sir accept my thanks. It was an amazing video

[@iscream3184]

you are a life saver

[@shivadumnawar7741]

Great tutorial

[@shaiksuleman3191]

Wow Super.No More Questions asked

[@csit3093]

You are a blessing 😊

[@sneha6999]

Hi Krish , You are doing an amazing job ,your vidios are really helpful . Could you please tell me why are we not performing sc.fit transform on X_test ?

[@thecosmiccases]

Great Explanation

[@manulavishvajith4551]

Hi Krish, great explanation. Thanks. Would you mind giving me an idea of your PC configurations, I plan to build a better PC for my machine learning projects. Basically I'm currently unable to execute high degree polynomial regressions on high dimensions. Would be a great help if you can tell me? Thanks

[@muhammadsaadmansoor7777]

Did I like this video, hell yes. Loved it.

[@arjunpukale3310]

I think in gridsearch.fit u must give X,y rather than Xtrain, ytrain coz cross validation in gridsearch will divide your entire dataset into train, test .

[@saxenarachit]

even i think so. Can you pl validate this @Krish Naik

[@arjunpukale3310]

@@saxenarachit no, i realized afterwards that u have to keep xtest for final testing on unseen data. So u can use only remaining dataset that is xtrain for grid search

[@saxenarachit]

@@arjunpukale3310 ok... In what situation we will use normal cross validation (not of grid search cv) to get the cross val score on whole data (X, y) and whats the purpose. Can you help me steps when to do cross validation on which data and when grid search cv on which data. I am bit confused here.

[@arjunpukale3310]

@@saxenarachit see 1st step is to divide your dataset into train and test. And keep the test data untouched till the end. Now you have your train data in your hand on which you have to fit your model. So now decide which model you will use to fit your train data. Suppose u select svm then use grid search on this model(use training data). And this will give you best parameter and cross val score of this model with best parameters. So you dont need to apply cross val again. Now using thise best features from grid search create your svm model and fit it with your train data. And now finally your model is created. So now test your model with unseen data that is your test data and see how well it works on your unseen test data based on accuracy, confusion matrix etc

[@saxenarachit]

@@arjunpukale3310 Thanks for this dear... One more thing - correct me where I am wrong .... 1- EDA, handling missing data, feature selection, scaling on whole data 2- Split the data for test and keep until very end using train test split on whole data 3- Applying algorithms, Imbalance techniques if needed, Handling Over/Underfitting probs. if needed, GridSearch CV to get best params on train data 4- Make the model with the best algo and best params on train data 5- Test the model accuracy with different measures 6- All Good - Deploy the model else goto 1 thru all steps except 2 to gain more accuracy.

[@shankar3109]

It's Crystel Clear... Thanks Krish..

[@panchaldp]

Good Explanation ...Thanks ...!!

[@amritanigote]

Very Helpful... Thank you!!

[@manthanladva6547]

Thanks a lot brother for the detail explanation . My topic get cleared. Thanks

[@lokeshsutrave4906]

Great explanation. Thanks for sharing.

[@sandipansarkar9211]

great explanation.thanks

[@KiranKumar-lq4td]

Nice explanation 💯

[@sharifdmd]

Very excellent detailed explanation ..

[@shanmukhasaratponugupati6308]

A very very very bigggg thanks

[@nevilparekh6400]

good explanation except difference between fit_transform() and transform() methods...

[@rahulsonvane8141]

Thank you so much, you explained it very nicely :)

[@tomtom-wv3hc]

Amazing Teacher !!!! Nice and clean explanation :)

[@lopabhattacharjee3845]

Very nicely explained. Do you have a similar video for LSTMs with hyperopt or Talos ?

[@jo1261]

Very good explanation! Thank you!

[@Vinit_Ambat]

Great video!

[@kiran082]

Thank You Krish

[@subbaraogannavarapu7405]

Hi Krish, This is amazing and i have one doubt.. what if we would like to use GridsearchCV for regression Problem? is this the same way we do for regression as well? if not, where it differs.

[@yigitsevim7741]

best vşdeo on the youtube

[@deepakgehani]

Hi krish. Can you make a video on hypermetric tuning using grid search on Random Forest Classifier

[@louerleseigneur4532]

Thanks Krish

[@kingolafff7739]

God bless you

[@yutishah1673]

Hi Krish! I have a question, while performing logistic regression when I want to perform gridsearch for hyper parameter tuning, I want to also find precision, F1 score, recall, ROC AUC, etc. So while trying to perform that gridsearch is returning me NAN values. How to handle this situation?

[@abdullahmahammadmir5541]

Really appreciate

[@nitinpatil1074]

Vary nice explanation

[@hilalkucuk5]

Excellent

[@ajaychhillar1033]

Hey, Krish please make video on Bayesian optimisation for hyperparameter tunning. Thanks in advance

[@krishnaik06]

Hey Ajay yes I will be uploading both random search and Bayesian optimization techniques in a couple of days

[@Kmysiak1]

You the man!

[@RajKumar-vm2kr]

Thank you for making this videos

[@karthickkk9427]

Thank You Krish, When GridSearchCV is performed on Random Forest, with scoring based on accuracy, best parameters identified seems to be overfit. Training data accuracy= 91% and test data accuracy=81%. Any suggestions to deal with this

[@adityadev2825]

Awesome

[@luanaleticia2644]

So great. Thanks!

[@gayatrikvr1111]

Hi Krish How do we choose values for the params_grid?

[@karanroy-vr1wn]

well explained , sir

[@_edd.ie_.o.8101]

Thank you!

[@rds6484]

you nailed it man...

[@prashantig1205]

Thank you so much! Shift+Tab is not working(jupyter notebook) for me to see the help, any settings need to do?

[@rashmir4742]

Thank you

[@awanishkumar6308]

And could you please tell me that what sections of Big Data and Hadoop is required for Data science and machine learning

[@pratheeeeeesh4839]

Wonderful !

[@sujithkumar_ga]

U r a genius bro

[@sujithkumar_ga]

Need u r help !.. am doing an internship they gave me task .. it would be very helpful if u help me plz.. give u r mail id . So that i can contact you

[@bajajchirag123]

How do we get to know that the provided range is the correct? For eg. in the given case, you used the range from 1 to 1000 for C value and for gamma the range was from 0.1 to 0.9. why we haven't taken the range to be .001 to 12130 or anything else for C values and similarly for the gamma values. and there are so many other parameters as well but we considered only these 2. Currently, I am trying to use this gridsearchCV on a linear regression model. then what should be the param_grid values I should take. Please provide a pseudo code and explain if possible. Thanks in advance.

[@subodhagrawal4087]

brother, this was just an example. I real world there will be 100s of values.

[@arjunpukale3310]

U must know the math behind it

[@write2ruby]

Nice

[@sandeepnigam757]

Is this good technique if we are applying feature scaling on test data??

[@mickaelsgro3370]

hi , please how do u chose "cv=10" in GridSearchCV ? Thanks a lot

[@NR_Tutorials]

thanks ... this optimisation for any classification ..?

[@maYYidtS]

no....every model has different non-default parameters. ex. knn has n_nighbours=10...

[@Ash-bc8vw]

Hello, can you suggest a good laptop for running machine learning codes Or the specification

[@zee4654]

Sir please can you provide a link where to I find the freight travel time prediction Dataset ??? 😔

[@kamal6762]

I have one doubt that why we only transform the X_test data set not fit first or we have to use fit data(mean and SD) from the X_train?

[@teacherHub6723]

Thanks u sir

[@kushalminachi445]

Can we use RandomizedSearchCV instead of gridsearchcv?

[@awanishkumar6308]

Hello sir sorry to ask, Here we have fitted the model without scaled features (I.e- X_train) then why you have scaled the features using StandardScaler??

[@harpreetsandhu7697]

can we use it on naive bayes algorithm

[@chinnam8053]

Too good..can u please help me in real time?? How can I contact u personally??

[@maYYidtS]

is that necessary to fit (x_train,y_train)again instead of fit(x,y) at 14:15 because the cv parameter will automatically split the data right?

[@oozzar2841]

Is this same for multi classification SVM or not?

[@AmitYadav-ig8yt]

Sir, You selected some values 10, 100, 1000 in Dictionary - How did you get these values for these parameters, Are they predefined or any ways to select these values?

[@krishnaik06]

No it is not. I have randomly selected it...you can put ur own values

[@AmitYadav-ig8yt]

@@krishnaik06 Okay Sir, Thank you very much.

[@ahmedbouchou6893]

Thank you Krish ! Where can find a simplified explanation of model parameters. Sometimes the python documentation is hard to understand.

[@junaidlatif2881]

Sir. If after scaling x_train, i build model. Now if i have validation data, (few new samples to check prediction). Now should i scale my sample data? Or should I do scaleback my X_train first? Then validate sample data?

[@junaidlatif2881]

6:15

[@ajaykushwaha-je6mw]

Hi Sir, after running this code: classifier.fit(X_train,y_train) you are getting various parameter in o/p section but i am getting just one. why sir ?

[@khanmohammedaamir8900]

But how to know , which parameter we can pass and what type of parameter is not important ?