Interview Prep Day3-How To Prepare Support Vector Machines Important Questions In Interviews🔥🔥

In this video we will be understanding the important interview questions that are usually asked regarding LSupport Vector machines
It is must for every data science aspirants
github: github.com/kri...
Data Science Interview Playlist: • Data Science interview...
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 for AI Tech news
/ @krishnaikhindi
Connect with me here:
Twitter: / krishnaik06
Facebook: / krishnaik06
instagram: / krishnaik06
#datascienceprojects
#machinelearninginterviews
#deeplearninginterview
#datascienceinterviews
#computervision
#nlp

[@kshitizomar6730]

Classification Algorithms such as Logistic Regression and Linear SVM (kernel= 'linear') which try to find an optimal hyperplane are generally prone to imbalanced dataset. The reason behind this is these linear classifiers are actually to solve their respective optimization problems in presence of a tradeoff, which is: Greater Generalization Power Vs Minimal Classification Error. Hence if the dataset itself is imbalanced, it will be only solving the optimization problem for the points of the majority class, given we haven't done any hyper-paramter tuning. Now for some aggressive hyper-parameter tuning, we get these classifiers to perform okayish for imbalanced dataset (tuning for minimising error over maximising generalization power), but then we will be sacrificing the Generalization Power for unseen datasets, which isn't desirable, also hyper-parameter tuning won't work for severely imbalanced datasets. Now, for Similarity based algos such as KNN are also prone to imbalanced datasets, as they employ a majority voting at the end to make their classification decision. Kernel SVM shouldn't be prone to imbalanced dataset, because it finds suitable optimal support vectors using the kernel itself, that is if the dataset is imbalanced, then their will be less number of support vectors for the minority class. (Please do correct if I am wrong on this one) Naive Bayes isn't affected by the imbalanced as it calulates the likelihood of each (seen) features during runtime. Tree based algorithms such as Decision Tress and Random Forest should also not be affected by imbalanced datasets, because they divide the given feature space into axis parallel hyperplanes. (Also, they try to maximise the information gain) Also, I think that MLP based algos would overfit the training dataset if it is imbalanced, hence making them immune to imbalanced datasets. This is a guess, please correct if it's wrong. I am certain that some of which I have said maybe rubbish, please correct if it really is! Great series Krish Sir!

[@peddaboinatribhuvan7749]

I think If Imbalance is present in Dataset itself(Even if Intentional for Experiment) No model would perform better as Bias is present in data itself .

[@kalyankrishna5902]

Svm, logistic regression, random forest classifier, decision tree classifier, knn, naive Bayes are affected by imbalanced data set

[@ritajyagupta4666]

all top of the line supervised learning algorithms such as Linear ,Logistic,KNN,SVM are affected by imbalanced datasets.

[@hardikvegad3508]

All the ml algo which contains gradient descent.... Are impacted by Imbalanced dataset... Such as ANN, Logistic regression, Linear Regression... Etc and KNN also

[@InspireXAI]

support vector machine, Xgboost, LightGBM, KNN, regression Algos can be impacted by imbalance data set .

[@LuciaCasucci]

Krish, you can tune the gamma to prevent the outliers to influence too much the SVM algorithm

[@utkar1]

Algorithms effected by imbalanced dataset - Linear Regression - Logistic Regression - KNN - SVM etc

[@Roblox5091]

can you explain the reason for it in short?

[@shashankshashi6361]

Because they are supervised learning.where target varaible are imbalced that makes over fitting of data

[@pankajnegi9278]

All are these algorithms are using distance as a metric, be it Euclidean distance or Manhattan distance,

[@basavarajbijali7342]

All ML algorithms deals with linear seperability and linear relationship are affect by imbalance dataset

[@siyays1868]

Thanku Krish so much for a valuable guidance, providing everything. Nice interview tips.

[@shahnawazkhan1636]

You are motivational like vivek bindra sir but in the domain of Data science.

[@manishsharma2211]

Bounce back

[@shahnawazkhan1636]

@@manishsharma2211 ya

[@arrigocoen7856]

Great video! A tiny recommendation would be to add it to the playlist of Interview Prep.

[@teluguforyou3992]

Hoo sir please make a video on how projects will be in product based companies(like how the tasks will be give to team, how is the deadline for task...)

[@rb7323]

Algorithms that rely heavily on accuracy metrics like Naive Bayes or k-Nearest Neighbors (kNN) can be easily fooled by unbalanced data.

[@rohanbura8978]

Logistic Regression, KNN affects most to the imbalanced data set

[@SahilKhan-yu3oh]

Sir you are great thanks you so much

[@anshusingh536]

krish sir SVM AND LOGISTIC REGRESSION are haviely impacted by imbalance dataset

[@mohitupadhayay1439]

Assumption 1: Linear Relationship. Assumption 2: Independence. Assumption 3: Homoscedasticity. Assumption 4: Normality.

[@tanmayvaidya8337]

Thanks Krish. This will surely help. Please upload similar interview preparation videos for other ML and DL algorithms. :)

[@tarunbilla1900]

Please make some videos on times series .

[@himanshuarora6822]

Machine Learning Algorithms such as Logistic Regression, KNN,SVM (basically which includes Gradient Descent and Euclidean Distance Computation) are affected by Imbalanced Datasets. Please correct me if I am wrong. Thanks

[@souravmzdr]

In a regression problem like linear regression we do not take into account the classes of the dateset.

[@vijendersinghshekhawat6244]

Hi Krish, The video should be reorderd at 3rd place in ML interview playlist. Thanks for wonderful videos..

[@kumareshbabu7951]

Can we use Principle Component Analysis along with SVM?

[@techspoc7442]

Algorithms which are not based on tree based algo impacted by imbalance datasets. Please correct me if I am wrong.

[@louerleseigneur4532]

Thanks Krish

[@nikhilkumarreddy5581]

where is day - 4 VIDEO??

[@akshaygudiyawar3180]

How is SVM affected by outliers? The separating planes will depend on just support-vector points right?

[@aravindnaidu1286]

try to relate outliers with cost function then you will get to know!

[@jeeveshkataria6439]

Sir please provide hugging face bert model

[@anjalisharma6543]

Naive Bayes, Linear Regression, logistic regression, and SVM, Yahh yes, DT also gets impacted by highly imbalanced datasets

[@saidurgakameshkota1246]

Logistic regression,knn,svm

[@aishwaryaandure6043]

please make a video on SVM loss function

[@navinchowdary5954]

logistic regression will be impacted by imbalanced data set

[@sahilraheja2491]

SVM is prone to over-fitting as it is s sensitive to noise and the amount of training data, please can you recheck and suggest.

[@roshankumargupta46]

Thank a lot sir. Can you make similar videos on XGBOOST, RANDOM FOREST?

[@PiyushKumar-tv6dr]

sir i think that before this in a video of yours you said tht we don't need feature scaling in svm

[@mr.kapoor3803]

Again D

[@reshusingh6005]

Where is the link?

[@anilshinde8025]

All classification related algorithms are affected by imbalanced datasets.

[@anandhiselvi3174]

Please do video on svm kernel

[@prasanna271992]

Why no assumptions for SVM? Didn't get proper answer in the internet.

[@prasanna271992]

Feedback: It would be great if you can give one liner reasoning like why, for all the advantages, disadvantages or any other point. As the points like advantages and disadvatages can be easily found in google but the reason behind it is not easy to find. Hence it can save lot time for us. Because the recruiter will go deeper with each answer we give. So we need to go down to all the whys and hows. This is just a feedback from my side.

[@barathkumarg9197]

sir please upload svm implementation video

[@nothing8919]

i think all the algorithms of classification

[@saibhaskar591]

Need a video on K means

[@ayushichoudhary6989]

plz upload kernel videos

[@pavankumar4368]

Except for the Ensemble Learning algorithms RF, XGboost, Gradient Boost, Ada Boost other algorithms may get affected by the Imbalanced dataset. why only these not effected 1. Ensemble Learning is of creating N number of Week models to make strong learners in RF. 2. XGBoost is a combination of Bagging and boosting, it takes the weighted average of the many weak models and by focusing on weak predictions and iterating through the next models it reduces the error. 3. same with gradient boost it starts with a weak prediction as average and builds decision trees by adding learning rate times the outcome of the residual tree to the initial prediction and process repeated. 4. Ada boost gives more weight to the incorrectly classified samples in creating further stumps. thank you guys hope you find the answer Correct me If I am wrong by replying back

[@shubhammarathe534]

Sir can u suggest me a good laptop under 70-80k for study purpose. Write now i am in last year of BscIT and doing research on email classification using SVM. I want a laptop which will last for 3-4 years for software development. Please suggest me

[@sofluzik]

Since your question is more about imbalanced dataset, I am assuming we can safely look at only supervised learning algorithms In this , cart, random forest , baggingz boosting type of algorithms are fairly resistant to imbalanced data. Logistic , svm , Ann,k-nn are sensitive .