Machine Learning Algorithm- Which one to choose for your Problem?
ฝัง
- เผยแพร่เมื่อ 30 ก.ย. 2024
- Here is a video which helps you understand which machine learning algorithm you should use for your use case.
You can buy my book of finance with ML
#Whichalgorithmtochooseforyourproblem
amazon url : www.amazon.in/...
![[UNCUT] “บอสณวัฒน์” ลากไส้!! ใครหนุนหลัง "แม่ตั๊ก ป๋าเบียร์" I คนดังนั่งเคลียร์ I 30 ก.ย. 67](http://i.ytimg.com/vi/QlD7mU2KKRk/mqdefault.jpg)
This is the only channel i guess who understands what students are expecting to know other than traditional algorithm knowledge.. Thank You so much..
sorry to be offtopic but does any of you know a way to log back into an instagram account??
I was dumb lost my login password. I would love any tricks you can offer me
Which algorithm did u choose, Nandini?
@@chandlerdarius6380 this dude 💀💀💀
hi krish, it was an awesome video. just a doubt popping up in my mind, when there is overlap of datapoints, why cant we use SVM, bacause that will also take the points (of different classes) to other dimensions and divide the points using hyperplane. please correct me if my understanding is not correct. Than you so much for your awesome videos.
The thing is, you are looking at the pair-plot for only two dimensions. Although everything is overlapped when projected into pairwise space, it doesn't necessarily mean the data is not linear separatable right?
You have no idea how useful all your videos were to help me find a job. I cannot thank you enough Sir, please post more.
Hey Krish, great video - just shows how important it is to understand the geometry/maths behind ML.
In regards to using KNN here for overlapping data points, would SVM (radial/polynomial) be a good choice as well, since it will use the kernel trick to apply a non-linear classifier in a higher dimension?
And how would SVM compare to the tree methods in terms of computation complexity?
Same doubt. Krish please help us get clarity on this.
Even logistic regression may work, he is just comparing two dimensions at a time, they may be linearly separable in higher dimensions
.
@@karandua6564 no, I think that is not true. LR will not work good, if single plots are pairwise highly overlapped and/or not dividible into straight lines.
ok, so ideally, we decide on the algorithm based on the charts we have plotted?
Shouldn’t we scale the features before plotting?
Same ques....mere bhi mind m that Kon si algo...suru kru bt phir Maine ek tutorial ko follow kiya and dekho decision tree ko hi Laker 5 days ho chuke hai bt uska code m error show ho rhi hai
Please Explain me,
Scenario 1 : All Categorical, ordinal, nominal features - Categorical target
Scenario 2 : All continuous features - Categorical target
Scenario 3 : Combination of categorical and Continuous features - Categorical target
Which model to use in these scenarios (particularly SCENARIO 1) ?
Your videos is a gift to all data science aspirants like me and working professionals too. Thank you for making our learning easy and fun. Please don't stop
Sir, What about SVM U didn't mention on what kind of Data it can be used after understanding pairplot,..
Please make a video on how a beginner can participate in Kaggle competitions.
brother you need to dive in to find out
Dont worry about the result see where you lack and work from it
but remember not to go down from your baseline
Hi Krish,
I really enjoyed your videos. The specialty of your videos are they fill the gap between theory and practical.
I have watched tones of other videos, most of them are telling "how to plot the graph" but no one is explaining "how to leverage information from these plots?"
Just like this video, we need to use knn as most of the points are overlapping. I would really love to such a great video for other algorithms as well. Not only me, but I believe your subscribers would also love them.
Tons of thanks to you.
Great session,
Please make an elaborative video on this topic covering all the pros and cons of different algorithms and if possible with codes.
Thanks for this session again sir🙏👍👍
I never saw a video like this description...
Can I use all kinds of all algorithms and choose which one give more accuracy? Is there any problem in this solution.?
There is no problem, but it takes lot of time
Great explaination Krishna.. I would like to know how pairplot will show categorical variable and how we will do the feature engineering for categorical variable?
Yes, this is my doubt also. Please help!!
@@sejalchandra2114 You first have to do the label encoding to your categorical features. and then use the pairplots.
@Krish Naik 11:30 if you double pick the pairplot in jupyter notebook. It zooms in
This is your best and most important video of all. Thank you 🙏
I have eye blink eeg data set in txt formet
Can you please say how to my model to predict eye blink in code. And which is best ml algorithm? It's something related to BCI project
Can you please put some tutorial separately for Brain computer interface (BCI)
Awesome Video! Thank you for sharing this. I have one question. When there is a lot of overlap, can I use SVM to solve this? thank you
what if there are more number of categories..the graph takes time and will it be useful then???
I tried to plot the same on my datset & it shows a mix of overlapping in some features & non-overlapping in others,so based on that which model should i go for?
Any help would be appreciated
It was really good video.. Please make more videos in DataScience...
It is a good video. Also explained in detail to understand pair plots. :) Thanks..
My Question is:
When should we decide to use SVM and Naive by visualizing and analyzing the pair plots? :)
@siddhant Naive Bayes works well with higher dimensionality (such as text corpuses) while support vectors whpith its kernel is applied when the data is non linear but you want to apply a linear classifier by transformation
In this case, non linear svm is much better
Does it works for regression techniques, I'm trying but not able to generate plots as you showed.
Your book is out of stock, please provide a way to buy. Great tutorial!
Hi Krish, if we have more numbers of features say 30, than in that case we wont be able to draw pair plot. What should be our approach in that scenario? Many thanks for your wonderful videos :)
In the pairwise plot, the points are only plotted on two dimensions. However, our data have dimensions more than two. This means that we do not plot the other features. It looks that we do some projection from higher dimensional space to two dimensional space. Can we conclude if the groups overlap with each other only by looking on two dimensional plots? Simple example, if we project the earth on two dimensional space, we can see that the norther and southern hemisphere will overlap on each other.
I have been looking for comments like this... I agree with you. I don't think one can conclude whether the data is linear separable, from just looking at pairwise plots.
'KNN is good for overlapping data points', can you please share your source or cite a paper for this info?
Completely incomprehensible, and I'm a PhD engineer, aerospace, not 'data science', .... first the accent is off the chart. Speak understandable English, or make the vid in your natvie language for your native country.
You explained the wrong concept of Logistic regression model
. You mainly explained the linear regression model
Hi bro, h r u, your videos are really very informative. I am from medical transcription field, but since 3 months learning data science, but need proper regarding good job with good salary package, it's very important bro, can u help?
Thank you for your contribution. Please note also that classification learner app in Matlab provides you such pair plots
This approach is not 100% correct. It shows which attributes are good. Those which can clearly separate the target into two classes are good variables. Thus this is at best a feature selection method. But often multiple variables together can also create a clear separation. So by doing pairplots, it may lead to high false negatives against the suitability of logistic regression. Better way is to take 10-20 smaller chunks from different locations and apply all algorithms. The ones which show consistently better performance (with least variaton)is a good algorithm for the given data.
Liked this video, but in case of large test set KNN might prove you costly at the time of prediction
how ? can you explain further Please
Thanks a lot for such a great explanation 😊
The silent badge suprisingly belong because stock phytochemically describe an a wry respect. defiant, flawless belgian
Selecting the appropriate machine learning algorithm is determined by aspects such as problem type, data qualities, and desired results.
why my python script giving erroe as key TARGET CLASS ? please help
Great video, friend!! Keep up the good work. Need to learn a lot through you.
Good Job! Clearly explained. Thank you so much for this video.
How to do model selection of Regression models ?? I m getting the cross validation scor of -2.8456
This is what is required .. thank you so much for sharing this 👍👍 great work
Hello Krish. I would like to order your book, but it seems to be unvailable on amazon for European zone. could you explain me how to order ? Thx
thanks a lot for this wonderful explanation, Krishna. You are my new hero now :)
At 13:30, you are making an assumption the data is not linearly separated based on just two dimensions. That's not correct, since the data can be linearly separable in multiple dimensions.
Great, this is the video i was looking for to explain the difference from a basic mathematical perspective.
What if few of overlaps and few are having classification, what algorithm works ?
hello Sir, can you make a project on detection of phishing website using machine learning
Nice explanation , you explained us like teaching a parrot love you from india .
Sir what about Naive Bayes classifier?? At what condition we should use Naive Bayes classifier?
One of best video's I have seen , won't forget to return u the favour....
great sir ...
can u plz make a video on the comparison of every ML algorithm
just awesome 💚💚💚i am just recharged after completing this video..thanks a lot.
How can you say that logistic only forms straight line division. I am confused.
Hello Krish Sir. Does df.corr() give same result as pairplot?
I used to be a fan...
.
.
.
But now.....
.
.
.
.
I'M AN AIR CONDITIONER
@krish.Naik Can you explain Techniques for regression too.
I'm tried for regression but I'm not able to generate plots can you help me if you tried?
What is the best technique to predict the QOS of services in microservices architecture please
can you provide link for code?.You have solved my query.... thanks again
Nice ... Please make more videos for data science, beacuse it better to understand into that way...... Thank you
Also make video for data cleaning, Solve any business problem
Hi Krish,
I've been following your channel for most of my learning.
The way you explain the things gives a very familiar approach and giving out what is needed in the way everyone can understand.
This let me watch any video of yours till the end and learn completely.
Happy to learn from you. Thanks
Hi sir , can u give any idea problem statement for loan approval prediction
Excellent video for an initial understanding! Thanks!
Thank you Krish! All doubts cleared now.
You are doing amazing job Sir,
I am too late to land on your channel
Keep it up
You are providing knowledge free of cost where people taking money for this in thousands
No need to join any private classes with huge pay if you will provide this material in sequential order
Great job all the best👍💯
Have you found an answer to your question?
Thank you Krish for the amount of effort and insights you put into your videos. Really helps a lot🙏❤️👍 May God bless you and keep you well🙏
Logistic regression is not what you are explaining here. Linear regression is where straight line comes
For unsupervised learning which algorithm i have to select
Anyone would like to join with me to do a new ml project? I will explain the task
can u suggest any book which explains this method?
Informative. Thank you so much
How do we decide which algorithm to use in case of Regression task?
why you didn't include Support Vector Machine
Hello, thank you for your contents. Your explanation really helps. I have a question, in the beginning you are mentioning Logistic Regression drawn as a straight line, however, is not Logistic Regression a curve line? While the linear regression is a straight line?
Thank you.
Yeah same doubt. Sigmoid func
very well explained thankyou so much sir
Great Vid. You made things clear. Thanks
Most important video. Explanation is just amazing.
Bro, y is there no subtitles, I have a problem in hearing😞
In my opinion there is no need for subtitles , I could hear him perfectly fine.
@@randomdude79404 hey bro, he meant deaf 🧏♂️ or any disability of hearing..
@@masthanjinostra2981 I humbly apologise sorry about that.
Can we get simular video with regression analysis thanks
Sir if possible please upload this python script on github
Best video till now. Thank you Krish.
This was very much helpful Thank you Krish
Hlo sir I need a problem statement for loan approval prediction
I have one python code. Will you help me make is error free?
What about naive bayes. Please answer fast if you can.
Thanks for this amazing video Krishna. However, some questions are striking on my mind. For binary class classification it is understood that if we have linearly separable data then we can use SVM or logistic regression or any other classifier which can make decision boundary. Suppose I have a dataset that can be separate by a decision boundary then I can make user of SVM, logistics regression or perceptron algorithm. How to choose best classifier out of these 3 classifier?
Check with Cross validation technique to understand a better classifier
Thanks a lot Krish......it's really very helpful ......
Which one is suitable for integrity of Iot device
sir, can you please provide the github link for this video? its not present in the descprition.
Which book explains this visualization of data?
Superb Explanation can you please do on regression
Thanks for the useful videos you post.
Nice video. But I like the way you are presenting at whiteboard. It feels like we are in actual class.
Woow, Thank you. its helpful
Is overlapping denotes Overfitting?
How to select the best model for a regression problem?
Sir thank you so much you solve my problem
All of your videos have a lot of useful information. Thank you...