Decision Tree Regression Clearly Explained!

Here, I've explained how to solve a regression problem using Decision Trees in great detail. You'll also learn the math behind splitting the nodes. The next video will show you how to code a decision tree regressor from scratch.
#machinelearning #datascience
For more videos please subscribe -
bit.ly/normaliz...
Decision Tree previous video -
• Decision Tree Classifi...
Love my work? Support me -
www.buymeacoff...
Join our discord -
/ discord
Facebook -
/ nerdywits
Instagram -
/ normalizednerd
Twitter -
/ normalized_nerd

[@hannav7125]

Shout out to this dude for the awesome visualization and clearly explanation.

[@drewplayz3765]

u kinda bad tho

[@jayo3074]

It looks hard at first but with a good teacher explaining it really is so simple

[@sai_sh]

Hi at 4:40 why did it go towards left node rather than the right coz x=16 and x

[@yashsaxena7754]

Great explanation! One question though is if the prediction is based on the average value of the target variable in the leaf node, it would mean that all the observations terminating at a node will have the same prediction. Is that right? For e.g., if 10 observations are terminating at a leaf node all will have the same predictions.

[@pratyushrout7904]

The nicest explanation video for DT on KZbin...

[@gustavoalcarde]

Thank you so much! Very simple and visual, that's all I needed!

[@jullienbeaufondcamacho2055]

Excellent visualization, kudos!

[@XuanTran-ri1hn]

Hi, may I ask 1 thing in minute 4:41, because x=16, so I think that condition is not true for x1

[@explovictinischool2234]

This condition is, indeed, true. You may have confused x0 and x1: x0 = 16 but x1 = -2. Here, we are talking about x1

[@nickgeo8250]

Really good explanation well done! Only one question, how do you calculate the wi weights?

[@AidarMHTV]

The weights, I think, it's just a fraction of items for each side. For example, 2 items on left and 6 items on right gives .25 and 0.75 weights respectively.

[@dgaphysics4026]

Great explanation!

[@shanks9758]

good job , thank you

[@jamalnuman]

great presentation

[@ahmedelsabagh6990]

Very good visualizations

[@polishettysairam6466]

Its good, can you provide the dataset used here

[@mihirsheth9918]

great explaination.thanks

[@kalpanakadirvel9220]

Excellent video.. Thank you

[@RohitAlexKoshy]

Great video. Keep up the good work!

[@NormalizedNerd]

Thanks!

[@kellymarchisio377]

Excellent!

[@tinyeinmoe5147]

you're the best, thank you Soooo much, india is the best

[@Mudiaga11]

this is excellent

[@samarthtandale9121]

Incredible 🔥

[@Rehul-gw3yj]

I actually came here to understand how we get the route note. anyone >?

[@tomgt428]

cool

[@dnyaneshjalamkar4257]

Your so called autopilot ruined the video!

[@SoubhikBhattacharya]

Can you explain why X0

[@onetapmanbbr]

8:24 He explains that the algorithm compares every possible split and finds the one with the best variance reduction. So the y in X0

[@soumyarm3851]

​@@onetapmanbbrv

[@sillem4337]

This video is next level teachning. Concept presented so clearly and so well. Thank you!

[@yuyang5575]

This is awesome! Clearly we use a binary tree to do the classification first (or build a decision tree first), and then we follow the tree to reach the target leaf node. Btw, which software do you use to make the animation? very impressive

[@NormalizedNerd]

Thanks! Well, I use Manim (a python library)

[@bhavinmoriya9216]

Thanks for the video! How do you decide x0

[@NormalizedNerd]

Actually, we try every possible value of the threshold and find which one produces the best split. If can go through the code for a better understanding. github.com/Suji04/ML_from_Scratch/blob/master/decision%20tree%20regression.ipynb

[@bhavinmoriya9216]

@@NormalizedNerd Thanks buddy :)

[@kalam_indian]

@@NormalizedNerd you have done well but plz first explain taking an example by showing every steps from first to the last with every maths used and the computation on how we get the results and why this value and not the other etc. plz teach like this so that every learner even the one who don't have basic can understand it is a request

[@saigopal5086]

brooooooo this is brilliant, I can't resist myself from pressing the like button, it's such a blessing to have people like you

[@jamiyana4969]

Honestly this is the most high end professional video that's so simply explained! Amazing job!

[@Ivan-cp2hn]

your explanation is so far the best in youtube up till now. Dont know why the view number and likes counts not that high. But keep doing the great work !

[@persevere1052]

Extremely helpful and easy to understand!

[@diegobarrientos6271]

Thanks for the explanation!, I have a question... I watched in some videos that use MSE instead of variance, so Should I use the sum of squared error or variance? It'd be great of someone could clarify this please

[@DaaniaKhalith]

MSE is used if both input and output is continuous ,variance is for discrete input n continuous output

[@rishabhgarodia410]

Thanks for these visualizations! Helps a lot

[@paultvshow]

At 1.28, what you mentioned was misleading and could be misinterpreted. A line is still a line in 2D, a line will never be a plane in 2D. You should have said “or a plane in 3D”, or simply call it a hyperplane instead of a line or plane.

[@siddhantpathak6289]

Awesome visualization and explanation, I went through the Github implementation and it seems you are using unique feature values as possible thresholds. How this approach would work for a continuous feature with millions of records, as there will be many unique values to test. Possible thresholds in the video were 1 and 2, right? Just checking my understanding.

[@DenisPolato-i3h]

amazing explanation!! thank you soo much!!

[@Gulshankumar-fg9ls]

Bro… I would suggest you to get the proper knowledge when you start teaching any topic in machin learning, sometimes your statement is vague

[@ckeong9012]

Awesome. May i know what kind of software you use for the visualization?

[@aminbakhtiari3026]

that was quite understandable ! thanks for the good explanation and visualization!

[@suryahr307]

does it take care of outlier data points?

[@himtyagi9740]

What a beautiful lecture..kudos to your efforts

[@inderkaur3012]

Hi, Though Var R1>Var R2, how do we conclude that R2 is best suited for split? Graphically I understand your logic as the colors are best segregated due to R1, we must choose R1, but I was unable to conclude the same from variance perspective. Could you please explain the same?

[@konstantinosmaravegias4198]

Do not confuse the variance reduction with the variance of each split. VarR1 < VarR2, hence the VarR1 split reduces more variance from the parent node (1 - VarR1 > 1 - Var2)

[@stuffy1021]

I can't understand

[@gatecseaspirant-dk9ze]

hello people from the future! you nailed it here

[@candicerusser9095]

Hi, great video! A small doubt, what does "desired depth" mean in decision tree regressor, does it mean that we reach a point where we can't split anymore, like variance becomes 0?

[@jakeezetci]

i think that's the depth of the tree you want, that you need to find by trying out yourself. you want to stop before variance becomes 0, as then the prediction really goes wild

[@JPDEV092]

how to do videos like this ?

[@osama11osama]

I like it really, thanks

[@takeshi7441]

Thank you sir

[@ckeong9012]

Awesome. Great video. Much appreciated if you could put the values or labels on the cartesian. TQ~

[@NormalizedNerd]

Suggestion noted...

[@subarnasubedi7938]

I am confused why you took x2

[@anuragshrivastava7855]

how did we get average

[@alexandrfedorov7297]

You are awesome

[@Nirjhar85]

Awesome!

[@sebastiaanvanhassel4664]

One extra like for the classical music!!! 👏😀

[@NormalizedNerd]

😁❤

[@Randomstiontastic]

How do you find the value of the inequalities for the filters?

[@NormalizedNerd]

By checking every possible value of a feature as the threshold and splitting the dataset based on that. Then taking that particular feature and the corresponding threshold that gives the maximum information gain. Please see the code provided in the next video for more clarity.

[@Rahul.alpha.Mishra]

Thanks a lot bro. And your viz helped me explain my Model in the presentation. Carry on foreward

[@yuvikayadav8203]

Great explanation! Just one doubt that in decision tree classifier we split the nodes until we get pure leaf nodes if hyperparameters are not clearly stated but in the case of regression problems how do it decide when to stop generating the tree if no hyperparameter is defined?

[@kalam_indian]

you have done well but plz first explain taking an example by showing every steps from first to the last with every maths used and the computation on how we get the results and why this value and not the other etc. plz teach like this so that every learner even the one who don't have basic can understand it is a request

[@illan731]

Great explanation. Feels too simple to have looked up on video, which means it is explained very well.

[@smartshoppingapp3585]

How do you find the best root node. ? Coz in video it's about finding the best split which really helped. But how to find the best root node???

[@razieqilham8327]

So in love with your explanation sir, but im confused with the dataset, could u build a dataset in table, not graph?

[@HuyLe-nn5ft]

You already had 1 more subsciption, Superb explanation and visualization!

[@ismailwangde580]

Bro, you are better than krish naik lol. Thank you for the efforts. really appreciate it

[@akashkundu4520]

Can you add post prunning of the tree and visual representation of the tree please.I have an assignment 😭

[@madhumatinarule4489]

Kindly make full course on fundamentals of machine learning as we are from not from computer science

[@akshaypr9164]

amazing man!! love your explanation and style

[@robm838]

Which values are -7 and -12 cannot be found on the grid.

[@rishiksarkar62]

Fabulous explanation sir! Thank you very much!!

[@prashantmandare2875]

Best explanation of decision tree for regression that I have come across

[@mindlessambient1791]

In the ending example, the weighted average of variance used weights with denominators of 20 (i.e. 11/20, 9/20, etc.) . Has anyone ever thought to adjust these weights using Bessel's correction? Not sure how much of a difference that would make but just curious. I am guessing that the weights would be something like 10/19 and 8/19 with this adjustment.

[@fabianaltendorfer11]

one hell of a explanation video. great!

[@diabl2master]

What a brilliant video!!

[@itsme1674]

Wonderful explanation

[@the_senthil]

Quality content... I never seen before 💯

[@leolei9352]

Really good teaching.

[@soniasnia]

ur video is just awesome!

[@sadjiajfiarei3498]

bro you made it so easy

[@kajalmishra6895]

I love all your videos.

[@SachinModi9]

Man... Loved it..

[@ronitganguly3318]

Dada tumi bangali?😁

[@NormalizedNerd]

Ha bhai :))

[@ronitganguly3318]

@@NormalizedNerd bengoli accent ftw!

[@MrHardgabi]

incredible

[@estefvasqu]

Thank you for sharing your knowledge. We appreciate it Greetings from Argentina

[@NormalizedNerd]

You're most welcome...Keep supporting!

[@Asmallpanda1]

Very Nice ty

[@ccuuttww]

U spend a lot of time to make an animation

[@ccuuttww]

I am not sure if u should use MSE for every split

[@saqlainshaikh5483]

Great Expectations ✌️

[@NormalizedNerd]

Thanks!! More to come :)

[@61_shivangbhardwaj46]

Thnx sir😊

[@NormalizedNerd]

Most welcome

[@asrjy]

Sick vid! Did you use manim to make this video?

[@NormalizedNerd]

Yeah!

[@asrjy]

@@NormalizedNerd nice

[@xINeXuSlx]

Great video! Helping me a lot in preparing for my Data Science exam soon. One thing I did not quite understand yet is when I should use Decision Tree Classification or Regression? I understand that one uses Information Gain and the other Variance Reduction, but how do I know in the first place what to apply?

[@NormalizedNerd]

It depends on the problem you are trying to solve. As a thumb rule: if the target variable takes continuous values then go for regression and if it takes discrete (and few) values then go for classification.

[@xINeXuSlx]

@@NormalizedNerd I understand, thanks a lot :)

[@piero8284]

Nice explanation, i was struggling a little bit to find some detailed material about this topic. As I thought, decision Trees in general always check for the best split looking for every possible feature, that means if there are k features and n samples, at each split the tree will perform O(m*k) variance computations, right?

[@p337maB]

It's not O(m*k) but exactly m*k computations at every split.

[@piero8284]

@@p337maB sure

[@DustinGunnells]

How are you determining the filter splits further down from the root of the tree? I don't see the reasoning that you're using to make this useful. I see the filtering, I see data points, but what is determining the other filters from the initial filter? Why is the partitioning valuable? How would the partitioning be applied? Why would you have two of the same filter between the two x variables, x sub 0 and x sub 1? Why is x sub zero represented in the root but not x sub 1? What is the relationship/difference between the two x variables? This looks initially useful, then it looks like a bunch of snow on a cathode.

[@NormalizedNerd]

"but what is determining the other filters from the initial filter?" The initial filter (at root) divides the data into two sets. The left one is then again divided so is the right one. We do this process recursively. While splitting a set we choose the condition that maximizes variance reduction. Please see the implementation to get more clarity: kzbin.info/www/bejne/hmO9c2uZaq2UZ7M

[@DustinGunnells]

@@NormalizedNerd So are you saying that x sub 0 and x sub 1 are two sets of decision sets? Rather, two collections of boundaries? Something still looks off. If it's an array of decision boundaries, how do you jump from 1 (of x sub 0) to -7 and -12 (of x sub 1)? I've even tried to figure out the symmetry in the tree to find logic. 4 elements of x sub 1 3 elements of x sub 0. 20 partitions in the grid for 20 elements in the set. I've watched several of your videos trying to understand your message. Explain this one where it makes sense and I'll definitely continue to watch your other content. I try to give everybody that says they're providing "knowledge" a chance. This is outstandingly bonkers to me. I'm also a programmer and MBA