DBSCAN Clustering Easily Explained with Implementation

[@nikitagupta8114]

@3:49 atleast should be >=4. Well explained. Thanks!

[@ashwanikumar4288]

Hats off to you. Very well explained. Thank you for the effort.

[@jacobmoore8734]

Really informative - hopefully this video blows up! Everybody needs explanations this intuitive :)

[@chinmaybhat9636]

Hatsoff to you @Krish Naik Sir, Very Neatly Explained..

[@yohoshivabasaraboyina8840]

when the silhouette score is near 1 the clustering algorithm works well but in this, we have a negative value it means the algorithm was not working well

[@vaibhavshah2175]

Thanks for the nice tutorial. However, I got a little confused at 10:50. As per the 'advantages' DBSCAN is great at separating clusters of high density vs clusters of low density. But the first line of the 'disadvantages' says it does not work well when dealing with clusters of varying densities. Could you please clarify on this?

[@toxicbabygirl]

Love this video so much. It helped me with my thesis! Thanks.

[@KiWiLUTSCHER]

Same here. His excitement in his voice got me Good 😂

[@SHUBHAMKUMAR-jv4kg]

Your videos are very helpful always.... keep creating... Thanks a lot for making us understand

[@tarams7775]

Very nicely explained, that too with python code was very impressive.

[@fidelca3679]

Thank you, Sir. I'll be using it for my malware analysis.

[@sarthaksinha9340]

Hey Krish can you discuss more about the silhouette score? Like how does it varies and how to determine if it is good silhouette score?

[@TheBjjninja]

The higher the score, the better the theoretical number of clusters is doing in terms of that particular algorithm. The score represents maximizing intra cluster distance and minimizing inter cluster distance. It is only a theoretical optimum and does not always use the result because it depends on the domain

[@sykumar_29]

@@TheBjjninja i guess its maximizing inter cluster distance and minimizing intra cluster distance

[@sandipansarkar9211]

Awesome explanation. Need to practice in jupyter notebook and get my hands dirty. thanks

[@amritakaul87]

How to solve the error "positional indexers are out-of-bounds" for my own data set...?

[@anuragkumar2735]

very well explained.. carry on making more videos on machine learning algorithms

[@Kmysiak1]

Great explanation but most of us have to utilize more than just two features. That's where DBSCAN will start producing 20, 30, 40..... clusters.

[@rvkrm9262]

That is 5 important points !!!

[@sijuas3863]

Simple and helpful. Thank you..

[@rezasoleimani6636]

I hoped this video included plotting different clusters.

[@JohnVandivier]

Dude this was fantastic. Well done.

[@vinitgalgali8856]

superb explanation!

[@kothapallysharathkumar9743]

how to Choose eps and minpts for DBSCAN

[@jayaprakashs4412]

hyperparameter tuning

[@abhishek-shrm]

Sir great video. But how you decide value of Epsilon and minPoints ? Is there any test like there is elbow test for finding K in Kmeans?

[@venberd]

simulated annhealing.

[@aminzaiwardak6750]

Thank you sir, you explain very good.

[@arunhbca]

Why the dataset was not scaled before calculating DBSCAN...? It's worked based upon euclidean distance right..?

[@CasuallyYoursTuhinBanerjee]

Sirji. I understood that agar ek point ka neighbour core point hai to usko border point bolenge. What if ek point ka neighbour ka neighbour core point ho..??

[@manabsaha5336]

Nicely explained.

[@jishnusen1470]

How do you visualize the clusters? What if I want to have only 4 clusters?

[@letslearnjava1753]

Hello Jishnu , if you want you can refer this video once , programming language is diff but anyway,you will be getting idea to visualise the clustering-- kzbin.info/www/bejne/f5KTkmd4Z9JsfrM Happy Learning 😊✌🏻

[@vinaylanjewar]

is it possible to have a border point in a noise point circle ?? what we can say for that point (noise) ?

[@mdashrafmoin1170]

How to do silhoutte validation in dbscan , showing error dbscan have no attribute n_clusters

[@pigno]

About DBSCAN inefficiencies for high dimension input data: how many components at most can a data point be for the results to be acceptable? 5-10? 50+?

[@sandrafield9813]

Thanks! You're good at this!!

[@AmitYadav-ig8yt]

Thank you sir. Have been waiting for this

[@joannawyrobek9260]

Did You include the center of the radius as one of these 4 points in the neighbourhood?

[@byebyedislikecount939]

yes.

[@alfredoderodt6519]

Excelent explanation! Thank you.

[@subodh.r4835]

The clustering is good when the silhouette gives a high value right? Then in this case DBSCAN has not performed well?

[@pramodyadav4422]

In the starting we have assumed value of epsilon and minimum_points. How we can find the optimal value of epsilon and minimum_points?

[@vedanti2358]

Confused about core points. COre point is that point when we have a cluster arounf it with core point being centre.But If there are no min points we cant callit as a clustenr and we cannot call the point around which the eps is used as core then how can we say while calculating border points that when atleast one core points is present Is that core point fo a different cluster present in another clustertoo? is overlapping possible?

[@rohanphuloria4111]

please explain the significance of the final score

[@avishakemaji4221]

Well explained Sir!!

[@snglvl]

Hey, nicely explained. I have a data points with 128d. I try to cluster the points with different combinations of EPS and minpts values. So far, it failed to group points reasonably. How to find the EPS and minimum points values for any situation???

[@DeepakKumar-uz4xy]

use 0.85 '

[@limavedaniazi7492]

very helpful

[@mohitkushwaha8974]

What is the unit of epsilon(radius) ??????

[@akashpoudel571]

Sir dbscan.core_sample_indices method isn't working out.....theory part was really clear...

[@neelakanthadolai5743]

You are the best

[@hasinthanawod5656]

This is GREAT!!!

[@chandinisaikumar2736]

Can you please let me know which evaluation method can be used for DBSCAN??

[@fitrianinasir4272]

i tried and practiced this tutorial but i got different number of clusters, is it possible? or I just did some mistakes?...

[@minurose3786]

Good video If possible can you make video on HDBSCAN algorithm too?

[@sofiarao7144]

Nice Video on DBSCAN. Can you pls make a video & explain Credit_Card Risk Assssment which you uploaded on github?

[@himalayasinghsheoran1255]

Good video.

[@brunosuwin328]

Sir i am studing B.E CSE i have a subject named Data warehousinh and data mining in that there is a topic named clustring,In text books in DBSCAN there is word density reachble,direct density reachable density connected what those words means please explain sir

[@Ishmaelstene]

Great video.

[@lam-thai-nguyen]

Thank you sir

[@akshatrailaddha5900]

Did anyone try to visualize the clusters?? If yes can anyone help me with code here. Thanks in advance

[@xyzrocks]

there is basic problem with your approach is you did not normalize the value and because of that too much noise and clusters were formed.your silhouette score also gave very poor result.

[@googlecolab9141]

thanks sir

[@somtirthamukhopadhyay5548]

Very sorry but can anyone make me understand about the accuracy or error or silhouette score which was done at last?

[@devanshadhikari9085]

Ur average silhouette coefficient is negative . Why so?

[@Lets_MakeItSimple]

I think this got confusing when you started talking about boundary point.

[@diosmorbodiosmorbo9547]

DBSCAN is one of the easiest cluster techniques to understand. You dont have things like euclidean or manhattan distance. Just the min_sample and the size of the ring of each point

[@YahYaAlabrash98]

greatttt!!! thanks

[@ridhimjain8170]

the explanation regarding sample_cores wasn't much clear, please make another video explaining better.

[@melihcelik9797]

This is not the implementation. Importing DBSCAN is not implementing it

[@pouryafarzi7635]

In computer science, we arent supposed to invent wheel again. there is no need to go for code from scratch.

[@melihcelik9797]

@@pouryafarzi7635 Yeah I know but I was looking for clever ways to implement it not use some libraries. If your code uses librarires just say DBSCAN code im python or something like that. That is not implementing the algorithm. And in data science you might not want to implement algorithms but I constantly try to find better and optised ways to implement algorithms. Even if they are full fledged and known algorithms. You never know when you gonna find something useful so I try it when I have the time. That was why I was looking for implementations, to have an idea about how people do it

[@arunkumarr6660]

can you pls share the ppt

[@camille_leon]

you could just use the medium article he stole the slides from. medium.com/@elutins/dbscan-what-is-it-when-to-use-it-how-to-use-it-8bd506293818

[@Orthagoni]

algaaarutum