How machine learning models take decisions?

What is happening is we have a large If I brought my targets there were four Ready to yeses way have worked 50 wasn’t on Obama’s knows my things will happen for sure That means production will be better for wife will be able to understand more on vice We’ll have a low understanding one, Yeah So what I’ve done No no If you look at the accuracy we award only 75% of backyards Even the accuracy is not that great And also if you located the ah 61 once that identify properly out off 62 and only nine zeros are identified properly out of talking So definitely zeros are known are not getting it a wide this what I have done it We have what No matter which usually uses forgetting.


When it’s called up the sample in apps sampling What I would do us uptime real That’s the smaller one You know those and I will offend again I mean random mutate some rules and I’ll try to upend it I do the process the so it’s likely for this 4 20 toe summoning and 420 or 4 22 should pickle So when I have this kind of re sure now I can say that at least there will be no by a thing happened Yes but the uh kind off see randomly choosing the rose and pushing So there are chances that similar rules might get pushed Yeah so they have been unknowing via creating a by us anyway But sometimes this matter looks very nice Sometimes it’ll give you pathetic is so be ready for this also.


In this case, other cases opposite of this mayor we will try to reduce the surrounding Lee start deleting those till it comes to the opposite again Here This I feel out of these two I feel this is kind of an okay matter but again we are losing data So losing their ties also not that gave us to these other two tickets, Yeah this type of writings may be the case to kiss me super No you like that sort of a project I think this is also has worked Liberte by us because a wrong prediction is born in done so panel calling customers as going to the greater if I do this method and it will give around prediction for me Ray Uh yes So there is a risk on bypassing happening Yes you are right.


But at least you can fight Are saying we cannot openly say that Okay We get a wrong prediction for sure But you can try this out up and down Something I said to accord with you You can write up you get a better reserve Lindus Just for her like a practical thing came doing You were out of those in customers and over seven customers 500 customers God people’s it on Um Well 50 artillery Because 50 customers around predicted that their car But But Linda it takes a no deposit in this place again Finder than 2 50 Uh by a separate that if I do this method Exactly what happens at 2 50 customers again Same thing Come to the later No Si Ok so you want to do first of all up Sam upscaling lab sampling.


You’re talking about this one or this one Yeah Example, Okay So what we will do is rather than copying the whole chunks So there are two ways Either you can copy the same chunk here which is I feel is not a good matter Logically what we can do is random will choose a rule and will upend it Here it becomes 500 So definitely there is a chance of bias I’m not disagreeing to that What At least it will be better than what was earlier The imbalance So I mean it depends on the attacker data So sometimes I will say it works very good Sometimes it will not work that way.


I’m sure about the example So in this case, really that is there with me The example is other except on bubbles I’ve chosen this data set where we are not getting a good result office Okay so you are right Definitely You are creating a by us unknowingly owns But that by us is much lesser than what Earlier we had it by us I know you wanted Yeah You know if you look at this it’s a very simple way Use resampling and in resembling They’re saying the places Yes Replace it till when will we reach this particular number and randomly choose the harmony at that time you want to place.


I would suggest you you can go ahead and your smaller number here as a random street rather than using a bigger number we use randomly chosen numbers are born Worry about his numb Yeah So when I did this and I use that triple-digit number because if I let us see if I do want if I do one by one it’s gonna take a lot of time for me to replace But I will feel this is a better recent past So let us say have just some or video like this So when I go through I have now several years isn’t a matter of knows that exactly equal how I have a bundle You Yeah So now our data set that the new does it that we have is not having a target in balance issue at least So now I can take my new did I said again through my model that is my lord is taking the regulation market and I can predict it.


I can print accuracies when I bring back You see I get inaccuracy or 68% But when you look at the original accusing that I got with by us was 75 miles So in my case up sampling did not help me much It reduced accuracy based 16 it So it reduced by around 7 to 8 Okay another cases that down simply one So where I’m saying I do not want to replace I want to delete Yeah And in the leading the little when I tell you to reach 1 92 was a number Yeah So in this case again if you observed both of them out a similar number.


There is no imbalance as an officer so in again Run the same chord I think My autistic regression I predicted I I give my unique wise and north and also give my accura, cities In that, case I’m seeing against around 60 So no matter I’m doing assembling or down Sampley for this particular leader set this data Both of the techniques are not working that great, Okay I mean there are a lot of factors on which it depends It depends on harmony number I’m taking orders Memorandums states when they deliver the court to you ways you didn’t try answering these factors it might change a little But I will not Do you not see immediately for sure Yeah not when you do this kind of stunt So basically I call this time Why Because we are playing around with our data It’s not that good.


In this case, if you observe for 22 in 1 92 is a huge difference Almost 100% difference When you do this you are supposed to use our very being a very strict algorithm or a penalizing a lot And in our case the penalizing over there Mr. Histories on us we and be Yeah So this is how you can use this Me a minute And then I use this Liam garden accuracy off 16 1964 So on an average I would say my criticism down seven peoples So when you do some are interrogation be ready with your s limos Or in this case he when you could have gone and used boosting Also nothing wrong.


What do you know about assuming that accuracy any so don’t use longer with uh something like this It’ll give you put reserves Okay You got the point I chose this date Does it purposely because of decision Have a good everybody So this is the original radio use Assault targeting violence This isn’t alternatively to avoid all the problems So now problem it This also is flying to Wrigley faster For so they were doing were saying while 83 quality for in college If I was saying that means we’re seeing the current was enough Android next version of Android and the next to next was in the van Driver will put it into one kid.

Leave a Comment