Support Vector Machine Algorithm

MOHAMMAD ALI ASHRAF

DEAR SIR/MADAM, I NEED THIS CHANNEL SPONSORSHIP FROM YOU TOO IF POSSIBLE https://www.dailymotion.com/mokkamodina19  DEAR SIR/MADAM, PLEASE ACCEPT AND PERMISSION DAILY MOTION, VIMEO AND RUMBLE.COM IN BANGLADESH.  Mohammad Ali Ashraf, [6/29/2024 7:07 PM] https://ea.ebs.bankofchina.com/contactUs_en.html  ashrafm703@gmail.com IS MY EMAIL AND DEAR SIR/MADAM, I NEED SUPPORT AND LOAN FROM CHINESE PRESIDENT AND GOVT.  DEAR SIR/MADAM, I WANT LOAN FROM YOU TO LEARN AND EARN MONEY BY THE SUPER AFFILIATE MARKETING OF JOHN CRESTANI.  DEAR SIR/MADAM.I AM FROM BANGLADESH MY PAYONEER ID IS  mokkamodina19@gmail.com   , IN ANY BANK ACCOUNTS OF BANGLADESH IS NOT SAFE FOR MY MONEY YOU KNOW THAT. WILL PAYONEER BANK ACCOUNTS OF MINE WILL BE SAFE FOR MY MONEY.? ALL OF MY ASSETS OR PROPERTIES ARE SEIZED BY FAKE WAYS IN BANGLADESH. I WANT TO LOAN FROM YOU TO BUY  READYMADE HOUSES OR HOMES FOR ME. ALSO NEED LOANS FROM YOU WORK OR JOB VISA FOR USA/CANADA. ALSO NEED LOAN FROM YOU TO SETTLE AND STUDY IN THE USA/CANADA.. I WANT TO HANDOVER ALL OF MY ASSETS OR PROPERTIES WHICH ARE SEIZED BY FAKE WAYS TO THE USA/CANADA UNIVERSITIES.  https://www.aiddata.org/how-china-lends  mokkamodina19@gmail.com is my payoneer id. i want my youtube channel sponsorship from you.https://www.youtube.com/channel/UCrNbtmpYbyYMgmzD9aPtGMQ and also help me to monetize my tiktok https://www.tiktok.com/@mohammadaliashr3   AFTER MONETIZATION I WANT THIS LINK SPONSORSHIP FROM YOU TOO. https://rutube.ru/channel/37901720/ ,,,,mokkamodina19@gmail.com is my payoneer id.

Transcript

00:00Hello everyone, let us move on to the next lesson here support vector machine algorithm

00:20which comes under the classification algorithm. We have learned linear regression, then we

00:26have come on to the logistic regression. Now, in linear we are using mostly linear equation

00:32straight line concept and in the case of a slope is considered, logistic we studied the

00:37s shape curve. In the case of support vector machine algorithm, we will be considering

00:44the line as well as the boundary, best fit line and the boundary.

00:49So, let us go on with in detail. Here it comes under supervised learning algorithm and it

00:57is used for both classification and regression, but mostly it is considered for classification

01:04and the goal here of SVM is to create best fit line or best decision boundary. The best

01:13fit line for linear and decision boundary for non-linear and this is how we categorize

01:20the data from the n dimensional concept for SVM. Now, the best decision boundary is also

01:28called as hyper plane and the SVM here chooses extreme points or the vectors of before creating

01:38the hyper plane. To create a hyper plane, the algorithm here chooses the extreme points

01:47and these extreme cases are also called as support vectors and hence it is termed as

01:55support vector machine algorithm. Since, we are considering the extreme points, it is

02:01called as support vectors and hence we named it as algorithm, support vector machine algorithm.

02:08Now, this algorithm is mostly used in the case of face detection, image classification,

02:16text categorization. This is where the SVM algorithm is used. Now, this is the figure

02:23here. We have two categories, category A that is green, category B this is blue and

02:32we are considering a line and a hyper plane. We are considering two extreme points and

02:40these are the two extreme points green and the blue which is coming on the line or nearest

02:47to the line. So, hence this is called as support vectors and a boundary which is being formed

02:54as hyper plane. So, there is both positive hyper plane. So, this is the basic concept

02:59here for you to understand. Now, types here we have two types which is linear and non-linear.

03:06Linear, it is used for linearly separable data which means the data can be classified

03:14into two classes using a straight line. In the previous figure, there are two categories

03:21of data and we are separating it with a single line. That is how we are specifying for linear

03:29SVM and hence such data is termed as linearly separable data and can be called as linearly

03:39SVM or linear SVM classifier. Now, in the case of non-linear SVM, we are using the boundary

03:48that is it is used for non-linearly separable data. You cannot separate the data using just

03:53a straight line which means if the data set cannot be classified by using a straight line,

03:59it is called as non-linear data and the classifier here is termed as non-linear SVM classifier.

04:07So, two types linear, non-linear. Linear, it is separable by a single straight line

04:14and non-linear, the categories of data cannot be separated by a single straight line.

04:21Now, we are coming into the concept of hyperplane and the support vectors. There can be multiple

04:27lines or decision boundaries to segregate the classes in an n-dimensional space. If

04:32you are considering an n-dimensional space, we can have a different multiple lines or decision

04:38boundaries where we can categorize the data into any form as per our idea.

04:44But here the goal is and of course, when you are able to create lot of decision

04:50boundaries and lines, we can have different forms of equations. We can solve it in any

04:54ways in possible, but that is not how it works. We need to find the best fit line or the best

05:01decision boundary that supports our problem. So, but the goal here is to find out the best

05:09decision boundary that helps us to classify the data points and this best boundary is

05:15called as hyperplane. The best boundary that we have been able to take from this n-dimensional

05:23space is called as hyperplane of the SVM. The dimension of the hyperplane depends on

05:29the features present in the data set, which means if there are two features, then the

05:34hyperplane can be a straight line. If there are two features, two categories, it can be

05:39a straight line. If there are three features, if there are more than two features or there

05:44are three features, then the hyperplane will be a two-dimensional plane. This is how we

05:50are categorizing it. We can also create a hyperplane that has a maximum margin, which

05:56means the maximum distance between the data points. In our previous figure, this is the

06:03maximum margin. That means from the points, see this point green and blue here, from this

06:11the maximum distance that is called as the maximum margin. So, this is what is the hyperplane

06:19concept. Now, coming to the support vectors, the data points or the vectors that are closest to

06:24the hyperplane and which affects the position of the hyperplane are called as support vectors.

06:32So, this is the hyperplane here, this dotted line and these two points are called the support

06:40vectors because they are the closest to the line or the hyperline and it gives us the best possible

06:48value. So, this is how the hyperplane and support vector concept comes into the algorithm.

06:57Now, we will study how the algorithm works for linear and non-linear type. First,

07:04coming into the linear type, the working of this SVL algorithm can be understood by the

07:10below figure. Here, we have a graph, we have two categories of data, which is in green and blue.

07:17Now, here it is in 2D space and we are using a straight line and we can separate these two

07:28classes and there can be different multiple ways also. If you want,

07:31you can separate the class line in this way also. There are different ways,

07:35but as I said the best fit line is considered here. So, this is how they have classified.

07:43We have classified it into two types and one in this linear form and another one in this linear

07:48form. The SVM algorithm helps us to find the best line or the best decision boundaries and the best

07:54boundary or the region is called as hyperplane. So, the algorithm here finds the closest point,

08:03closest point of the lines from both the classes. From both the classes, we try to

08:10understand the closest point and these points are called as support vectors.

08:16The distance between the vectors and the hyperplane is called as margin. The distance

08:21between the vectors and the hyperplane is called as margin. Now, as we have said, the goal of SVM

08:28here is to maximize the margin and the hyperplane with maximum margin is called as optimal hyper

08:38plane. So, you can see both the figures here. We have two categories of data. We are dividing it

08:45with one single line and we are trying to find out the support vectors. So, now here they have

08:54categorized into three lines and we have been able to find out the support vectors and what is the

09:01optimal hyperplane and what is the maximum margin between the data points. So, this is just one of

09:09the figure here. Now, coming into the non-linear data form, the non-linear SVM here definitely we

09:16cannot just arrange the data in a linear equation form. It is not just separable by a single line

09:22that is what is non-linear SVM. So, if the data here is linearly arranged, we can separate it

09:29using a single line or a straight line, but in the case of non-linear data, we cannot use a

09:36straight line. We need to set a boundary for it. So, to separate these data points, we can add one

09:43more dimension here that is the z. We are giving a two-dimensional for linear one that is why we

09:50have used x and the y plane, but in the case of non-linear, we will give one more dimension that

09:56is the z dimension which is the third dimension here. Optimal hyperplane and what is the maximum

10:10margin between the data points. So, this is just one of the figure here. Now, coming into the

10:16non-linear data form, the non-linear SVM here definitely we cannot just arrange the data in

10:23a linear equation form. It is not just separable by a single line that is what is non-linear SVM.

10:29So, if the data here is linearly arranged, we can separate it using a single line or a straight

10:36line, but in the case of non-linear data, we cannot use a straight line. We need to

10:43set a boundary for it. So, to separate these data points, we can add one more dimension here

10:50that is the z. We are giving a two-dimensional for linear one that is why we have used x and

10:56the y plane, but in the case of non-linear, we will give one more dimension that is the

11:02z dimension which is the third dimension here. So, the equation becomes in the form z is equal

11:08to x square plus y square. Why? Because we are giving one more dimension here as you can see

11:14one more z dimension. And this is how the data sets are being classified and since it is in a

11:233D space, we are looking at a plane. In this figure, we are looking like you know looking

11:28at a plane parallel z axis which is parallel to the x axis and we can convert it into 2D

11:36space with z is equal to 1 as well and when z is equal to 1, the figure turns out in this

11:43form and we consider the best hyper plane which falls in the circumference of radius of 1.

11:51So, in that circumference of radius of 1, the points that falls under that radius is considered

11:58here as the best data point and that is the best boundary here. So, this is how the non-linear SVM

12:05works. Now, the implementation here of support vector machine is also in the form of with respect

12:11to the python algorithm. We are pre-processing the data, pre-processing step is being done.

12:16We are trying to find the best fitting SVM classifier from the training set. We are

12:22predicting the test result and for the accuracy of the test result, we are using confusion matrix.

12:28Then we are visualizing the training test and as well as we are visualizing the test result.

12:34The training result and the test result both are visualized and compared and then we get our

12:41model or the required output. This is how the implementation of the support vector machine

12:48works. Now, coming into the advantages and disadvantages. SVM works relatively well when

12:54there is a clear margin of separation between the classes. Between the classes when there is

13:00a proper separation line, it helps us to understand which is the best fit line or the

13:07best decision boundary. So, SVM is highly advantageous when you are considering a clear

13:13margin. Now, it is more effective in high dimensional spaces and it is much more effective

13:21in cases where the number of dimensions is greater than the number of samples, ok.

13:27Always say that, always see to it that the number of dimensions here is much greater

13:32considering the number of samples and relatively it is memory efficient because we are considering

13:39only the part which is required. We are neglecting the other areas. So, that is why it is much more

13:46memory efficient. Now, coming to disadvantages here. It is not suitable for large data sets

13:53and it does not perform very well when the data has more noise. When the data is noisy,

14:01it does not give us the accurate results which leads to overlapping concept, ok.

14:07The noisy data leads to overlapping concept and in cases where the number of features for each

14:13data point exceeds the number of training data samples, the SVM will not perform well. That is,

14:20it will underperform when the features of the data points exceeds the number of training data,

14:26ok. Now, the support vector classifier works by putting data points above and below the

14:33classifying hyperplane, ok. The data points are put above and below the classifying hyperplane

14:41and there is no probabilistic explanation for this type of classification. We do not know why

14:47this kind of classification is done because there is no probabilistic value that

14:51gives us the accurate decision on why we are classifying the data points in such way.

14:56So, this is one of the disadvantage because when you are getting a large number of data points,

15:01you can classify it in any form you want linear, non-linear form, but we do not know why we are

15:06giving a linear and non-linear form to it. So, this is one of the disadvantages of this SVM.

15:13So, from this we conclude this session. Thank you so much.

Category

Transcript

Comments

Recommended