WEBVTT 00:00.990 --> 00:03.270 Now let's discuss about the aggression. 00:04.470 --> 00:11.760 So let us assume that this is the of which we have in time and we have only one input variable, which 00:11.760 --> 00:13.500 is the education. 00:15.810 --> 00:20.580 And based on this education, we want to predict the income of the. 00:22.770 --> 00:31.380 Now, we can predict these by creating a linear regression and what linear regression suggests is that 00:31.620 --> 00:40.770 we will be drawing a particular line such that the line will be able to map all the values if possible. 00:42.730 --> 00:46.710 So just as you what lions can we create here? 00:48.120 --> 00:51.030 There could be a line created like this. 00:52.860 --> 00:54.420 There could be another nine. 00:55.400 --> 00:57.470 Created like this. 00:58.970 --> 01:01.680 And there could be another line creator. 01:03.410 --> 01:04.250 Like this. 01:08.340 --> 01:15.180 Now, each of these lines will be able to capture some of the data points, but none of the line is 01:15.180 --> 01:17.970 actually happening, capturing most of the data. 01:18.930 --> 01:25.430 So we have to find out which line will be able to capture maximum data points. 01:26.620 --> 01:30.560 So we will be finding the error which is present to you. 01:31.330 --> 01:37.150 Now let us see how will we find out if the values I correctly predicted or not? 01:37.900 --> 01:44.770 So for finding out if the values are predicted correctly or not, we will be dropping a perpendicular 01:44.920 --> 01:49.380 to each data point to find out the distance or the error value. 01:50.930 --> 01:54.350 So let us consider that this is the line. 01:55.980 --> 02:00.060 Which we have actually created for the Swedish. 02:03.080 --> 02:05.360 So when we are dropping the ball, when Nikolas. 02:11.100 --> 02:13.080 The distances will be this. 02:15.880 --> 02:16.510 This. 02:18.100 --> 02:18.730 This. 02:20.500 --> 02:21.160 This. 02:52.020 --> 02:52.650 And this. 02:54.830 --> 03:03.950 Now, let us compare the distances now, these instances when compared to distances from this particular 03:03.950 --> 03:04.300 line. 03:07.120 --> 03:14.570 Will be a lot more so the distances for this particular line will be larger than the earlier one. 03:14.590 --> 03:17.600 So let me drop distances to this particular line. 03:17.620 --> 03:21.070 So if we drop the distances here, the distance would be less. 03:22.010 --> 03:26.630 Hear the distance would be less here, the distance would be less. 03:29.280 --> 03:30.180 This is same. 03:32.860 --> 03:38.680 Now, after this point, all the distances will be more than the previous distance. 03:42.120 --> 03:51.630 So you can see that in comparison with these two lines, this first line actually provides a better 03:52.050 --> 03:52.550 result. 03:54.830 --> 04:01.850 So we will be creating the front lines and we will be comparing the distances between these lines. 04:02.450 --> 04:08.840 So how will it be creating these lines and how we will be reducing the distance is the next thing in 04:08.840 --> 04:09.080 my. 04:10.100 --> 04:16.580 So let us try to solve this and let us try to see that, how will we actually create this equation of 04:16.580 --> 04:20.090 this line and how will we create this model? 04:25.100 --> 04:30.950 So what are the different type of equations which we have, there are limited equations, quadratic 04:30.950 --> 04:34.400 equations, Kubic equations, so a linear equation. 04:35.510 --> 04:37.250 Will look something like this. 04:40.090 --> 04:49.270 And this equation will be something like, why is it going to be one x one plus two plus B three x three 04:49.270 --> 04:51.250 plus B for export and so on? 04:52.230 --> 04:55.580 And in the end, one constant V will be added to it. 04:56.800 --> 05:04.570 Because the original line of formula is Y is equal to a mixed blessing, which is the original equation 05:04.570 --> 05:07.510 for line where C is the. 05:08.460 --> 05:11.690 Intercept and M is the slope. 05:13.340 --> 05:19.580 Similarly, we will be creating this particular equation because here we have multiple X values, let 05:19.580 --> 05:21.230 us get back to the data which we have. 05:22.510 --> 05:29.860 So this is the data which we have, so the age amount, salary dependent sex and children will be different 05:29.860 --> 05:33.310 X values, that is X one, two, three, four, five. 05:33.550 --> 05:37.390 And the interest rate will be the value which we are predicting. 05:37.870 --> 05:43.750 So the equation will be some confusion multiplied with E under the age and a definition multiplied with 05:43.750 --> 05:50.200 amount on a double multiplied with salary under the compulsion to multiply the dependent on the condition 05:50.200 --> 05:54.970 multiplied by six, another coefficient multiplied by children and some of all of these. 05:56.580 --> 06:03.420 Should be equivalent to the value of Interest-free now out of this, the only thing which we need to 06:03.420 --> 06:10.410 find out is the value of the efficient because value. 06:11.780 --> 06:17.990 And the one, two, three values is something which we already have, so the only thing which we have 06:17.990 --> 06:23.050 to formulate is the value of BE1 be to be three before and so on. 06:24.790 --> 06:31.060 Now, while we are creating this equation, we have to make sure that the prediction, which we are 06:31.060 --> 06:33.470 making good predictions. 06:33.850 --> 06:36.830 So what are different features of a good prediction? 06:37.270 --> 06:41.040 So a good prediction model will give accurate results. 06:41.650 --> 06:45.050 So a good model will give accurate results. 06:45.220 --> 06:51.250 So a model which will build this particular equation will be a better model in comparison to the model 06:51.250 --> 06:53.110 which will give this particular equation. 06:55.460 --> 06:59.940 Next is the accuracy should not be limited to just one data point. 07:00.440 --> 07:07.700 So let us say we are looking at a particular city for which we are trying to find out how much water 07:07.700 --> 07:10.160 or how much food is needed in that particular. 07:11.090 --> 07:15.040 And similarly, we are looking for multiple cities in the country. 07:16.000 --> 07:26.470 So now our task is to find out the amount of food required as accurately as possible so that no specific 07:26.470 --> 07:33.010 city doesn't have extra food or no specific city has lesser amount of. 07:34.550 --> 07:41.870 So we're going to have to make sure that all the cities have sufficient amount of food available so 07:41.870 --> 07:43.780 that no one dies of hunger. 07:46.400 --> 07:54.740 We cannot do something like we predict one requirement for one particular city correctly, and the predictions 07:54.740 --> 08:02.240 are very wrong or very inaccurate for the other cities, because in that case, other cities, the people 08:02.240 --> 08:04.530 who are living in other cities will die of hunger. 08:05.210 --> 08:05.550 Right. 08:05.750 --> 08:15.110 So the predictions which we are making have to be in sync and accurate or good enough for all the data 08:15.110 --> 08:18.890 points or at least for the maximum number of data points. 08:20.830 --> 08:27.910 So now the model which we will be creating, the model which we will be creating or the predictions 08:27.910 --> 08:35.350 which we will be making, should be performing equally well on the testing data as on the training data. 08:35.780 --> 08:36.280 Now. 08:37.650 --> 08:44.280 As of now, we don't know what testing device and what printing device, so let us try to understand 08:44.280 --> 08:44.940 that a little. 08:46.830 --> 08:53.310 So let's take the example of preparing for examination again. 08:54.930 --> 09:02.610 Now for an examination we are preparing from a particular information or from a particular book. 09:03.730 --> 09:12.010 Now, this particular book is the invasion force material, which we have now, what happens is the 09:12.010 --> 09:16.200 book will have certain questions at the end of each and every chapter. 09:17.310 --> 09:24.540 So when we are learning from this particular book, we will be practicing on those practice questions, 09:24.540 --> 09:28.200 which we will be having at the end of each and every chapter. 09:29.560 --> 09:37.930 So when we will be doing self evaluation, while we will be training ourselves to give the exam, we 09:37.930 --> 09:44.290 will be self evaluating ourselves on the basis of these practice questions which are present at the 09:44.290 --> 09:45.190 back of the book. 09:46.420 --> 09:51.470 So we will be able to perform really well if we are really studying hard. 09:51.490 --> 09:56.070 We will be able to perform really well on the training data. 09:56.080 --> 10:01.720 That is, the evaluation questions or the practice questions is in the back of the book. 10:02.940 --> 10:11.700 So now the testing data is basically the exact exam, the exam, which we will be giving Igby after 10:11.700 --> 10:20.250 tomorrow, then we will have 14 questions, which will be not present in the book, but we will be based 10:20.250 --> 10:22.780 on the concept which are present in the book. 10:23.430 --> 10:29.070 So when we are learning from the data and we are learning from the information which is provided to 10:29.070 --> 10:35.630 us, we want to perform well on the training data or on the practice questions also. 10:35.790 --> 10:43.400 And we are performing good on practice questions, but we also want to perform well on the final examination 10:43.420 --> 10:43.650 on. 10:45.110 --> 10:51.620 So this is what it means, that it should perform equally well on the training data, that is a practice 10:51.620 --> 10:58.430 question on the testing data, which is the final evaluation, so that if this. 10:59.510 --> 11:00.310 A woodmore. 11:02.450 --> 11:10.340 Now, next thing is that a simpler model is a better model, complex models than do or would Woodfork. 11:11.090 --> 11:14.720 Now let's ignore what Overfitting means. 11:15.700 --> 11:19.310 OK, let us ignore what Overfitting feels for now. 11:19.500 --> 11:19.790 Five. 11:19.820 --> 11:26.130 Just try to understand that we want to make our model as simple as possible. 11:27.380 --> 11:33.710 And what complex model does we will look at in something so far? 11:33.740 --> 11:37.670 Now the model should give accurate results. 11:38.000 --> 11:41.680 The accuracy should not be limited to just one data point. 11:41.960 --> 11:47.410 It should perform equally well on the testing data as on the training data. 11:47.630 --> 11:52.640 And a simpler model is better model than the complex model. 12:05.780 --> 12:13.880 Now, let us see what this regression so did not what we have discussed is that regression is when we 12:13.880 --> 12:16.820 are trying to predict the continuous value. 12:17.990 --> 12:25.190 So regression takes a group of random variables to be predicting why? 12:26.610 --> 12:30.570 I'm trying to find a mathematical relationship between them. 12:32.260 --> 12:39.610 So this relationship is typically in the form of a straight line that best approximates all the individual 12:39.610 --> 12:40.330 data points. 12:43.050 --> 12:49.650 So what this means is that I have certain data, I have certain data, the gardening, education and 12:49.650 --> 12:57.420 income, so I want to set up a relationship between the education and the income of the person so that 12:57.600 --> 13:05.130 for when the next time someone tells me that this is their education, then I can easily guess what 13:05.130 --> 13:06.110 their income would be. 13:09.680 --> 13:17.560 So what is needed aggression, linear regression is used for continued static good values like Ege, 13:17.810 --> 13:21.290 sales interest, industry or house prices. 13:22.490 --> 13:29.930 So we can let us see if we can have data related to the number of bedrooms, the number of bathrooms, 13:30.170 --> 13:37.400 the the floor of the building and the details, the location of the building, and based on that, we 13:37.400 --> 13:40.580 will try to find out what is what should be the price of the house. 13:41.150 --> 13:49.280 Similarly, we can get details about a person's height and weight and try to predict that age. 13:50.240 --> 13:53.180 So these are different things which we can predict. 13:53.390 --> 13:59.450 So when we are trying to predict a continuous value, then it is called the linear regression from. 14:01.270 --> 14:08.650 Now, what is the expectation, the expectation is too long, the relationship between the independent 14:08.650 --> 14:19.210 input data and the dependent on good values, the independent input data is the legacy for the for the 14:19.540 --> 14:20.610 loan problem. 14:20.890 --> 14:29.290 The independent data could be the age of the person, the gender of the person, the amount which the 14:29.350 --> 14:33.410 user is requesting the loan for the number of dependents. 14:33.550 --> 14:35.020 So all of these will be the. 14:36.230 --> 14:41.930 Independent input data and independent target value will be the interest rate, which we are trying 14:41.930 --> 14:42.470 to predict. 14:44.240 --> 14:52.860 Now for finding out this will calculate the slope and the positions to create a linear equation which 14:52.860 --> 14:55.150 could align do given data property. 14:55.850 --> 14:59.230 So we want to create a linear equation. 14:59.420 --> 15:03.860 So the equation consists of the slope and the. 15:07.510 --> 15:09.670 Coefficients of the values. 15:11.920 --> 15:14.230 So that is what we are trying to predict here. 15:15.210 --> 15:18.390 And in this entire thing, we want to minimize the cost. 15:19.150 --> 15:26.730 Now, let us try to ask you, what exactly is a predictive word like we have discussed about what the 15:26.730 --> 15:27.420 question is? 15:27.630 --> 15:32.680 But our main concern here is to know what caused this. 15:32.880 --> 15:35.040 We want to understand what the cost is. 15:35.400 --> 15:38.230 So how will we understand what is cost? 15:38.820 --> 15:44.220 So to understand the cost, we have to understand what our predictive models. 15:45.630 --> 15:55.890 So for us, a predictive model is just like the application or just like a box in which we provide our 15:55.970 --> 15:56.820 training data. 15:58.090 --> 16:05.050 So we will provide the training data, which is the independent input data and the dependent that it 16:05.080 --> 16:13.720 values, and we will provide it to this box, which we have, and then this box will try to decrypt 16:13.930 --> 16:18.610 the relationship between these independent, independent variables. 16:20.270 --> 16:26.830 So it will try to establish a relationship between this and a group that I do for you to function internally 16:26.960 --> 16:34.130 so that it will apply that function or it will apply that equation on top of these X values. 16:34.400 --> 16:38.570 And using this equation, it will be able to find out the Vivan. 16:39.920 --> 16:48.380 So when we are planning the model or when we are letting the machine learn from the data, we are basically 16:48.380 --> 16:55.760 providing the input data and the output data, the input X values and the output values to the machine. 16:55.970 --> 17:03.160 And the machine will try to find old patterns in the relationship from the state of digitize and create 17:03.180 --> 17:04.220 the equation in. 17:06.490 --> 17:15.250 And then once this equation has been created, then what happens is whenever I have any of the problem 17:15.250 --> 17:22.300 in hand, whenever I have any other input data, again, I can just provide this input data to the machine 17:22.450 --> 17:30.010 and it will apply the formula on this data and automatically provide me the output value as a prediction. 17:34.450 --> 17:42.580 Now, when it is providing me this output data is a prediction, this output data, which is very hard 17:42.580 --> 17:48.070 to value, will not be exactly the same as the value which is expected. 17:48.980 --> 17:54.500 OK, the value which we have, the value which we are predicting will not be. 17:55.480 --> 17:59.590 Exactly the same as the value which is in the real life scenario. 18:01.150 --> 18:09.730 So what do we do now, this gap, this difference between the original value and what we have predicted? 18:10.060 --> 18:14.180 So, for example, we are trying to predict the height of a person. 18:15.200 --> 18:21.650 And so when we are creating a model, we will have given some detail about the person and the height 18:21.650 --> 18:23.330 of the person to the model. 18:23.990 --> 18:31.250 So now the model will try to analyze the characteristics of the person and it will find try to find 18:31.250 --> 18:36.340 out the relationship between the characteristics and the height of the person. 18:37.490 --> 18:42.850 So it will try to do that by creating information internally. 18:43.700 --> 18:51.140 Now, while it is creating the equation internally, it means that I do think the equation to the data, 18:51.140 --> 18:52.430 which it already has. 18:53.120 --> 18:56.210 So it will try to check for the data again. 18:56.420 --> 18:59.800 So it will at one moment it will learn from the data. 19:00.620 --> 19:07.850 At one moment it will learn from this data on this value of the interest rate, and later it will try 19:07.850 --> 19:10.660 to predict this increased interest rate. 19:11.090 --> 19:13.210 So it will try to practice again. 19:13.850 --> 19:20.300 You remember the example for the book, so it will try to evaluate itself based on the questions on 19:20.300 --> 19:22.040 the back of the chapter. 19:23.140 --> 19:29.450 So this jocking from the back of the job, though, so we try to find out what is the exact sort of 19:29.480 --> 19:36.390 the question and what answer has the machine given to us and the difference between the exact answer, 19:36.400 --> 19:43.120 the actual answer and the answer, which the machine has given, is called the error. 19:44.260 --> 19:49.860 It is the editor value between the actual design and what the machine has predicted. 19:51.030 --> 19:59.970 And when we add on the edit for all the observations, because we will not have just one observation 19:59.970 --> 20:04.770 rate, we will not have just one practice question, there will be multiple practice questions. 20:05.400 --> 20:12.060 So then we will be evaluating on all the practice questions, the added value for all the questions 20:12.060 --> 20:16.310 will be summed up, and that is the cost of the model. 20:20.170 --> 20:28.090 That is called the cost of the model, which is the sum of all the ED films and. 20:30.050 --> 20:40.040 Reveals just one particular error, though, I guess, is to reduce the entire cost that invasion cost 20:40.040 --> 20:48.890 might reduce by reducing land value also or by a cumulative reduction in all the edibles. 20:49.730 --> 20:57.890 So our target is to reduce the cumulative Edwardo instead of reducing the error for just one. 20:57.890 --> 21:05.360 But because if we are reducing the error for just one value in that case, the for other values might 21:05.360 --> 21:05.940 increase. 21:06.230 --> 21:07.570 So we don't want that. 21:07.850 --> 21:10.490 We want to reduce the overall. 21:16.290 --> 21:25.580 So let us see, so we have this right is equal to a forfeit, so we have different X values. 21:25.980 --> 21:34.780 So out of those different X values, we want to find out a function which we give right as the output. 21:35.670 --> 21:41.710 So this function is actually the linear equation which we will be forming. 21:42.300 --> 21:51.480 So the equation will be doubled plus the next one plus B, that extra plus Lusby, that's been too three 21:51.750 --> 21:54.120 plus wait for export and so on. 21:56.970 --> 22:05.010 So why is actually that target value, for example, it is the price of the flight ticket or height 22:05.010 --> 22:11.100 of the person or the interest rate, which we are trying to find out, which is the fundamental value. 22:13.450 --> 22:22.510 I'm the one extra extra export, these are the independent values or the features or the attribute, 22:22.900 --> 22:28.900 which is the values which we provide so that the relationship between these values and this value can 22:28.900 --> 22:29.440 be found. 22:29.830 --> 22:31.180 So what are these values? 22:31.180 --> 22:36.510 These values could be season or decision destination, the nearest holiday date. 22:36.780 --> 22:43.480 So these kind of events which actually impact the value of the target value. 22:46.090 --> 22:55.140 And when we are having the training, they originally Devinder X values are already present with us. 22:55.390 --> 23:01.580 So we already know what the V and X values are, hence the lion X value constant. 23:04.070 --> 23:11.120 And then we try to say to various be the values we want to bring, bring the predictions closer to the 23:11.120 --> 23:12.340 actual value. 23:14.310 --> 23:17.730 Now, the value, which we predict is very. 23:18.540 --> 23:24.120 So when we subtract Vizag from the actual value, we get the added value. 23:25.140 --> 23:30.930 So we have sold in value, which we have predicted and subtracted from the actual value. 23:30.960 --> 23:32.570 And this is all the added value. 23:35.210 --> 23:42.290 So in this equation, we already know what is we already know what makes one extra extra is we only 23:42.290 --> 23:44.690 need to find out the be done all be done. 23:44.900 --> 23:46.490 That will be that revalues. 23:51.770 --> 23:59.660 So what is the process, the process will be to convert all the variables into numerical, whether they 23:59.660 --> 24:09.360 are or the that they are categorically that we will have to convert all these variables into numerical. 24:10.190 --> 24:16.720 And after that, the categorical variables should be converted into that many variables. 24:17.420 --> 24:20.190 We have to remove the outliers in the data. 24:20.450 --> 24:26.480 And in case there are any missing values, we will have to imbue those basic values by. 24:28.660 --> 24:33.020 Placing the mean value or the median value in the people? 24:34.800 --> 24:41.460 Once we have done this, the transformations made to bringing data should also reflect in the testing 24:41.460 --> 24:41.860 data. 24:42.360 --> 24:48.840 So whatever transformation we will be doing, all the training data, the same information we will be 24:48.840 --> 24:50.170 doing all the testing, the. 24:52.460 --> 25:00.020 Because the machine does not really know what the interest rate is or what the age of a person is, 25:01.070 --> 25:09.770 the machine will not understand what the machine will only understand what the numbers are. 25:10.920 --> 25:19.730 So it will simply apply the same operations on the numbers which see sees in the training data, all 25:19.740 --> 25:22.380 the lines which it sees in the testing data. 25:23.190 --> 25:29.150 So that is the reason why the transformation's, which are made in the training data, should be made 25:29.160 --> 25:37.530 in the testing data so that the formula which the machine has created should be applicable on both the 25:37.950 --> 25:38.390 six. 25:43.670 --> 25:55.190 Metrics are the types of measures which are used to evaluate the more so different type of things, 25:55.190 --> 25:59.290 that might mean square error and the mean absolute. 26:01.280 --> 26:02.810 So you mean square? 26:02.810 --> 26:07.850 And it is doing that would also mean squared error. 26:08.900 --> 26:12.160 I mean, of the absolute is. 26:12.920 --> 26:17.280 And why have the applied, the squared and the absolute value. 26:17.660 --> 26:21.740 This is because there is no detection in the. 26:23.030 --> 26:29.090 We will understand this in a while, but for now there is no specific direction in the error. 26:29.100 --> 26:35.930 That is why we use one square and the mean absolute and the goodness of the predictions would depend 26:35.930 --> 26:38.690 on the scale of the target values. 26:38.930 --> 26:45.440 So if we are trying to predict how good the values, how how would our predictions. 26:46.040 --> 26:54.260 So that is not something which will be decided on our own basis, but it will be decided on the scale 26:54.260 --> 26:54.980 of the data. 26:56.130 --> 27:03.210 So let's say we are walking with some chemicals, we are working in a chemical industry and we have 27:03.210 --> 27:05.340 to create some products based on that. 27:06.990 --> 27:14.850 So when we are working on a chemical industry and we are creating a particular kind of company or some 27:15.180 --> 27:17.880 specific kind of chemical. 27:18.120 --> 27:27.570 So in that case, even a small millimeter or small animal of the chemical can change the composition 27:27.570 --> 27:39.390 of what we think is right when we are looking in a legacy soil the packaging industry, so that if we 27:39.390 --> 27:48.060 are dealing with gallons or we are dealing with tons of salt, then small milligram of salt will not 27:48.060 --> 27:48.960 make a difference. 27:49.800 --> 27:53.950 So you see, it is depending on the target value. 27:53.970 --> 27:56.280 What exactly are you dealing with? 27:56.550 --> 28:00.080 The scale of the production will change. 28:00.900 --> 28:07.440 So if the production value is off by, what, even one million, the chemical industry, it will cause 28:07.440 --> 28:08.460 a lot of issues. 28:08.910 --> 28:17.970 But if the prediction is off five, five, five grams or 10 grams in the packaging industry, then it 28:17.970 --> 28:19.730 will not make much difference. 28:21.620 --> 28:24.860 So this is what witness prediction? 28:27.600 --> 28:35.760 So now let's say we have the actual price as a thousand nine eighty nine fifty one thousand twenty one 28:35.760 --> 28:41.430 thousand fifty, and we have predicted the prices using the formula. 28:41.610 --> 28:43.980 And these are the predicted prices, which we have. 28:46.000 --> 28:53.880 Now, these predicted prices, you can see the difference between this is the difference between diesel 28:54.250 --> 28:54.820 will be. 28:55.970 --> 28:58.460 Here, the differences here. 28:58.490 --> 29:02.650 The differences mistake here, the difference is minus Hundert. 29:03.020 --> 29:11.860 So now if I have these errors, if I add these errors, the sum comes out to be only 10. 29:15.120 --> 29:19.710 But that does not mean that the issue is of only the. 29:21.250 --> 29:29.320 The editing is actually of 371, but because these negative signs are being canceled out, the edited 29:30.400 --> 29:38.020 out the week, that this is the reason why they need to consider the absolute leader, the absolute 29:38.020 --> 29:46.450 edit will actually keep in consideration all the added values, whether it is a negative or positive 29:46.450 --> 29:46.600 it. 29:48.780 --> 29:53.710 So the main focus and predictive analysis is to reduce the overall. 29:54.900 --> 29:59.420 Not what is the overall rate of the overall and it is VITAC. 30:00.560 --> 30:07.370 This is the revenue which is generated from the function which we have applied, so the added value 30:07.370 --> 30:14.360 will be via minus by half and the sum of all the other values will be the cost of the. 30:15.540 --> 30:16.860 Model, which we have done. 30:23.310 --> 30:27.660 Now, us have a look at this, so this is the input value. 30:31.050 --> 30:39.510 This is the model which we have generated, so this input value of input into the machine learning model 30:39.840 --> 30:48.270 should have already should have generated value, but because our model is not perfect, so it will 30:48.270 --> 30:49.350 generate a value. 30:49.350 --> 30:49.710 Right. 30:51.900 --> 30:55.590 So this is the actual output and this is the predicted. 30:56.160 --> 30:57.670 This is what we have predicted. 30:57.960 --> 31:01.140 So the added value will be the actual value. 31:02.640 --> 31:09.770 And predicted values difference, so we will find out the difference between the value and the actual 31:09.770 --> 31:11.740 value, and that is the added value. 31:13.170 --> 31:21.900 And then we will add these added values after applying the absolute function or by squaring these values 31:21.900 --> 31:25.620 so that the clients do not actually cancel out the devil values. 31:27.690 --> 31:29.580 So here we have a block. 31:30.900 --> 31:37.140 This plot is between height and weight, so you can see there is a positive relationship between height 31:37.140 --> 31:40.710 and weight, so as the height increases, the breathing freezes. 31:41.040 --> 31:43.980 So we have the situation of like. 31:45.290 --> 31:46.640 Between height and weight. 31:48.250 --> 31:54.970 And then there the line intercepts that Y-axis is called the intercept. 31:56.870 --> 32:04.550 And the scale of the equation is change of five divided by the changing X. 32:04.670 --> 32:06.820 This is called the slope of the equation. 32:07.910 --> 32:12.620 So the equation will be very easy, equal to. 32:14.750 --> 32:25.250 The school into X plus that B double, so this is the creation of the line, which we have, that is 32:25.250 --> 32:35.840 why is equal to a mixed blessing where M is the slope, X is the X value, and C is the intercept where 32:35.840 --> 32:38.050 the point intercept the Y axis. 32:43.540 --> 32:48.460 In the next section, we will discuss about the cost and the function.