WEBVTT 00:00.080 --> 00:00.760 Welcome back. 00:00.760 --> 00:03.800 We finished the first step, which is preparing the data. 00:03.840 --> 00:07.320 Now let's train the linear regression model. 00:07.320 --> 00:10.200 We'll use the linear regression to fit a line. 00:10.240 --> 00:13.720 Price equals to W times size plus b. 00:13.760 --> 00:17.640 This is the linear regression relationship. 00:17.680 --> 00:21.080 Linear regression is a model from sklearn. 00:21.320 --> 00:22.560 Linearmodel. 00:22.600 --> 00:30.080 Because we used here linear regression that tries to predict a numeric value like price based on one 00:30.080 --> 00:32.600 or more input features like size. 00:32.880 --> 00:38.080 It assumes that the relationship between the features and the target is linear. 00:38.080 --> 00:47.040 For example, it can be approximated by a straight line in 2D or a flat plane hyperplane in higher dimensions. 00:47.040 --> 00:50.360 So think about it as a linear line. 00:50.360 --> 00:54.800 So if we go up here we have uh a straight line. 00:54.800 --> 00:58.720 If you if you see it like it's a straight line. 00:58.720 --> 01:00.320 So this is the linear. 01:00.360 --> 01:04.720 This is the linear relationship between size and the price okay. 01:04.760 --> 01:08.570 So this is the basic linear regression model. 01:08.610 --> 01:15.250 At first we need to set the model and tell the Python interpreter. 01:15.410 --> 01:19.330 With that we need to use this as a linear regression. 01:19.330 --> 01:22.650 So model equals to linear regression. 01:22.690 --> 01:26.890 Here we create an instance of the linear regression class. 01:27.050 --> 01:32.010 At this point the model exists but hasn't learned anything yet. 01:32.210 --> 01:35.650 It's like a blank sheet waiting to be trained. 01:35.930 --> 01:36.490 Okay. 01:36.730 --> 01:41.170 And this is from the library that we imported before. 01:41.370 --> 01:46.090 The second step is to use this model to train it. 01:46.250 --> 01:49.850 So here we are training the model. 01:49.970 --> 01:52.530 We use model dot fit. 01:52.530 --> 02:00.370 So I created an instance of the linear regression class and use this model instance to train the model. 02:00.570 --> 02:02.090 Model.fit. 02:02.370 --> 02:06.690 The fit method is used to train the model. 02:06.690 --> 02:08.010 So fit. 02:08.050 --> 02:10.930 I love to write notes down. 02:10.930 --> 02:15.290 Please pay attention to those notes because they are very important. 02:15.290 --> 02:23.170 Please write them down and because it's the revision and helps you understand the code better. 02:23.330 --> 02:26.370 It looks like a very simple. 02:26.410 --> 02:27.130 Yes. 02:27.130 --> 02:30.250 So Model.fit is the function. 02:30.250 --> 02:33.090 This is the method that trains the model. 02:33.290 --> 02:40.930 It looks at the training data, exit train for sizes, white train for prices, and calculates the best 02:40.970 --> 02:49.650 slope and intercept B that minimizes the difference between predicted and actual prices. 02:49.650 --> 02:53.210 So here training the model involves many steps. 02:53.210 --> 03:01.610 So this function gets xtrain and ytrain, which are the sizes and prices respectively, and calculates 03:01.610 --> 03:10.530 the best slope and intercept b that minimizes the differences between predicted and actual prices. 03:10.730 --> 03:17.420 After this step, the model knows the relationship between house size and the price. 03:17.620 --> 03:21.140 Again, guys, I want from you to focus with me here. 03:21.140 --> 03:25.580 We have a linear relationship between size and the price. 03:25.580 --> 03:26.900 We have this straight line. 03:26.900 --> 03:34.140 And as you know we have this relationship price equals to three size plus two. 03:34.180 --> 03:38.180 This is normally a linear relationship between size and price. 03:38.220 --> 03:41.660 It's in form of y equals ax plus b. 03:42.060 --> 03:45.020 So A is the slope or w is a slope. 03:45.020 --> 03:55.540 You can make it like w plus uh here like linear relationship y equals to w x plus b. 03:55.700 --> 04:00.300 So the slope is called w and intercept is b. 04:00.700 --> 04:07.340 Those are used by the model to minimize the difference between predicted and actual prices. 04:07.380 --> 04:13.020 After this step the model knows the relationship between house size and price. 04:13.060 --> 04:13.500 Okay. 04:13.740 --> 04:17.820 The third step is the learned parameters. 04:17.870 --> 04:21.550 Here we have the learned parameters w. 04:21.830 --> 04:27.950 The model coefficient contains the weights or the slopes learned from this feature. 04:27.950 --> 04:34.670 Since we have only one feature, which is the size, we take the first element with zero. 04:34.670 --> 04:36.750 So model dot coefficient. 04:37.110 --> 04:43.510 Since we have only one feature, we take only the zero from this array. 04:43.710 --> 04:54.990 So since we have only one feature, B contains the bias or intercept the value of price when size equals 04:55.030 --> 04:56.150 to zero. 04:56.190 --> 04:59.830 Let's run this cell and here we go. 04:59.870 --> 05:08.110 Model trained learned equation price equals to 2.9 times size plus 2.7. 05:08.270 --> 05:19.030 So after training this model this model gets and learned the equation that price equals to 2.9 times 05:19.070 --> 05:21.750 size plus 2.7. 05:21.790 --> 05:24.110 So this is the magic. 05:24.150 --> 05:27.670 This is how this model get to train. 05:27.790 --> 05:29.310 How this model. 05:29.510 --> 05:40.230 Get this equation and conclude this equation y equal 2.9 x plus 2.7. 05:40.430 --> 05:45.230 So price equals to 2.9 times size plus 17. 05:45.390 --> 06:03.830 So each 1000ft² increases in house size adds roughly um two $2,900 to the price starting from 2007 06:04.150 --> 06:07.670 $2,170. 06:07.910 --> 06:16.670 So in general, what we've done here, we created a linear regression model, training it on the training 06:16.670 --> 06:24.390 data, extract the slope x and intercept b, print the learned linear equation.