WEBVTT 00:00.120 --> 00:01.040 Welcome back. 00:01.080 --> 00:04.080 We learned about distribution of MPG histogram. 00:04.200 --> 00:13.480 Now let's create a heatmap visualization to show relationships between numerical features in data set. 00:13.520 --> 00:15.840 Create a new code set. 00:15.920 --> 00:19.000 Start with plt dot figure. 00:19.200 --> 00:25.560 Figure size ten by eight and set the heatmap using SNS. 00:25.600 --> 00:34.040 This is the newly created and imported library that we used before, which is the seaborn as SNS. 00:34.160 --> 00:41.040 And as I told you, Seaborn data visualization library built on top of Matplot library. 00:41.080 --> 00:47.880 It provides a high level interface for creating attractive statistical graphics. 00:47.880 --> 00:54.240 So we're going to use this library to create our heat map SNS dot heatmap. 00:54.240 --> 00:56.040 Creating the heatmap. 00:56.080 --> 00:58.560 Passing the parameters. 00:58.720 --> 00:59.910 Data set dot. 00:59.910 --> 01:02.630 Correct the annotation true. 01:02.790 --> 01:03.790 See map. 01:03.790 --> 01:04.230 Cool. 01:04.230 --> 01:04.710 Worm. 01:04.910 --> 01:08.270 Worm and line width is 0.5. 01:08.630 --> 01:09.350 The title. 01:09.350 --> 01:13.910 Give it the coloration heatmap and the plot dot show. 01:13.950 --> 01:16.950 I'm using the Matplot library. 01:17.070 --> 01:23.150 Also we can name it as feature coloration, heatmap or matrix. 01:23.190 --> 01:25.870 Okay, let's run and see. 01:26.070 --> 01:26.950 And here we go. 01:26.990 --> 01:29.030 This is our heatmap. 01:29.070 --> 01:33.790 It's very very amazing I love this type of heatmaps. 01:33.790 --> 01:34.750 And here we go. 01:34.950 --> 01:41.390 Let me explain what we've done here and what this heatmap means. 01:41.430 --> 01:42.830 Back to our code. 01:42.830 --> 01:48.110 Here we have the coloration dataset dot coloration. 01:48.150 --> 01:55.110 The coloration calculates the the numerical values between the features. 01:55.350 --> 02:03.370 Coloration values range between -1 to 1, so please write those notes down. 02:03.410 --> 02:12.730 Dataset dot correlation calculates the correlation matrix between all numerical columns minus one to 02:13.010 --> 02:13.890 plus one. 02:13.930 --> 02:20.090 The correlation values ranges from -1 to 1 one. 02:20.130 --> 02:23.650 Perfect positive correlation zero. 02:23.690 --> 02:27.050 No correlation minus one. 02:27.090 --> 02:29.290 Perfect negative correlation. 02:29.290 --> 02:32.730 Now we have the heat map. 02:32.730 --> 02:35.490 And this is the main heat map creation. 02:35.690 --> 02:39.650 As I told you we passed the correlation matrix. 02:39.850 --> 02:44.330 The annotation through displays the actual correlation values on the heatmap. 02:44.330 --> 02:50.090 Cool warm colors were blue represents negative coloration. 02:50.090 --> 02:54.330 So the negative coloration would be in blue. 02:54.570 --> 02:59.400 So blue color represents the negative coloration. 02:59.680 --> 03:04.280 The red color represents the positive color correlation. 03:04.520 --> 03:13.880 Okay and zero centers the color map to zero, which is neutral, and the title is uh feature coloration 03:13.920 --> 03:15.760 correlation heatmap. 03:15.920 --> 03:20.520 And let's display the final result like this. 03:20.520 --> 03:22.440 So we're going to interpret it. 03:22.480 --> 03:30.840 Diagonals always show 1.0 with each variable perfectly correlated with itself. 03:30.840 --> 03:41.000 So MPG is correlated with mpg origin is present and correlated with origin symmetric mirrored across 03:41.000 --> 03:42.040 the diagonals. 03:42.040 --> 03:43.600 So we have asymmetry. 03:43.920 --> 03:47.840 Color intensity indicates coloration strength. 03:47.840 --> 03:58.940 As I told you, blue is, uh, the high blue and the more intense blue is perfectly negative coloration 03:58.940 --> 04:02.820 and positively coloration is in red. 04:02.820 --> 04:08.340 So NPG is strongly correlated with NPG with itself. 04:08.380 --> 04:13.420 Numbers show exact coloration correlation coefficients. 04:13.460 --> 04:21.900 Okay, so this coloration heatmap shows relationships between automotive features from what appears 04:21.900 --> 04:24.340 to be a vehicle data set. 04:24.340 --> 04:33.780 So here we have we have eight columns and cylinders placement horsepower weight acceleration model year 04:33.780 --> 04:35.220 and origin. 04:35.340 --> 04:43.860 So this is a very important way to display the feature coloration using this heatmap. 04:44.180 --> 04:47.780 Strong positive correlation red high value. 04:47.780 --> 04:52.680 So here we have this square this red square. 04:52.720 --> 04:57.040 So for example cylinders and displacement. 04:57.040 --> 04:58.240 So this is the cylinders. 04:58.240 --> 05:03.520 And this is displacement 0.95 very strong cylinders. 05:03.520 --> 05:08.240 And weight is 0.93 cylinders. 05:08.360 --> 05:09.920 Uh sorry cylinders. 05:09.920 --> 05:14.000 And weight is 0.90.93. 05:14.040 --> 05:15.680 Displacement and weight. 05:15.680 --> 05:17.440 They are very very strong. 05:17.720 --> 05:23.080 Displacement and horsepower is 0.9. 05:23.080 --> 05:25.040 Also very strong here. 05:25.080 --> 05:34.320 As an interpretation, we conclude that vehicles with more cylinders tend to have larger engines displacement 05:34.320 --> 05:38.160 more, horsepower and weight more. 05:38.320 --> 05:43.760 Also, we have strong negative correlation blue high negative values. 05:43.760 --> 05:46.520 Here we have this blue box. 05:46.520 --> 05:57.510 And here we have a blue box the mpg and weight, so mpg and the weight is zero -0.83. 05:57.750 --> 06:14.150 Strongest and negative mpg and displacement -0.81 mpg and cylinder 0.75 78 and mpg and horsepower -0.78. 06:14.270 --> 06:24.350 Fuel efficiency mpg decreases as vehicles get heavier, have larger engines, more cylinders, and more 06:24.510 --> 06:25.590 horsepower. 06:25.590 --> 06:32.270 So this is the coloration that this heatmap tries to gives us. 06:32.270 --> 06:36.990 And to, to uh, allow us to conclude this conclusion. 06:36.990 --> 06:43.790 So positive relationships between MPG and model year is 0.58. 06:43.830 --> 06:52.580 You notice that this mpg and model year 58 mpg and origin is 0.57. 06:52.620 --> 07:00.100 Newer vehicles and certain origins tend to have better fuel efficiency, but this is not a very big 07:00.100 --> 07:00.540 deal. 07:00.540 --> 07:05.260 So I want from you to understand the colors and the coefficients. 07:05.500 --> 07:12.060 Vehicles with more horsepower tend to have slower acceleration, likely because they are heavier. 07:12.220 --> 07:15.340 If we get the horsepower and the acceleration. 07:15.340 --> 07:21.500 So here the acceleration and the horsepower -0.69. 07:21.540 --> 07:26.500 Vehicles with more horsepower tend to have slower acceleration. 07:26.740 --> 07:29.620 Those are the negative correlation. 07:29.660 --> 07:30.140 Okay. 07:30.380 --> 07:39.580 So this heat map reveals the classic automotive engineering trade off between performance and fuel economy.