WEBVTT 00:01.030 --> 00:05.990 Higher in this session, we will be implementing hierarchical clustering. 00:06.340 --> 00:13.840 So let us begin the first thing which we will be doing is imposed a different library, such as import 00:14.780 --> 00:17.310 fundus as Phoebe. 00:20.570 --> 00:28.760 No, as and then we will import matplotlib. 00:31.350 --> 00:32.580 Don't buy plot. 00:35.120 --> 00:42.260 As a lot, next, we will import Seabourne. 01:03.080 --> 01:06.170 So we lean forward from this. 01:08.900 --> 01:13.900 We don't really need this one, let us import. 01:17.190 --> 01:18.420 And Skillern. 01:19.620 --> 01:20.350 Dr.. 01:21.500 --> 01:23.330 Pre processing. 01:27.340 --> 01:28.320 Import. 01:29.890 --> 01:35.230 Sked, this particular library will actually help us to scale the data. 01:37.240 --> 01:45.160 Next from Escalon dot cluster, we will import the. 01:48.000 --> 01:50.190 Argumentative flustering. 02:01.670 --> 02:07.760 And the last we will implode this school, which is a metric for evaluation. 02:28.960 --> 02:30.790 Let me run this. 02:40.410 --> 02:42.090 Thought Matrix. 02:45.470 --> 02:53.210 So after this, we will be importing the items dataset, which is the data set, which we will be using 02:53.210 --> 02:55.670 for this particular tutorial. 02:56.030 --> 02:59.780 So let me import defined by other. 03:02.150 --> 03:06.140 Is equal to the filename that is itis. 03:07.260 --> 03:08.490 NSV. 03:10.020 --> 03:19.350 Next, I will give the data frames while the iris is equal to the dot read. 03:21.210 --> 03:22.080 CSFI. 03:25.030 --> 03:26.970 Inside this, I give the other. 03:27.610 --> 03:36.520 So it should be treated like this only and I will give I just thought hard for you to see the detective. 03:37.120 --> 03:38.980 So this is the data that we have. 03:39.520 --> 03:42.100 So this data contains idee. 03:43.510 --> 03:51.400 The sibilant separate the battlement battle delivered on the species of the. 03:52.540 --> 04:00.460 Fly, which we have now, because we don't want to classify anything, we are just implementing the 04:00.460 --> 04:01.690 clustering algorithm. 04:02.180 --> 04:09.520 So clustering algorithm, we don't need the target values and we also don't need this idea because it 04:09.520 --> 04:11.230 is a completely redundant column. 04:11.240 --> 04:13.090 So I will drop these columns. 04:13.630 --> 04:18.640 So I simply say iris dot drop. 04:23.680 --> 04:30.730 And in this, I will give me a list of the columns which I want to drop, that this idea. 04:33.210 --> 04:38.040 And the other one is species. 04:41.720 --> 04:51.800 Next, I will give the axis, which is one, and in place equal to two. 04:53.550 --> 04:58.860 Next, I will print the data set for you, so I'll say I just started. 05:00.350 --> 05:09.350 And this is the data center, which we have now, now let us create clusters out of the first two columns. 05:10.350 --> 05:11.190 So. 05:12.310 --> 05:19.750 For this, we will, first of all, let us create the new grandmotherhood, so let us actually visualise 05:19.750 --> 05:23.010 how hierarchical clustering would be working. 05:23.560 --> 05:29.800 So we will import the library from sci fi. 05:29.810 --> 05:35.380 So we will see from sci fi dot Plaistow. 05:36.830 --> 05:37.520 Dorte. 05:38.880 --> 05:39.880 Hirers. 05:41.210 --> 05:44.420 He import. 05:46.570 --> 05:47.320 Dendle. 05:49.250 --> 05:50.270 Then, Joe. 05:51.590 --> 05:52.130 Gramme. 05:53.490 --> 06:04.120 And we want linkage from it, so we will import these libraries and now we will need to scale this particular 06:04.120 --> 06:07.210 data to so how we will scale the Zetas. 06:07.210 --> 06:11.430 So to scale this data, we have imported this particular package. 06:13.950 --> 06:17.130 So we will simply see Iris. 06:18.790 --> 06:20.860 Underscores the steady. 06:21.930 --> 06:25.650 Is equal to feed dot data frame. 06:28.350 --> 06:29.280 And the. 06:30.550 --> 06:31.240 Give me. 06:34.270 --> 06:41.460 Details, so we will skew the details of this skin and insight designed to give the dataset so Iris, 06:41.710 --> 06:43.540 so this will simply scale the data. 06:43.580 --> 06:46.630 Now, now I will have to give the columnists. 06:46.630 --> 06:58.600 So the column will be columns equal to I can put this into a list for them and I get ideas dot columns 06:58.600 --> 06:58.900 from. 07:00.600 --> 07:06.870 So this will basically give me the list of the columns and I'm converting these columns into list from 07:07.770 --> 07:10.370 next, let me display the data for you. 07:10.380 --> 07:16.820 So I see Iris Dot sorry, underscore Steve. 07:18.150 --> 07:24.730 Not this time, so that we can see what exactly is present in this data. 07:25.320 --> 07:26.250 So this is the. 07:27.580 --> 07:29.590 Data which has been circulating now. 07:34.890 --> 07:42.110 Next, less let us create our program out of it, so to create an anagram, letters of creative, the 07:42.120 --> 07:42.680 good for us. 07:42.690 --> 07:45.750 So we will see if not not finger. 07:46.840 --> 07:53.110 And inside this, we will give the big size, the big size is. 07:55.270 --> 07:57.250 Let us say 10, Gummow five. 07:58.770 --> 08:01.560 And we want to have a lot. 08:02.640 --> 08:05.120 Got X neighbor. 08:06.660 --> 08:09.090 So in the X level, I want to give. 08:13.100 --> 08:13.960 Somphone. 08:17.400 --> 08:21.400 Or simply say IDEX for simplicity. 08:22.140 --> 08:23.970 And next, we will have. 08:26.810 --> 08:28.700 Or just a simple index. 08:32.540 --> 08:35.850 Next, we will have the available. 08:41.700 --> 08:44.610 So in the vilely will let us have the distance. 08:47.360 --> 08:49.220 Next, we will create. 08:50.100 --> 08:51.180 The language. 08:52.870 --> 09:00.610 So this will create the linkage for us, so we have given linkage and the data set and the method which 09:00.610 --> 09:08.590 you want to create the linkage with and then from the data, we want to print the linkage which we have 09:08.590 --> 09:16.340 created and provide the rotation and different details about the levels on the road. 09:17.110 --> 09:18.670 So we will just print this. 09:19.960 --> 09:21.730 It will take a little time. 09:22.880 --> 09:28.040 And this is the dental of which has been created, so here you can see the. 09:29.450 --> 09:35.600 No signs of the clusters, the values which are present, the indexes. 09:36.570 --> 09:41.670 So here you can see that this is the second and next, so this is in the next. 09:46.610 --> 09:52.240 You can see that these are the glass doors which have been created, so if we want to have two clusters, 09:52.240 --> 09:55.980 we can simply break this distance equal to 15. 09:56.320 --> 09:59.680 And these would be the two clusters which will be developing. 09:59.890 --> 10:06.580 If you want three clusters, then we can break it at this point, and then we will have one, two and 10:06.580 --> 10:07.360 three cluster. 10:07.360 --> 10:09.760 That is green, red and blue. 10:11.760 --> 10:18.180 So these are we have been highlighted, so these are the actual clusters which are present in this dataset, 10:18.690 --> 10:21.610 which is actually being depicted here. 10:22.290 --> 10:28.590 So now we will go ahead and plot a cluster. 10:28.590 --> 10:35.310 So we report that we will again take a subset of data so that we can visualize it more properly. 10:35.340 --> 10:39.510 So for that, let's take only these two columns. 10:40.970 --> 10:42.320 So we will see. 10:43.580 --> 10:44.510 Iris. 10:45.630 --> 10:47.760 Is equal to Iris. 11:09.680 --> 11:13.580 We have updated the data set, now we can see it is. 11:14.990 --> 11:16.070 Describe. 11:20.780 --> 11:29.390 So here you can see the values now begins and the scale, the data, or we could have simply done this 11:29.390 --> 11:31.410 from the standardized details. 11:31.430 --> 11:33.380 And it does do it with the standardized data. 11:42.430 --> 11:47.920 So here you can see we have simply used these standardized data, which we had created earlier. 11:51.210 --> 11:59.280 Now we will big of the clusters, so we will see clusters now also for creating the clusters we value 12:00.000 --> 12:03.510 from several number of clusters, so. 12:04.420 --> 12:13.510 Create this particular for loop, then we are outrating on top of clusters ranging from two to 19 and 12:13.660 --> 12:16.460 we have created the cluster model object. 12:16.900 --> 12:22.870 So this is the model object which has the object of argument, of clustering where we have provided 12:22.870 --> 12:30.970 the number of clusters, the affinity, which is the distance, calculating metric which we have used 12:31.330 --> 12:34.120 and the type of linkage which we have decided. 12:35.250 --> 12:44.550 And we have provided the data, so we have the models which then predicted the model using the to standardize 12:44.550 --> 12:47.850 data, which has only these two columns for now. 12:48.890 --> 12:52.190 And we have gathered the label in Kluster label. 12:54.170 --> 13:00.230 And after getting these values, we have actually calculated this old school. 13:01.630 --> 13:02.650 Using the. 13:03.650 --> 13:08.390 Standard data set and the label which has been generated. 13:09.680 --> 13:18.560 And we are next printing the score for each number of clusters, so when we will run this, it will 13:18.560 --> 13:20.580 give us all the number of clusters. 13:20.840 --> 13:28.640 Now, we already know that the score, which is closest to one, is the best Szilard school. 13:28.910 --> 13:34.380 And unfortunately, here the score is around zero point four. 13:34.850 --> 13:39.920 So we will select this particular Senate school, which appears to be the maximum one. 13:40.980 --> 13:48.420 Out of all the Senate scores, which we have, so we will select and is equal to three. 13:49.330 --> 13:54.460 So let us actually implement this for these three cluster's. 13:56.100 --> 13:57.980 So we will pick this up. 14:09.340 --> 14:15.820 In this, we will put three, which is the best question which we have found. 14:18.930 --> 14:22.470 So we can see these live average values for this. 14:23.680 --> 14:28.320 Which comes out to be zero point forty, which is exactly the one which we have got. 14:29.590 --> 14:33.620 So you can see that even out there running this particular method. 14:33.640 --> 14:39.040 Again, it did not change the index value, it came up to be seen. 14:39.520 --> 14:49.330 That is because the idea of clustering or we can see the hierarchical clustering is always giving the 14:49.330 --> 14:53.110 same results, no matter how many times we are running it. 14:54.160 --> 15:00.880 Because it is not dependent on any external criteria, it is dependent on the distances between the 15:00.880 --> 15:02.500 points which are not changing. 15:05.320 --> 15:10.870 Next, what we will be doing is so next we will create a blog. 15:12.530 --> 15:16.160 So to create this particular plot, let us. 15:17.790 --> 15:25.080 Get the neighbors for this, so for Lebas, these are the neighbors which we have. 15:26.720 --> 15:34.430 And let us look at these levels to the data set, which we already have, so we will see Iris. 15:35.600 --> 15:41.750 Under Iris and we will save Mehboob. 15:42.790 --> 15:48.010 And unbelievable, we will for the values from the cluster labels. 15:51.670 --> 15:55.790 Next, we will look at of. 15:56.620 --> 16:01.990 So for that, we will see there's not a lot. 16:03.180 --> 16:09.770 And inside this, we will give the details so we don't want to further aggression, so we will save 16:09.940 --> 16:10.470 dreg. 16:12.280 --> 16:13.330 Equal to. 16:14.430 --> 16:15.450 Foy's. 16:19.570 --> 16:20.950 Next, we will. 16:22.600 --> 16:30.600 Provide the X and Y values so that X and Y values will be the values which we have here. 16:34.740 --> 16:42.660 So these are the X and Y values, so X is equal to this and this is why want. 16:44.610 --> 16:45.750 Next, we will. 16:46.900 --> 16:49.990 Provide the data set, so data is. 16:50.920 --> 16:51.970 Iris. 16:54.140 --> 16:55.370 And for the. 16:57.580 --> 17:00.820 Now, this, again, has to be in the eye of standard. 17:02.450 --> 17:05.990 And this also has to be the standard data. 17:07.630 --> 17:15.730 Next, which we will be doing is selecting the Hill, so we will give you as the cluster labels, which 17:15.730 --> 17:16.240 we have. 17:21.970 --> 17:24.250 Or we can simply give this one to. 17:29.910 --> 17:32.420 So let's create this plot. 17:34.280 --> 17:42.140 So this is the plot which have been generated, so here you can see that we have got three clear clusters 17:42.140 --> 17:42.800 created. 17:45.350 --> 17:54.020 So this is how we can use and limit on how people clustering for generating clusters for any type of 17:54.020 --> 17:54.630 dataset. 17:55.220 --> 18:01.980 Now we can use a different number of variables here. 18:02.000 --> 18:07.770 We have used only two because it is easy to visualize for these two videos. 18:08.270 --> 18:15.500 But we could have used all four of these and it would have created as good clusters as these ones. 18:16.310 --> 18:20.300 But the only drawback here is we cannot apply this. 18:20.690 --> 18:28.970 Or I must say that it does not prescribe to apply this for a large dataset, because then it would take 18:28.970 --> 18:33.020 a lot of time to generate the data from all the clusters. 18:33.620 --> 18:43.310 So this is about how the clustering next we will we will generate the gaming's clustering in comparison 18:43.310 --> 18:48.290 to the hierarchical clustering once we have learned about the gaming's clusters.