WEBVTT 00:00.990 --> 00:07.260 In this session, we will discuss the mode stacking and stacking and melting of data, so the first 00:07.260 --> 00:11.310 thing which we will do is we will impose the required library's. 00:13.360 --> 00:20.440 So we are importing Seabourne as Asness, Fondas as BDI and no as an. 00:25.150 --> 00:33.040 So now we will go ahead with looking at different through volume transformations, so stacking and stacking 00:33.040 --> 00:39.300 and arming the rebels have their own documentation act does not find that out. 00:39.370 --> 00:47.370 Or so you can refer to those in case you want to check the full documentation and learn more about them. 00:47.680 --> 00:52.240 But we will cover only what is actually required and the important parts of that. 00:53.150 --> 01:00.320 So this is the dip state, does it, so we are loading the students from the library. 01:01.580 --> 01:10.460 So the data set contains the total bill, the six more days, time and size of the of the particular 01:10.460 --> 01:10.740 date. 01:11.270 --> 01:13.780 And this is the detail of the bills. 01:13.790 --> 01:18.590 The amount the what is the amount? 01:18.620 --> 01:20.240 This is the bill amount. 01:20.720 --> 01:22.450 This is the gender of the person. 01:22.460 --> 01:29.690 This is if the person is smoking or not, then that the average someone has visited the restaurant, 01:30.020 --> 01:36.050 then the time when they have visited and the size of the group of people who have visited. 01:37.870 --> 01:43.510 So for this dataset, we will check details further. 01:43.720 --> 01:47.180 So we would, first of all, group the data. 01:47.410 --> 01:51.950 So here what we are doing is we have this entire data. 01:52.150 --> 01:56.800 So now we are grouping the data by the and then by six. 01:57.310 --> 02:04.370 And we are creating an aggregation with size of the group of people. 02:04.600 --> 02:05.860 So, like, we run this. 02:08.220 --> 02:16.110 So here you can see what happens here is that while we have the data, so it has brought all the data 02:16.590 --> 02:22.110 with the B, so we have all the details of each and every day, which we have. 02:22.410 --> 02:29.640 And for each day, it has further created a further subgroup of male and female. 02:29.970 --> 02:36.270 And on the male and female, it has created a size which has been aggregated. 02:37.360 --> 02:38.770 So this aggregation. 02:41.340 --> 02:49.690 So this aggregation is actually the sum of the size of the data, so it is adding the value for it. 02:49.950 --> 02:57.540 So it is actually finding out all the Sundays then for all the Sundays, a finding of all the Fenians, 02:57.840 --> 03:04.090 and it is just adding the Google, the number of people who are dead in the group. 03:04.530 --> 03:06.570 So this is how the abused group, but. 03:08.520 --> 03:17.070 Now we are getting to find out all these values, but it is a little difficult to how to analyze this 03:17.070 --> 03:17.990 kind of data. 03:18.210 --> 03:24.080 So we want the data to be in a column form where we can actually analyze it more to you. 03:24.510 --> 03:33.130 OK, because what happened now is that when we have created this grouping, this specific indexing available, 03:33.480 --> 03:40.340 so I cannot actually find out the particular value based on some particular index. 03:40.620 --> 03:46.800 I want the data which looks something like this so that I can retrieve certain values from it and apply 03:46.800 --> 03:49.860 for those based on the bottom does, which we have. 03:50.580 --> 03:52.500 So what we can do is. 03:53.620 --> 04:02.560 We can use Unstopping, so we have needed this tips, gva just dimps biodata, so what we are doing 04:02.560 --> 04:10.470 for that is we are applying the unstaffed method on top of and we are creating a on unstuck from this. 04:10.660 --> 04:11.890 So let me run this. 04:13.890 --> 04:19.210 So here I have the unstaffed version of this group. 04:19.740 --> 04:26.370 So what happened here is it has unstaffed the media and we have the indexing. 04:26.370 --> 04:34.560 So now the index, instead of being up to level the index, it is now just a single level index. 04:35.460 --> 04:41.340 So it has only one level of indexing, which is Thursday, Friday, Saturday, Sunday, which I can 04:41.340 --> 04:42.370 refer to easily. 04:42.540 --> 04:47.090 Now it is more understandable and we can use it more easily. 04:47.490 --> 04:49.860 So this is an unstoppable version of it. 04:50.400 --> 04:58.110 Now when we are giving the barometer so indiv is what it is doing is it is giving Thursday, Friday, 04:58.110 --> 04:59.280 Saturday, Sunday here. 04:59.280 --> 05:01.700 And it has unstaffed mean Fehmi. 05:02.040 --> 05:05.160 So how it was actually the first nibble here. 05:05.160 --> 05:07.670 We had those during this time in the sun. 05:07.950 --> 05:11.050 Then we have male and female as the next level. 05:11.490 --> 05:19.190 So on unshocking it has removed the male and female and only the first level, which was actually that 05:19.590 --> 05:24.490 now we will unstaffed on the basis of an argument which is zero. 05:24.840 --> 05:30.440 So when we strike on the basis of this, what happens is it is unstopping the level. 05:31.620 --> 05:32.220 Zero. 05:32.760 --> 05:36.030 So what it does is it does stacking the zero to. 05:37.520 --> 05:42.380 So it is unshocking zero levels or the keeping the level one. 05:43.990 --> 05:50.920 As of this summer, male and female remains as it is, and it has on Thursday, Friday, Saturday, 05:50.920 --> 05:51.300 Sunday. 05:51.310 --> 05:58.570 So we have the Thursday, Friday, Saturday, Sunday on the column by vigneron stacking on the basis 05:58.570 --> 05:59.130 of one. 05:59.380 --> 06:03.570 What it does is it unstaffed the level number one. 06:03.760 --> 06:06.280 So the level number one, which is. 06:08.110 --> 06:17.890 Female lend me get unstaffed, I'm it is represented on the lip, so which levies you want to unstaffed, 06:17.890 --> 06:21.970 you need to provide that as an input barometer for this function. 06:22.120 --> 06:23.050 Onestop function. 06:24.980 --> 06:33.800 Now, further, what we can do is if we want to have or new object, like when we are of looking at 06:33.800 --> 06:40.870 the columns of the upstart unstaffed, what we get is we get this multi-level kind of indexing. 06:41.720 --> 06:44.280 So this is the type start on stack. 06:44.900 --> 06:46.940 So in type, start on stack. 06:47.180 --> 06:50.390 We have this multilevel column name. 06:53.290 --> 06:57.970 So it is kind of difficult to understand how this actually works. 06:58.240 --> 07:05.920 So when we talk about the balloons, the balloons have size and undersize, it has male and female. 07:07.590 --> 07:15.060 So this is kind of difficult to interpret how we will filter out the values, how we will apply any 07:15.060 --> 07:19.030 operations on the volume, that becomes difficult in this particular case. 07:19.650 --> 07:21.660 So what we can do is. 07:23.670 --> 07:34.150 We can use tips us this was the deed of name name, and we have this size Godmamma, which is the double 07:34.170 --> 07:35.100 meaning of the. 07:36.600 --> 07:43.530 On the colony, so we done this, we can get the volume as it is, so I'm just getting the double reading 07:43.830 --> 07:48.480 this particular thing, which is the first colony and this is my second colony. 07:49.110 --> 07:59.290 So I can use this or what I can use I can simply use of copy this entire day Dufrene and then would 07:59.310 --> 08:03.570 dive into different and different levels and use that. 08:03.720 --> 08:08.500 But this kind of a tricky thing and it would take some time to understand also. 08:08.790 --> 08:11.990 So a better thing of a better thing. 08:12.000 --> 08:13.230 What we can do is. 08:14.270 --> 08:20.000 Instead of using this kind of a structure, it is convenient, if it is convenient, you can use this 08:20.270 --> 08:21.140 kind of a structure. 08:21.440 --> 08:26.150 If not convenient, then what we can do is we can simply use. 08:29.160 --> 08:34.410 So how we can do the stacking, so this is the start us them for me. 08:35.420 --> 08:40.520 This is what we have generated, so they absorb us, has the problems. 08:41.500 --> 08:44.890 Size me, size women, so these are problems. 08:45.280 --> 08:52.270 So when we do the US stuff, what it does is it again stacks back all the stuff on me. 08:52.960 --> 08:56.200 So I again get the level kind of a structure. 08:58.300 --> 09:06.130 Now, if you want to start with the level of zettl, it will start from the the zeroth level. 09:07.880 --> 09:14.650 And if you want to step on the basis of one, then it will stand the possible. 09:18.320 --> 09:28.040 So if we want to make the data by now, then either we can use the stack and sucking for changing the 09:28.040 --> 09:31.040 levels of data or we can use melting. 09:31.730 --> 09:33.710 So let us see how melting works. 09:33.980 --> 09:39.910 So we have this data, which is of having forced last height weight. 09:40.280 --> 09:44.570 So all the data is just in the form which we usually like to work with. 09:45.140 --> 09:48.890 Now what we can do is we can use dort make. 09:50.020 --> 09:55.560 And began giving the lady and we are giving the lenient on bubbleheads. 09:55.600 --> 09:59.900 We want to make it so we are giving the first and last name. 10:00.130 --> 10:02.140 So what did we do with it? 10:02.140 --> 10:02.890 Will simply. 10:04.370 --> 10:14.900 Change the look of fear, so it will change the height and weight into column name itself and it will 10:14.900 --> 10:16.280 give the value in front of you. 10:16.700 --> 10:18.190 So this is how it will make it. 10:18.200 --> 10:26.810 It will just by making you alone as it will increase the number of rules when we make something, it 10:26.810 --> 10:31.280 will increase the number of rules and when we are. 10:32.610 --> 10:40.440 Converting this so we can also look at this as cheese box index, so if you will just run this particular 10:40.440 --> 10:41.160 piece of fruit. 10:41.610 --> 10:46.590 So here we are just converting this first and last as the index theme's. 10:48.590 --> 10:50.400 So this is what we can do. 10:50.420 --> 10:54.800 We can have the next visas index names and. 10:55.910 --> 11:02.990 When we do this and the next, it will simply convert this and we will stop them together and we can 11:02.990 --> 11:05.200 convert it into the original. 11:05.960 --> 11:16.040 So it is a better use of the stuff because it kind of looks flawlessly might even be used on stacking. 11:16.040 --> 11:23.750 If my positive job is so, I would feel like you can try to follow what you like and follow what you 11:23.750 --> 11:33.050 are more comfortable with, but may think I'm stacking works more flawlessly and it's a lot more versatile 11:33.050 --> 11:33.680 in nature. 11:35.280 --> 11:40.410 Now, the other thing which we want to learn here is to convert into dummy variables. 11:40.680 --> 11:46.070 Now, we have discussed A of about political benefits. 11:46.350 --> 11:54.150 So let's say we have some variable as gender and it has two values, male and female, and we want to 11:54.150 --> 11:56.220 convert that into a zero one. 11:56.820 --> 11:59.760 So what we can do is we can create a dummy. 12:00.870 --> 12:03.870 So how we can do is here we have started. 12:05.480 --> 12:11.360 So it has this Volume six, which has values, female and me, so what I want to do here is I want to 12:11.360 --> 12:12.520 get the Emmys for this. 12:12.520 --> 12:15.310 So what I can do is I can simply these dummies. 12:16.730 --> 12:25.510 So if we can work the column six to six and the school, me, I'm six and the school Fehmi, I'm wherever 12:25.520 --> 12:32.360 the value of sex was me, it would be on the values to one under me and column. 12:32.660 --> 12:38.710 And wherever the value of Fehmi value was female and the female column, it will there one. 12:39.260 --> 12:43.660 So this is how we can dummy column instead of having this factual data. 12:43.820 --> 12:49.180 Now same thing can be done when we have a multiple number of captivities. 12:49.190 --> 12:51.440 Let's see if we have a job. 12:51.690 --> 12:58.640 Nine, then we can have as many number of categories as we want so we can simply use get dummies for 12:58.640 --> 12:58.990 that. 12:59.210 --> 13:09.190 And there are other methods also by which we can convert something into a dummy and then put Sadan for 13:09.320 --> 13:15.680 those on top of that so that we don't have extra number of of categories. 13:15.680 --> 13:22.850 I make sure no problems so we even know when we will talk about feature selection, how we reduce the 13:22.850 --> 13:23.780 number of volumes. 13:23.990 --> 13:29.930 But for now, you just need to know that we have this dummy function, which helps us in converting 13:30.230 --> 13:33.260 the categorical columns into dummy variables. 13:33.530 --> 13:42.770 And for the ordinary columns, we will have another process which will allow us to convert those of 13:42.950 --> 13:46.460 ordinary values into numeric values. 13:47.860 --> 13:52.230 In the next session, we will learn about somebody else. 13:52.350 --> 13:55.180 We will see how we can generate that.