WEBVTT 00:01.320 --> 00:02.280 All right, everyone. 00:02.820 --> 00:10.950 So the next step in this spam message classification project is to separate out your data into two different 00:10.950 --> 00:15.030 buckets that will be training sets of data and texting sets of data. 00:15.570 --> 00:21.240 Now, why we need to separate out this data because this machine learning system is kind of training 00:21.270 --> 00:22.950 plus testing kind of system. 00:23.310 --> 00:28.960 So modern building machine learning happens when you apply some subset of data from your audience. 00:29.220 --> 00:31.380 I say that is called as a training dataset. 00:31.800 --> 00:37.590 And create a model out of it and evaluate your model that how good your model is. 00:38.250 --> 00:40.830 What is the accuracy of running behind your model? 00:41.130 --> 00:48.720 You can apply all those remaining testing datasets on your model to get to know about how good your 00:48.720 --> 00:49.200 model is. 00:49.470 --> 00:55.920 So the basic rule behind is that you just segregate your data into two different buckets, like a training 00:55.920 --> 00:58.860 data set and testing the CSA and know what it was. 00:58.870 --> 01:06.660 So basically to say who your model really never, ever use your testing data set for the model training, 01:06.990 --> 01:11.820 because once you learn the model, the model got created. 01:11.910 --> 01:17.520 We will apply those testing dataset to know about how good or accurate our model is. 01:17.850 --> 01:22.090 So the first task we will do from this data datastream object. 01:22.510 --> 01:24.350 Let me display it. 01:26.060 --> 01:32.820 We will divide it into two different buckets, like a training dataset and testing that a. 01:35.210 --> 01:35.640 All right. 01:36.250 --> 01:40.900 So we have a fourteen hundred ninety four records that I really want. 01:41.350 --> 01:48.970 So let us keep some 30 was in charge of the guy to pasting buckets and 70 percent of the time to cleaning 01:48.970 --> 01:53.020 buckets and thought that we are going to use this sacred land library. 01:53.050 --> 01:57.610 Now, this sacred land is also pre Mendon as a part of this CoLab environment. 01:57.640 --> 02:02.890 So no need to worry about any kind of installation from Skillern. 02:04.430 --> 02:08.780 Let's go with the model selection and let me put. 02:10.970 --> 02:12.890 Thirteen days split. 02:13.840 --> 02:14.320 All right. 02:15.640 --> 02:21.610 So this train is flic function will give us two different buckets. 02:21.730 --> 02:25.900 So the first argument we need to supply for which data you want to do it. 02:26.350 --> 02:29.110 So the first data we want to do it for. 02:30.710 --> 02:38.000 Data message, because that is what our training input data and we are not going to give this length 02:38.570 --> 02:44.090 and punctuation mark column for about model billing, because as we have seen, that is not very much 02:44.120 --> 02:48.140 useful, not creating any kind of differentiating factors. 02:48.590 --> 02:58.220 So what we can do the feature vector first, we can possibly hope it's not only if it's the data and 02:59.570 --> 03:00.420 message column. 03:00.860 --> 03:02.390 So data will be able to input data. 03:02.850 --> 03:07.850 And our objective is to predict the label that is nothing but our output because the data. 03:10.820 --> 03:12.500 In now, one more argument. 03:12.530 --> 03:20.540 You can pass it like a test size or it will be a brain size, so test size. 03:20.930 --> 03:23.810 So here you can view the numbers between zero to one. 03:24.200 --> 03:25.740 Let's see if I give zero point three. 03:25.760 --> 03:29.760 That means you are looking at a total percentage of daytime to testing. 03:29.960 --> 03:30.350 It's. 03:32.000 --> 03:32.990 Let's just make it. 03:33.580 --> 03:34.310 No state. 03:38.290 --> 03:38.660 Zero. 03:39.010 --> 03:44.920 Now, this zero indicates here, whenever you want to recreate exactly the same reason why you are so 03:44.930 --> 03:45.300 insecure. 03:46.090 --> 03:47.770 Same thing on your site. 03:48.400 --> 03:51.280 So zero here indicates it's the same reason. 03:51.310 --> 03:56.800 So suppose if I keep some other numbers, you also have to keep the same number to recreate exactly 03:56.920 --> 03:57.670 the same reason. 03:57.970 --> 04:00.370 And another one is you want to shuffle. 04:00.370 --> 04:03.150 It felt like just make it through. 04:05.450 --> 04:06.820 All right. 04:06.850 --> 04:10.820 Now this whole function will return for retirement. 04:11.290 --> 04:14.070 So that will be x rayed. 04:17.770 --> 04:25.190 Then we have our X, X, Y and escort rain in Reinders could test. 04:25.520 --> 04:32.470 So you have X that refers to a feature input, whereas Y refers to the label as an output. 04:32.860 --> 04:40.020 So corresponding to this particular message, it will be split into two different variables huckstering 04:40.180 --> 04:43.040 exodus and corresponding to this particular variable. 04:43.420 --> 04:45.850 It will be split it into Whiteley and Vytas. 04:46.090 --> 04:49.610 So this is the standard way of writing this for variable. 04:49.690 --> 04:52.620 Generally, everyone in a community use that. 04:53.320 --> 04:54.010 So let me. 04:54.260 --> 04:55.720 Executing an. 04:57.690 --> 05:02.880 So the best way to observe the result is to apply the shape matter on the top of. 05:03.330 --> 05:11.130 So what we can do out of one, I mean, 1494 records that are relevant and it's the total percentage 05:11.130 --> 05:14.170 of value will be 448. 05:14.550 --> 05:21.440 So let's us try to observe how many records containing, say, this train dog --. 05:21.840 --> 05:22.320 Whoops. 05:27.400 --> 05:27.680 It. 05:30.160 --> 05:35.520 And we have a 1045 a court has been look at raining buckets safely. 05:35.930 --> 05:37.330 And let's do this for. 05:38.360 --> 05:41.660 X this like ship. 05:44.900 --> 05:49.240 So it's a 449 cost because obviously you just cannot take slow to lose. 05:49.640 --> 05:54.290 So 449 Lacourt has been assigned to testing it, put it simply. 05:54.590 --> 06:00.110 If you just try to observe how many records are available in a training output limit, our testing output 06:00.110 --> 06:00.440 level. 06:00.470 --> 06:05.130 So that will be respectively 1045 and it will be 449. 06:05.210 --> 06:07.750 So you can verify your sense of light. 06:07.760 --> 06:12.890 Solar database has been created into two different buckets, training and testing. 06:13.190 --> 06:19.330 So in the next video, we'll create our first model and do the training on what training data set. 06:19.820 --> 06:21.030 So see you in the next video. 06:21.310 --> 06:28.280 We'll get started with creating our first random forest model, applying on our spam classification 06:28.280 --> 06:28.610 project.