1 00:00:00,720 --> 00:00:04,410 Now, the next model we want to create is a double smarted. 2 00:00:05,430 --> 00:00:11,030 So the steps are almost seem reinforced import at Abus classifier. 3 00:00:12,240 --> 00:00:20,180 And then we can provide hyper parameters and I was classifier and created our redoubles classifier object. 4 00:00:20,640 --> 00:00:22,800 We will use our Ekstrand in y train data. 5 00:00:23,880 --> 00:00:29,330 To train this model and then we'll find no accuracy score on our test data. 6 00:00:32,890 --> 00:00:35,940 Let's look at the documentation of that most classified. 7 00:00:40,180 --> 00:00:42,820 Here, our first parameter is base estimate. 8 00:00:44,350 --> 00:00:49,300 Unlike other techniques here, you can provide your base estimate there as well. 9 00:00:50,380 --> 00:00:52,060 By default, that's the decision. 10 00:00:52,060 --> 00:00:54,190 Greek classifier of maximum deafen. 11 00:00:55,760 --> 00:01:03,680 But you can also provide random forest or any other classified as a base, a scimitar for doubles and 12 00:01:03,920 --> 00:01:08,300 doubles will perform, boosting technique on that classifier. 13 00:01:09,960 --> 00:01:12,770 Now, the next parameter here is an estimated. 14 00:01:12,910 --> 00:01:15,030 Again, this is the number of trees we want. 15 00:01:16,690 --> 00:01:21,960 Then we have learning rate this, the learning rate for our Addabbo Smardon. 16 00:01:22,210 --> 00:01:24,700 We have also to discuss, in a word, to re lecture. 17 00:01:28,220 --> 00:01:34,700 So for our first mardon, we will use a learning rate of zero point zero two and the number of crease 18 00:01:34,730 --> 00:01:35,960 to be five tosing. 19 00:01:39,190 --> 00:01:40,360 Let's sit on this, Martin. 20 00:01:43,380 --> 00:01:45,440 We have created a world object. 21 00:01:45,680 --> 00:01:47,750 We will photoed explain why drin due to. 22 00:01:53,440 --> 00:01:54,790 No more delays already. 23 00:01:55,390 --> 00:02:00,250 So by default, we have created this model on their decision, Tree of Depth one. 24 00:02:01,450 --> 00:02:07,660 You can see the by default value of our base basis, someone that is mission creep classifier maximum 25 00:02:07,670 --> 00:02:07,890 depth. 26 00:02:10,330 --> 00:02:18,100 So we have frittered our air, the boost on that classifier, and we can calculate the accuracy score 27 00:02:18,100 --> 00:02:21,640 by just running accuracy score and using the predicate function. 28 00:02:22,870 --> 00:02:25,390 So the accuracy here is 62 percent. 29 00:02:28,230 --> 00:02:28,700 No. 30 00:02:30,210 --> 00:02:33,200 Let's create another classifier object. 31 00:02:33,390 --> 00:02:37,490 Here we are using a word random forest classifier. 32 00:02:37,770 --> 00:02:41,070 We have already created this classifier about. 33 00:02:44,710 --> 00:02:45,550 This is Overclassify. 34 00:02:46,190 --> 00:02:52,700 We are using this classifier with our doubles classifier here. 35 00:02:53,450 --> 00:02:56,900 I have reduced the number of Greece to 500. 36 00:02:57,740 --> 00:02:59,440 And I will also change. 37 00:02:59,680 --> 00:03:02,030 Learning that put zero point five since I have. 38 00:03:03,020 --> 00:03:05,300 Decreased my number of grease. 39 00:03:06,860 --> 00:03:07,850 Let's run this. 40 00:03:09,910 --> 00:03:13,090 Green, there were more done on this new object. 41 00:03:15,080 --> 00:03:18,600 Now, let's get the accuracy score for this morning. 42 00:03:18,670 --> 00:03:23,060 This way again, the new accuracy score is 64 percent. 43 00:03:23,900 --> 00:03:27,500 So you can see that here we were using just this January. 44 00:03:27,830 --> 00:03:35,600 Here we are using random forest as our base classified and that have increased the accuracy score from 45 00:03:35,600 --> 00:03:37,690 62 percent to 64 percent. 46 00:03:39,160 --> 00:03:47,260 Again, you can use grid search to optimize this learning rate and estimates, and you can also provide 47 00:03:47,560 --> 00:03:50,980 different models as your basis simulator in the grid search. 48 00:03:51,910 --> 00:03:57,760 So try a different combination and try to find the model which will give you accuracy scale more than 49 00:03:57,760 --> 00:03:58,390 this is school. 50 00:04:00,030 --> 00:04:00,450 Thank you.