1 00:00:00,240 --> 00:00:01,020 Welcome back. 2 00:00:01,290 --> 00:00:07,940 It's the last video we covered are to score and so this one we're gonna look at mean absolute era. 3 00:00:08,200 --> 00:00:09,240 Let's come down here. 4 00:00:09,450 --> 00:00:13,710 We'll create a little heading mean absolute error. 5 00:00:13,900 --> 00:00:14,990 Okay I am. 6 00:00:15,020 --> 00:00:16,600 Ay ay. 7 00:00:16,620 --> 00:00:22,820 Now again we could always look up what mean absolute error is but what we're going to do is check in 8 00:00:22,830 --> 00:00:25,410 with the code first and then we'll look into it. 9 00:00:25,410 --> 00:00:26,460 How do we calculate it. 10 00:00:26,460 --> 00:00:32,990 I mean absolute error you might be able to deduce from its name what it actually does if not. 11 00:00:32,990 --> 00:00:33,460 That's right. 12 00:00:33,460 --> 00:00:34,210 We'll look into it. 13 00:00:34,710 --> 00:00:38,220 Import mean absolute needed square. 14 00:00:38,220 --> 00:00:44,010 That's gonna be the next one to evaluate it we need to make some predictions because remember a valuation 15 00:00:44,010 --> 00:00:48,210 metrics are always comparing truth labels to predicted labels. 16 00:00:48,200 --> 00:00:54,950 So we'll predict on the test data set and go and they E equals mean absolute error. 17 00:00:55,170 --> 00:01:00,120 And we're going to compare y test which is the truth labels that we made up here. 18 00:01:00,120 --> 00:01:02,730 Why test to our predictions. 19 00:01:02,730 --> 00:01:08,400 Which is why we've just made here and then going to have a look at M.J. OK. 20 00:01:08,560 --> 00:01:12,000 That wasn't too hard but two point one too. 21 00:01:12,100 --> 00:01:13,360 What does that mean. 22 00:01:13,360 --> 00:01:21,120 Well again this might be a little bit easier to visualize so we might create a data frame side by side 23 00:01:21,130 --> 00:01:30,370 we're going to posit data equals a dictionary and we're gonna have actual values vs. what we want here 24 00:01:31,300 --> 00:01:44,350 is predicted values go y parades wonderful and then we'll have a look at f huh. 25 00:01:44,380 --> 00:01:45,670 So what do we have here. 26 00:01:46,150 --> 00:01:50,360 Well this is the truth labels and this is what our model has predicted. 27 00:01:50,860 --> 00:01:54,880 So can see twenty three point six is twenty three point 0 0 2. 28 00:01:55,020 --> 00:01:55,590 Okay. 29 00:01:55,630 --> 00:01:59,350 Thirty two point four thirty point nine point two. 30 00:01:59,350 --> 00:02:04,000 Looking at these right and then looking at this number can you think about where this number might be 31 00:02:04,000 --> 00:02:06,710 coming from. 32 00:02:06,810 --> 00:02:07,200 That's right. 33 00:02:07,200 --> 00:02:13,660 He can't what we might do is add another column that may add a little bit of clarity equals and we're 34 00:02:13,660 --> 00:02:14,940 going to do. 35 00:02:15,250 --> 00:02:19,370 Def predicted value. 36 00:02:19,390 --> 00:02:22,030 So this is this is a column we're creating for the differences right. 37 00:02:22,030 --> 00:02:27,390 So the difference between our predicted value and the actual values. 38 00:02:27,730 --> 00:02:29,340 There we go. 39 00:02:29,390 --> 00:02:30,690 We'll be IDF again. 40 00:02:30,880 --> 00:02:32,310 Differences. 41 00:02:32,340 --> 00:02:32,720 OK. 42 00:02:32,750 --> 00:02:36,300 So we see that this one is minus point six or. 43 00:02:36,520 --> 00:02:42,670 This one is minus point five and this one is three point ones was a bit off here and this one was really 44 00:02:42,670 --> 00:02:43,060 off right. 45 00:02:43,060 --> 00:02:45,250 It's minus four point eighty seven. 46 00:02:45,310 --> 00:02:45,820 Okay. 47 00:02:46,750 --> 00:02:52,570 So now you might be starting to get an idea of where mean absolute error comes from if not mean absolute 48 00:02:52,570 --> 00:02:58,870 error it's formally defined as the average of the absolute differences between predictions and actual 49 00:02:58,870 --> 00:03:02,860 values and so it gives you an idea of how wrong your predictions are. 50 00:03:02,860 --> 00:03:05,410 So what we might do is write that here. 51 00:03:05,620 --> 00:03:19,090 So amazing is the average of the absolute differences between predictions and actual values. 52 00:03:19,090 --> 00:03:29,260 It gives you an idea of how wrong your models predictions are and what it means by absolute is that 53 00:03:29,260 --> 00:03:35,110 it would turn these values positive and then add them all up and then take the main and then that's 54 00:03:35,110 --> 00:03:36,510 going to give us this number. 55 00:03:36,520 --> 00:03:43,750 So basically what this is saying is that on average every prediction our model makes is going to be 56 00:03:43,750 --> 00:03:50,190 on average plus or minus two point one two away from the actual value. 57 00:03:50,290 --> 00:03:51,910 And you can kind of see that right. 58 00:03:51,930 --> 00:03:57,640 Some are only point six six seven away but others are more than two point one. 59 00:03:57,640 --> 00:04:03,910 So this is 2.8 9 and this is negative 3 point 7 9 4 and you can keep going throughout all the 102 rows 60 00:04:03,910 --> 00:04:09,820 and figure out this figure manually or you could just take the psychic learn implementations word for 61 00:04:09,820 --> 00:04:09,910 it. 62 00:04:10,600 --> 00:04:14,190 So what we're gonna do next that's covered mean absolute error. 63 00:04:14,260 --> 00:04:16,240 We're going to have a look at Main squared error.