1 00:00:01,330 --> 00:00:02,230 Hi, welcome back. 2 00:00:02,260 --> 00:00:09,010 In this video, we are going to generate a plot containing different lines and each line is going to 3 00:00:09,010 --> 00:00:14,030 represent the average rating by month for a particular course. 4 00:00:14,260 --> 00:00:15,980 So we have multiple courses. 5 00:00:16,000 --> 00:00:18,160 We are going to have one line for each course. 6 00:00:18,800 --> 00:00:19,540 Let's do that. 7 00:00:20,020 --> 00:00:25,280 Let's put a heading h3. 8 00:00:25,540 --> 00:00:31,150 This would be average rating by month by course. 9 00:00:32,630 --> 00:00:32,990 Right. 10 00:00:33,950 --> 00:00:35,400 This is how we do that. 11 00:00:35,660 --> 00:00:36,920 First of all, we need 12 00:00:39,590 --> 00:00:43,970 a month column in the data data frame. 13 00:00:44,150 --> 00:00:46,150 Now, I know that we did that here. 14 00:00:46,160 --> 00:00:51,920 We already have that month column, but I want to keep the code separate so it doesn't harm if you do 15 00:00:51,920 --> 00:00:53,860 it again here just for consistency. 16 00:00:54,140 --> 00:00:58,810 So to see that this cell is doing that particular thing. 17 00:00:59,000 --> 00:01:01,640 So it's building a graph of multiple lines. 18 00:01:01,820 --> 00:01:04,080 And for that graph, we need to this expression. 19 00:01:04,080 --> 00:01:05,630 So it's the same expression. 20 00:01:05,810 --> 00:01:07,940 I can just copy and paste it in here. 21 00:01:09,140 --> 00:01:09,490 Right. 22 00:01:09,500 --> 00:01:11,210 And we don't need to change anything there. 23 00:01:11,210 --> 00:01:13,190 So we just need to extract the month 24 00:01:15,800 --> 00:01:17,290 and let's check. 25 00:01:17,960 --> 00:01:20,480 So we have the month for each row. 26 00:01:21,350 --> 00:01:24,500 Since we are grouping by months, it makes sense to have that. 27 00:01:25,730 --> 00:01:31,730 Then we need to use the groupby method to create a new data frame, which would be an aggregation. 28 00:01:32,090 --> 00:01:42,260 Now, just so that we compare it to compare it to let's see what the previous month average data frame 29 00:01:42,260 --> 00:01:43,160 looks like. 30 00:01:43,700 --> 00:01:47,300 So again, that is what we had. 31 00:01:47,990 --> 00:01:52,180 This data frame is missing the information about the courses. 32 00:01:52,400 --> 00:01:54,860 So it's missing the course name information. 33 00:01:55,130 --> 00:02:02,570 You see here we have this column that we need to have that dimension somehow in our new aggregated data 34 00:02:02,570 --> 00:02:02,930 frame. 35 00:02:03,620 --> 00:02:06,370 So what we need to do here is something different. 36 00:02:07,040 --> 00:02:11,090 Again, we need to create a new variable. 37 00:02:11,270 --> 00:02:16,820 Let's call this a month_average_crs maybe for course. 38 00:02:17,030 --> 00:02:25,040 So to distinguish between that data frame and that month_average, month_average_crs. 39 00:02:25,040 --> 00:02:25,820 This again will be 40 00:02:27,750 --> 00:02:31,860 a product, a byproduct of that data frame. 41 00:02:32,490 --> 00:02:39,810 So using the groupby methods, but in this case, let me delete that so you can see the other codes here 42 00:02:40,080 --> 00:02:42,800 to see the difference. In this case, 43 00:02:44,400 --> 00:02:52,830 we are going to have not only the month here in the list of columns to be used for the aggregation process, 44 00:02:53,280 --> 00:02:57,480 but we are also going to have can you guess what? Course name. 45 00:02:58,530 --> 00:03:02,330 So by month, by course. That's what we need. 46 00:03:03,600 --> 00:03:07,950 Let's apply now mean() and and see what we get so far. 47 00:03:14,190 --> 00:03:24,180 So this is a data frame and basically this one has actually two levels of indexes and it has the month 48 00:03:24,180 --> 00:03:28,830 and the course name, you can see that if you do that index. 49 00:03:30,270 --> 00:03:31,980 So you see it's a multi index. 50 00:03:32,820 --> 00:03:34,620 You see the month column. 51 00:03:35,910 --> 00:03:42,420 And also the course name, which doesn't show here, but it's there and as you saw it, you could 52 00:03:42,420 --> 00:03:43,620 see it in here. 53 00:03:43,910 --> 00:03:45,960 You see that these are in bold. 54 00:03:47,640 --> 00:03:49,990 These values, that that means they are indexes. 55 00:03:51,120 --> 00:03:54,450 This here is a column, so the columns are 56 00:03:57,540 --> 00:03:59,820 rating only, only one column. 57 00:04:01,080 --> 00:04:10,620 Now, this is not very useful yet because of the way this is constructed is that we have a row here, 58 00:04:10,950 --> 00:04:13,130 so this is a group of rows. 59 00:04:13,620 --> 00:04:14,720 It starts here. 60 00:04:14,790 --> 00:04:19,140 So the first course, is the second course, the third, fourth, the fifth. 61 00:04:19,500 --> 00:04:25,980 And there's also some other courses here, which actually we can see them by applying here as lines. 62 00:04:26,170 --> 00:04:28,770 So let's see the first 20 records. 63 00:04:31,560 --> 00:04:41,730 Now we see the first 20 records in full, so complete first day, sorry first month of 2018, second 64 00:04:41,730 --> 00:04:51,510 month of 2018 and so on, so that course has that rating for that month, that course has that average 65 00:04:51,510 --> 00:04:53,360 rating for that month. 66 00:04:53,880 --> 00:05:01,800 And so on, all the other courses. Then the same pattern such as this one here is repeated over and over 67 00:05:01,800 --> 00:05:10,070 again for all the months. Now to have this data are in a better structure, 68 00:05:10,500 --> 00:05:20,940 what we would do is we would apply here and dot unstack method to basically unstack this data frame 69 00:05:21,510 --> 00:05:24,450 and end up with this better structure. 70 00:05:25,410 --> 00:05:33,180 So now we have sort of a pivot table, so to say. We have the month here, first month, second month, 71 00:05:33,180 --> 00:05:37,710 third month, and each column now is representing a course. 72 00:05:41,020 --> 00:05:47,920 So if you want to know, for example, the rating of 'The Complete Python Course: Build 10 Profesional 73 00:05:47,920 --> 00:05:49,390 OOP Apps for 74 00:05:50,420 --> 00:05:58,700 a particular month, let's say this one, it's NaN because the cause was not published yet in that date, 75 00:05:58,940 --> 00:06:04,730 but if you look at the end, let's say minus 20, 76 00:06:11,440 --> 00:06:19,450 then you see that from the first month of 2020, we have a rating for that course, an average 77 00:06:19,450 --> 00:06:20,760 rating, right? 78 00:06:24,550 --> 00:06:27,860 And so on, what can we do this data frame now? 79 00:06:28,570 --> 00:06:36,280 Well, we can plot it, but this time we are going to use a different approach because we cannot use 80 00:06:36,280 --> 00:06:39,310 that plt.plot. 81 00:06:41,330 --> 00:06:49,100 Since we have multiple, multiple columns, so this expects an X and a Y, but which one is the X 82 00:06:49,640 --> 00:06:49,890 right? 83 00:06:49,920 --> 00:06:54,020 The X of course is month, we are clear about that. 84 00:06:54,890 --> 00:06:59,830 But which one is the Y? Is it that or that or that? 85 00:07:00,080 --> 00:07:02,270 So we have multiple columns. 86 00:07:02,450 --> 00:07:09,860 One way would be to write this plot function multiple times or maybe create a loop that iterate through 87 00:07:09,860 --> 00:07:10,550 the data frame. 88 00:07:10,550 --> 00:07:18,620 But another easier way is to just point to average_crs.plot. 89 00:07:20,380 --> 00:07:28,740 And voila, we have of course, it looks a mess, but let's improve it a bit since we used the plot 90 00:07:28,750 --> 00:07:38,110 function directly from the data frame, not from the plt, then we can use a figsize argument in here. 91 00:07:38,410 --> 00:07:40,720 So let's say 25, 3. 92 00:07:42,360 --> 00:07:44,280 And now it looks a bit better. 93 00:07:46,220 --> 00:07:47,550 Still not ideal. 94 00:07:48,050 --> 00:07:50,630 Maybe we could increase that to eight. 95 00:07:52,740 --> 00:08:00,270 OK, now it's working so that we have a legend that shows the color of each of the causes. 96 00:08:01,510 --> 00:08:08,220 Again, I'm not a fan of this kind of graph, but when we use the Highcharts library with the Web interface, 97 00:08:08,320 --> 00:08:10,390 this is going to look much better. 98 00:08:10,840 --> 00:08:13,060 Now, what if we use count here? 99 00:08:17,460 --> 00:08:25,560 We would get the graph correctly, but the legend would be a bit of a mess, that is because in the 100 00:08:28,830 --> 00:08:35,130 month_average_crs dataframe 101 00:08:40,480 --> 00:08:49,540 we have not only the counts of the ratings, but we also have other kinds of timestamps and things 102 00:08:49,540 --> 00:08:50,460 we don't need. 103 00:08:50,770 --> 00:08:54,360 You see that, for example '100 Exercises I' 104 00:08:54,370 --> 00:08:57,400 has the time stamp here. 105 00:08:57,400 --> 00:09:00,330 You see it's on another level of columns. 106 00:09:00,610 --> 00:09:04,570 So all these belong to timestamp, right? 107 00:09:05,050 --> 00:09:07,380 The count of time stamp. 108 00:09:08,890 --> 00:09:13,240 And then we have the count of rating starting from here and up to somewhere. 109 00:09:13,360 --> 00:09:17,500 We don't see all the columns because Jupiter is truncating them. 110 00:09:18,310 --> 00:09:20,840 But you get the idea day and week. 111 00:09:21,640 --> 00:09:24,280 So how do we extract only the rating? 112 00:09:24,430 --> 00:09:25,950 Well, that is easy, actually. 113 00:09:26,830 --> 00:09:39,700 I told you that groupby which starts in here and ends in here, groupby returns a data frame, now out 114 00:09:39,700 --> 00:09:40,660 of a data frame 115 00:09:40,660 --> 00:09:43,300 we extract only the rating. 116 00:09:46,860 --> 00:09:56,130 And problem solved now we get a clear graph showing the number of ratings left for each month of the 117 00:09:56,130 --> 00:09:56,460 year. 118 00:09:58,200 --> 00:10:05,760 So starting from the very beginning, 2018, up to 2020. You see this pink line here, which is the 119 00:10:05,760 --> 00:10:14,690 new course, 'The Complete Python Course', it starts somewhere in 2021, so the first month of 2021. 120 00:10:15,360 --> 00:10:22,770 And of course, you can again get the mean, just like you did previously, but with mean we again get 121 00:10:22,770 --> 00:10:30,500 the same results since the mean method ignores those timestamps of those nonnumeric columns. 122 00:10:30,720 --> 00:10:35,850 So it's any way it drops them, but count does not drop them. 123 00:10:35,850 --> 00:10:39,680 It's counts them up and gives us so much data. 124 00:10:40,380 --> 00:10:40,700 Right. 125 00:10:40,800 --> 00:10:41,820 I hope this was clear. 126 00:10:41,820 --> 00:10:43,300 And I'll talk to you later. 127 00:10:43,350 --> 00:10:43,710 See you.