1 00:00:00,520 --> 00:00:04,330 What day of the week are people the happiest? 2 00:00:04,780 --> 00:00:10,780 Can we answer that question by using Python and the data we have at hand right now? 3 00:00:11,320 --> 00:00:13,030 Well, I think, yes, we can do that. 4 00:00:13,240 --> 00:00:15,810 And that is the beauty of data analysis. 5 00:00:16,100 --> 00:00:20,900 If you are a bit creative, you can answer very interesting questions about your data. 6 00:00:21,430 --> 00:00:27,640 So in this video, we are going to find out together which day of the week are people the happiest? 7 00:00:28,000 --> 00:00:29,420 How do we do that? 8 00:00:29,440 --> 00:00:31,890 Well, by generating a graph. 9 00:00:32,320 --> 00:00:34,010 So how can we do that? 10 00:00:34,330 --> 00:00:41,500 Well, in my opinion, we can use the logic that if the average rating for all the courses, let's say, 11 00:00:41,500 --> 00:00:48,250 on Wednesdays is of the highest of the week, then we can see that people maybe in that day 12 00:00:48,560 --> 00:00:51,610 so on Wednesdays are more positive. 13 00:00:51,790 --> 00:00:55,690 And I'm sure we can tell that because we have so much data here. 14 00:00:55,840 --> 00:00:58,090 So thousands of ratings. 15 00:00:58,090 --> 00:01:03,610 And statistically, that I think gives us a lot of confidence in the outcome. 16 00:01:04,000 --> 00:01:11,110 So what we need to do is we need to create a graph which is going to have seven days in the horizontal 17 00:01:11,110 --> 00:01:11,540 axis. 18 00:01:11,740 --> 00:01:19,510 So from Monday to Sunday and in the vertical axis, we are going to have the average for each day. 19 00:01:20,470 --> 00:01:24,160 That means we have to do a lot of data aggregation. 20 00:01:25,000 --> 00:01:25,870 So let's do that. 21 00:01:29,770 --> 00:01:34,210 What day are people that are happiest? 22 00:01:34,510 --> 00:01:43,690 Now, we are going to use, as always, the data data frame on so where we have these ratings left at a particular 23 00:01:43,690 --> 00:01:44,310 time stamps. 24 00:01:44,920 --> 00:01:50,530 So what do we need to obstruct where the weekday, let's call it like that. 25 00:01:52,690 --> 00:02:00,250 Which is equal to data, timestamp, so out of timestamp, we access the dt property and out of that 26 00:02:00,250 --> 00:02:06,000 we need strftime and the day of the week. 27 00:02:06,070 --> 00:02:13,510 So, like, if you want Sunday or Monday or Tuesday, that name of the day of the week can be extracted 28 00:02:13,510 --> 00:02:18,010 using the capita A format code. 29 00:02:18,370 --> 00:02:19,730 You can look that up on Google. 30 00:02:19,750 --> 00:02:20,720 I do that as well. 31 00:02:20,740 --> 00:02:22,990 It's impossible to to remember everything. 32 00:02:23,540 --> 00:02:29,980 Now, we can print out the data again and we see that we have a weekday column. 33 00:02:30,430 --> 00:02:33,190 So Friday, Friday, Friday is repeating over and over again. 34 00:02:33,400 --> 00:02:37,740 And we need to aggregate now to get the average for each weekday. 35 00:02:38,470 --> 00:02:47,970 So let's say a weekday_average for the data frame that is going to contain the aggregated data. 36 00:02:48,370 --> 00:02:51,420 Again, groupby is our savior. 37 00:02:51,930 --> 00:02:52,290 Hmm. 38 00:02:52,900 --> 00:02:54,480 What do we use this time? 39 00:02:55,120 --> 00:02:56,140 Well, weekday. 40 00:02:57,860 --> 00:03:09,980 And then do mean and let's see weekday average what we get and that's it, then all we have to do is 41 00:03:10,640 --> 00:03:11,090 leave that. 42 00:03:11,720 --> 00:03:21,770 All we have to do is plt.plot the x axis would be weekday_average 43 00:03:22,190 --> 00:03:29,630 dot index and the Y axis would be weakday_average, weekday, sorry rating. 44 00:03:31,760 --> 00:03:38,720 Plot and this is the answer, I know the order is not as you would expect, but we can fix that. 45 00:03:39,840 --> 00:03:48,300 Either way, we can tell that Friday is quite different from the other days, so on Fridays, people 46 00:03:48,300 --> 00:03:58,260 leave a four point four, five five average rating on a scale of one to five, and that makes sense. 47 00:03:58,560 --> 00:04:02,970 I think I also fall in that range, so I tend to be happier on Fridays. 48 00:04:03,780 --> 00:04:07,740 So end of the week and the weekend is coming, so it makes sense. 49 00:04:08,310 --> 00:04:14,710 Now, how can we order these data, so we need to order the weekday column. 50 00:04:15,750 --> 00:04:17,130 Well, we can do that. 51 00:04:17,520 --> 00:04:20,760 Let me comment this out for a while. 52 00:04:21,510 --> 00:04:29,940 Weekday_average equal to weekday_average.sort_values. 53 00:04:31,410 --> 00:04:44,940 By weekday and then print out and see what weekday_average is, and you're going to see that this is not 54 00:04:45,180 --> 00:04:52,470 the order we would expect. That is happening, of course, because the index column, this one here, 55 00:04:53,850 --> 00:04:55,470 has string values. 56 00:04:55,500 --> 00:04:58,800 So Friday, Mondays are they are all strings. 57 00:04:59,460 --> 00:05:05,050 Therefore, Python is ordering them in alphabetical order. 58 00:05:05,220 --> 00:05:07,230 So Friday, first Monday and so on. 59 00:05:07,770 --> 00:05:12,060 They are strings because this method produces a string 60 00:05:12,060 --> 00:05:16,800 weekday column, you see, string from time. 61 00:05:17,460 --> 00:05:21,150 So timestamp was time, but we converted to a string. 62 00:05:21,660 --> 00:05:25,080 Now there may be different workarounds to fix this. 63 00:05:25,560 --> 00:05:31,110 The one that comes to my mind is that we could add a weekday 64 00:05:32,490 --> 00:05:33,840 number. 65 00:05:35,520 --> 00:05:38,280 Or simply daynumber. 66 00:05:43,220 --> 00:05:51,200 To the data data frame and all of that again uses string from time, but this time we are going to 67 00:05:51,200 --> 00:05:59,190 use percentage lowercase w and what that would give us is. 68 00:05:59,240 --> 00:06:07,040 Let me show you the data now is this data frame and it says daynumber. 69 00:06:07,280 --> 00:06:16,430 And so now we have a number attached to each row besides the name of the weekday that allows us now 70 00:06:16,430 --> 00:06:20,420 to sort the values by this day number column. 71 00:06:21,020 --> 00:06:22,550 So Monday is the first one. 72 00:06:23,910 --> 00:06:25,680 Friday is the fifth and so on. 73 00:06:27,750 --> 00:06:28,410 Let's do that. 74 00:06:28,950 --> 00:06:36,600 I'm going to delete that data data frame from there, the printing uncomment this and change that to 75 00:06:36,750 --> 00:06:38,250 daynumber. 76 00:06:41,200 --> 00:06:42,580 And here weekday 77 00:06:42,790 --> 00:06:50,200 average, delete, uncomment this, execute, we have this Key Error day 78 00:06:50,210 --> 00:06:50,830 number. 79 00:06:51,810 --> 00:06:55,320 Oh, because I forgot to add daynumber in here. 80 00:06:59,050 --> 00:07:00,100 Again, we have an error. 81 00:07:00,130 --> 00:07:02,400 Let's see where the error occurred. 82 00:07:06,190 --> 00:07:16,180 So these are the errors deep in the libraries, you see in matplotlib and these shows the error in our 83 00:07:16,840 --> 00:07:17,340 code. 84 00:07:18,070 --> 00:07:20,560 So it's in this line, plt.plot. 85 00:07:24,480 --> 00:07:29,250 And it says that value must be an instance of string or bytes, not a tuple. 86 00:07:30,450 --> 00:07:33,380 So let's try to troubleshoot the error here. 87 00:07:33,640 --> 00:07:39,510 I'm going to uncomment this and see what's weekday average is. 88 00:07:41,910 --> 00:07:51,630 So the problem here is that weekday_average.index returns a double index, which looks like 89 00:07:52,170 --> 00:07:52,770 this. 90 00:07:55,410 --> 00:07:59,330 So it's a multi index, it contains weekday and daynumber. 91 00:08:00,030 --> 00:08:10,470 Therefore, the plot method will be confused so it doesn't know which of these arrays has to plot along 92 00:08:10,470 --> 00:08:11,380 the x axis. 93 00:08:12,690 --> 00:08:17,130 So we need a way to extract only these columns. 94 00:08:17,130 --> 00:08:26,370 So the names of the weekdays, you can do that by using the get level values. 95 00:08:26,820 --> 00:08:29,220 That is a method that expects an argument. 96 00:08:29,430 --> 00:08:35,910 In this case, we can either input zero or one. Zero means we are extracting, we are accessing the first 97 00:08:35,910 --> 00:08:40,280 column of the multi index and one would be the second column. 98 00:08:40,530 --> 00:08:41,850 So we need the first column. 99 00:08:43,790 --> 00:08:54,800 And that gives us those days, so the names of the days. Therefore now we can make use of that in here 100 00:08:54,800 --> 00:08:58,580 so we can copy that expression. 101 00:08:58,730 --> 00:09:03,050 That is the array we want to plot along the X axis. 102 00:09:03,300 --> 00:09:06,310 So we put that in here, execute. 103 00:09:06,350 --> 00:09:08,000 And this is the output. 104 00:09:08,930 --> 00:09:20,140 If we want to make the picture wider, we need to do plt.figure figsize equal to a list of 25 ,3. 105 00:09:20,150 --> 00:09:22,970 It works well or maybe a bit. 106 00:09:25,650 --> 00:09:26,010 Smaller. 107 00:09:27,030 --> 00:09:28,060 Yeah, this is better. 108 00:09:28,260 --> 00:09:32,670 So you have to to find a good ratio between the width and the height. 109 00:09:33,120 --> 00:09:39,500 And again, we see that Friday is the day that has the highest average rating. 110 00:09:39,840 --> 00:09:41,670 And so that was about this video. 111 00:09:42,150 --> 00:09:43,170 Thanks a lot for following! 112 00:09:43,170 --> 00:09:45,420 I'll talk to you in the next videos! See you!