1
00:00:11,120 --> 00:00:16,180
Everyone, and welcome back to the course, we are now going to introduce the next section.

2
00:00:16,930 --> 00:00:22,820
This section is about our first not naive forecasting method called Etes or Exponential Smoothing.

3
00:00:23,770 --> 00:00:28,240
This lecture will give you an outline for this section and a brief description of what you will learn.

4
00:00:30,130 --> 00:00:33,400
So the basic premise of this section is the moving average.

5
00:00:33,940 --> 00:00:35,990
The moving average is a simple concept.

6
00:00:37,030 --> 00:00:40,540
Imagine taking a window and sliding it across a time series.

7
00:00:40,840 --> 00:00:45,090
And at each point in the Time series, you take the average of all the numbers in that window.

8
00:00:45,460 --> 00:00:47,230
That's called the simple moving average.

9
00:00:51,870 --> 00:00:57,570
So the simple moving average is equally weighted since each point you, including your average, matters

10
00:00:57,570 --> 00:00:58,480
the same amount.

11
00:00:59,130 --> 00:01:04,850
In contrast, we also have the exponentially weighted moving average, which weights each point exponentially

12
00:01:04,890 --> 00:01:06,060
going back in time.

13
00:01:07,080 --> 00:01:11,580
This kind of makes sense because it's saying Newar points matter more than older points.

14
00:01:12,150 --> 00:01:13,650
This would make sense of your time.

15
00:01:13,650 --> 00:01:16,470
Series's is dynamic and changing over time.

16
00:01:21,270 --> 00:01:26,370
From this concept, we're going to build three different models that we can use to make forecasts.

17
00:01:27,480 --> 00:01:32,910
The first model is called simple exponential smoothing, which can be used to forecast non trending,

18
00:01:33,120 --> 00:01:34,750
non seasonal time series.

19
00:01:35,370 --> 00:01:40,320
The second model is called the whole model, which can be used to forecast trending but non seasonal

20
00:01:40,320 --> 00:01:41,220
time series.

21
00:01:42,090 --> 00:01:46,950
The third model is called the whole Winters' Model, which can be used to forecast Time series with

22
00:01:46,950 --> 00:01:48,690
both trends and seasonality.

23
00:01:53,350 --> 00:01:58,660
So throughout this section, we'll look at how to apply these methods to various data sets as well as

24
00:01:58,660 --> 00:02:01,450
some real life applications of these concepts.

25
00:02:03,110 --> 00:02:06,900
Now, I want to note that the theory in this section is essentially optional.

26
00:02:07,430 --> 00:02:10,370
There are basically two kinds of students who will take this course.

27
00:02:10,820 --> 00:02:14,930
One kind will see the equations and realize immediately why they are useful.

28
00:02:15,410 --> 00:02:18,260
It's because all these equations take on similar forms.

29
00:02:18,530 --> 00:02:22,730
So when you see these forms, it's immediately obvious how they are being applied.

30
00:02:23,270 --> 00:02:27,370
This should give you an intuitive understanding of these models and why they work.

31
00:02:28,550 --> 00:02:32,270
The second kind of student will see equations and get kind of intimidated.

32
00:02:32,750 --> 00:02:36,950
So for the second kind of student, I would recommend only watching the lecture.

33
00:02:37,250 --> 00:02:40,670
It's for beginners and then skipping straight to the code.

34
00:02:41,150 --> 00:02:45,620
You won't have an in-depth understanding of each model, but you'll know how to apply it in the real

35
00:02:45,620 --> 00:02:48,800
world without having to be intimidated by math.

36
00:02:49,400 --> 00:02:53,090
So in other words, only look at the math if it actually helps you.

37
00:02:53,420 --> 00:02:56,750
If it doesn't, then feel free to settle for the plug and play approach.