1
00:00:01,110 --> 00:00:04,410
The first step in the process is Gorie definition.

2
00:00:05,730 --> 00:00:13,980
This basically means that we should be clear about our objective data mining and clearly defining the

3
00:00:13,980 --> 00:00:18,420
forecasting goal is essential for arriving at useful results.

4
00:00:20,430 --> 00:00:25,950
Here is a list of some of the questions that you must answer before starting the analysis.

5
00:00:27,600 --> 00:00:31,320
You must first determine the purpose of generating forecasts.

6
00:00:32,340 --> 00:00:34,110
The type of forecasts that are needed.

7
00:00:35,040 --> 00:00:37,540
How the forecasts will be used by the organization.

8
00:00:39,360 --> 00:00:43,110
What are the costs associated with forecast errors?

9
00:00:44,160 --> 00:00:50,110
What data will be available in the future, etc. in the next few slides?

10
00:00:50,880 --> 00:00:55,440
I'm covering some of the important considerations during the goal setting step.

11
00:00:57,270 --> 00:01:05,100
Pay attention because these issues will affect every step in the forecasting process from data collection

12
00:01:05,520 --> 00:01:09,180
through data processing, modeling and performance evaluation.

13
00:01:12,330 --> 00:01:17,880
First comes whether our goal is descriptive analysis or predictive analysis.

14
00:01:19,980 --> 00:01:27,750
Time series data modeling, as done for either descriptive or predictive purposes in descriptive modeling

15
00:01:27,960 --> 00:01:35,070
or time studies, is modelled to determine its competence in terms of seasonal patterns or trends or

16
00:01:35,070 --> 00:01:37,830
relation to external factors and the like.

17
00:01:39,660 --> 00:01:43,380
These can then be used for decision making and policy formulation.

18
00:01:44,700 --> 00:01:51,780
For example, if you want to find out the effect of rainy season on travel bookings, that is descriptive

19
00:01:51,780 --> 00:01:52,380
analysis.

20
00:01:54,680 --> 00:02:01,760
In contrast, predictive analysis uses the information in a time cities to forecast future values of

21
00:02:01,790 --> 00:02:02,310
parties.

22
00:02:04,190 --> 00:02:09,050
For example, if you want to find out the number of travel bookings that will be done in the coming

23
00:02:09,050 --> 00:02:12,710
three months, that is predictive analysis.

24
00:02:14,480 --> 00:02:20,750
The difference between descriptive and predictive goals leads to the differences in the types of methods

25
00:02:20,750 --> 00:02:23,040
used and in the modeling process itself.

26
00:02:24,740 --> 00:02:32,690
For example, in selecting a method for describing a time cities or even for explaining its patterns,

27
00:02:33,640 --> 00:02:37,600
priority is given to methods that produce explainable results.

28
00:02:38,510 --> 00:02:47,300
And two models which are based on causal arguments, on the other hand, are predictive model is judged

29
00:02:47,300 --> 00:02:53,190
by its predictive accuracy rather than its ability to provide correct explanation.

30
00:02:55,810 --> 00:03:04,380
In short, for descriptive analysis, those models are preferred, which are more explainable, and

31
00:03:04,390 --> 00:03:11,500
for predictive analysis, we prefer those models which give higher accuracy, even if they are complex

32
00:03:11,680 --> 00:03:13,390
and not easily explainable.

33
00:03:14,320 --> 00:03:15,790
Let's consider an example.

34
00:03:15,940 --> 00:03:24,670
To understand this more clearly, Amtrak is a U.S. railway company which routinely collects data on

35
00:03:24,760 --> 00:03:25,360
ridership.

36
00:03:26,890 --> 00:03:34,930
In this example, we will look at these series of monthly Amtrak ridership between January 1991 and

37
00:03:34,930 --> 00:03:37,900
March 2004 in the United States.

38
00:03:39,460 --> 00:03:46,960
This dataset is part of the course resources, not what could be the different analysis goals for a

39
00:03:46,960 --> 00:03:47,770
railway company.

40
00:03:48,280 --> 00:03:48,820
Let's see.

41
00:03:51,630 --> 00:03:59,880
One possible analysis goal that Amtrak might have is to forecast future monthly ridership on its trains

42
00:04:00,510 --> 00:04:01,950
for the purpose of pricing.

43
00:04:03,900 --> 00:04:06,240
Clearly, this is a predictable.

44
00:04:08,180 --> 00:04:13,550
Setting the prices as per demand is a popular practice by airlines and hotel chains.

45
00:04:14,450 --> 00:04:21,470
You may have noticed hotel and flight prices are higher in a peak travel season as compared to other

46
00:04:21,470 --> 00:04:21,860
season.

47
00:04:23,930 --> 00:04:26,390
This strategy of pricing is called revenue management.

48
00:04:28,560 --> 00:04:37,460
Anyway, predicting future ridership is a predictable or different goal for which Amtrak might want

49
00:04:37,460 --> 00:04:45,530
to use these ridership data is for the impact assessment or for evaluating the effect of some, even,

50
00:04:47,120 --> 00:04:54,410
for example, airport closure due to bad weather or opening of a new large national highway.

51
00:04:56,330 --> 00:04:58,530
This goal is retrospective in nature.

52
00:04:59,510 --> 00:05:07,310
That is airport was closed probably last month and we are trying to establish the impact of that event

53
00:05:07,820 --> 00:05:08,860
on rail ridership.

54
00:05:08,900 --> 00:05:16,480
Now, this goal is therefore descriptive as it will describe the impact of that event.

55
00:05:18,560 --> 00:05:25,130
This analysis would compare the series before and after the event, and no direct interest is shown

56
00:05:25,250 --> 00:05:26,960
in the future values of the series.

57
00:05:28,730 --> 00:05:34,700
With these two examples, I hope that you can clearly differentiate between descriptive and predictive,

58
00:05:34,700 --> 00:05:43,700
good predictive analysis is also called Time series forecasting, whereas descriptive analysis is called

59
00:05:43,970 --> 00:05:45,230
times in his analysis.

60
00:05:46,920 --> 00:05:53,100
The focus in this course is on time to these forecasting, where the goal is to predict future values

61
00:05:53,190 --> 00:05:54,090
of the times it is.

62
00:05:55,720 --> 00:06:02,190
However, some of the methods that we will take, you can also be used for descriptive purposes.