FORECASTING

Considerations in Forecasting
Forecast Object :
What is the object that we want to forecast? Is it a time series, such as
sales of a firm recorded over time, or an event, such as devaluation of a
currency, or something else?

Information Set
• On what information will the forecast be based? In a time series
environment, for example, are we forecasting one series, several, or
thousands? And what is the quantity and quality of the data?
Considerations in Forecasting
Model Uncertainty and Constant Improvement :
Does our forecasting model match the true data generating process. Of
course not. All models are false: they are intentional abstractions of a
much more complex reality.

Forecast Horizon
• What is the forecast horizon of interest, and what determines it? Are
we interested, for example, in forecasting one month ahead, one year
ahead, or ten years ahead (called h-step-ahead forecasts, in this case
for h = 1, h = 12 and h = 120 months)?
Considerations in Forecasting
Structural Change
• Are the approximations to reality that we use for forecasting (i.e.,
our models) stable over time?

Model Complexity and the Parsimony Principle

• The phenomena that we model and forecast are often
tremendously complex, but it does not necessarily follow that
our forecasting models should be complex. Bigger forecasting
models are not necessarily better, and indeed, all else equal,
smaller models are generally preferable (the “parsimony
principle").
Considerations in Forecasting
• Unobserved Components
• In the leading time case of time series, have we successfully
modeled trend? Seasonality? Cycles? Some series have all such
components, and some not. They are driven by very different
factors, and each should be given serious attention.
Steps to improved forecasts
Quantitative Methods of Forecasting
Time Series Data vs Cross Section Data
Time Series Data
Date Sales of Private Jets
Jan-2016 92
Feb-2016 91
Mar-2016 91
Apr-2016 89
May-2016 94
Jun-2016 93
Jul-2016 90
Aug-2016 89
Sep-2016 91
Oct-2016 87
Nov-2016 93
Dec-2016 98
Time plots and time series patterns: Horizontal Pattern
• HORIZONTAL pattern exists when the data values fluctuate around a constant
mean. (Such a series is called “stationary“ in its mean.) A product whose sales do
not increase or decrease over time would be of this type. Similarly, a quality
control situation involving sampling from a continuous production process that
theoretically does not change would also show a horizontal pattern.
Time plots and time series patterns: TREND
• A trend exists when there is a long-term increase or decrease in the data.
It does not have to be linear. Sometimes we will refer to a trend as
“changing direction,” when it might go from an increasing trend to a
decreasing trend. There is a trend in the antidiabetic drug sales data
shown in Figure
Time plots and time series patterns: SEASONAL
• A seasonal pattern occurs when a time series is affected by seasonal
factors such as the time of the year or the day of the week. Seasonality is
always of a fixed and known frequency. The monthly sales of antidiabetic
drugs above shows seasonality which is induced partly by the change in
the cost of the drugs at the end of the calendar year.
Time plots and time series patterns: CYCLE

• A cycle occurs when the data exhibit rises and falls that are not of a
fixed frequency. These fluctuations are usually due to economic
conditions, and are often related to the “business cycle.”

• The duration of these fluctuations is usually at least 2 years.

CYCLE vs SEASONS
• Many people confuse cyclic behaviour with seasonal behaviour, but they are really quite
different.
• If the fluctuations are not of a fixed frequency then they are cyclic; if the frequency is
unchanging and associated with some aspect of the calendar, then the pattern is seasonal.
• In general, the average length of cycles is longer than the length of a seasonal pattern, and
the magnitudes of cycles tend to be more variable than the magnitudes of seasonal
patterns.
• Many time series include trend, cycles and seasonality.
• When choosing a forecasting method, we will first need to identify the time series patterns
in the data, and then choose a method that is able to capture the patterns properly.
The monthly housing sales (top left) show strong seasonality within each year, as well as some
strong cyclic behaviour with a period of about 6–10 years. There is no apparent trend in the data
over this period.

The US treasury bill contracts (top right) show results from the Chicago market for 100 consecutive
trading days in 1981. Here there is no seasonality, but an obvious downward trend. Possibly, if we
had a much longer series, we would see that this downward trend is actually part of a long cycle, but
when viewed over only 100 days it appears to be a trend.
The Australian quarterly electricity production (bottom left) shows a strong increasing
trend, with strong seasonality. There is no evidence of any cyclic behaviour here

The daily change in the Google closing stock price (bottom right) has no trend, seasonality
or cyclic behaviour. There are random fluctuations which do not appear to be very
predictable, and no strong patterns that would help with developing a forecasting model
The Forecasting Process
• THE FORECASTING PROCESS is a series of connected activities that
transform one or more inputs into one or more outputs. All work
activities are performed in processes, and forecasting is no exception.
The activities in the forecasting process are:
• 1. Problem definition
• 2. Data collection
• 3. Data analysis
• 4. Model selection and fitting
• 5. Model validation
• 6. Forecasting model deployment
• 7. Monitoring forecasting model performance
The Forecasting Process
• Problem definition involves developing understanding of how the forecast will be used
along with the expectations of the "customer" (the user of the forecast). Questions that
must be addressed during this phase include the desired form of the forecast (e.g., are
monthly forecasts required). the forecast horizon or lead time. how often the forecasts
need to be revised (the forecast interval) and what level of forecast accuracy is required
in order to make good business decisions.

• Data collection consists of obtaining the relevant history for the variable(s) that are to be
forecast, including historical information on potential predictor variables. The key here is
"relevant"; often information collection and storage methods and systems change over
time and not all historical data is useful for the current problem.
The Forecasting Process
• Problem definition involves developing understanding of how the forecast will be used
along with the expectations of the "customer" (the user of the forecast). Questions that
must be addressed during this phase include the desired form of the forecast (e.g., are
monthly forecasts required). the forecast horizon or lead time. how often the forecasts
need to be revised (the forecast interval) and what level of forecast accuracy is required
in order to make good business decisions.

• Data collection consists of obtaining the relevant history for the variable(s) that are to be
forecast, including historical information on potential predictor variables. The key here is
"relevant"; often information collection and storage methods and systems change over
time and not all historical data is useful for the current problem.
The Forecasting Process
• Data analysis is an important preliminary step to selection of the forecasting model to be used.
Time series plots of the data should be constructed and visually inspected for recognizable
patterns, such as trends and seasonal or other cyclical components.
• If potential predictor variables are available, scatter plots of each pair of variables should
be examined.
• Unusual data points or potential outliers should be identified and flagged for possible
further study. The purpose of this preliminary data analysis is to obtain some "feel" for the
data, and a sense of how strong the underlying patterns such as trend and seasonality are.
This information will usually suggest the initial types of quantitative forecasting methods
and models to explore.

• Model selection and fitting consists of choosing one or more forecasting models and fitting
the model to the data.
The Forecasting Process
• Model validation consists of an evaluation of the forecasting model to determine
how it is likely to perform in the intended application. This must go beyond just
evaluating the "fit" of the model to the historical data and must examine what
magnitude of forecast errors will be experienced when the model is used to
forecast "fresh" or new data.
• Forecasting model deployment involves getting the model and the resulting
forecasts in use by the customer. It is important to ensure that the customer
understands how to use the model and that generating timely forecasts from the
model becomes as routine as possible.
• Monitoring forecasting model performance should be an ongoing activity after
the model has been deployed to ensure that it is still performing satisfactorily.
FORECASTING METHODS
Naive Methods of Forecasting (See Excel)
• Average method
• Here, the forecasts of all future values are equal to the average (or “mean”) of
the historical data

• Naive method
• Here, the forecasts of all future values are equal to the average (or “mean”) of
the historical data

• Seasonal naïve method

• A similar method is useful for highly seasonal data. In this case, we set each
forecast to be equal to the last observed value from the same season of the
year (e.g., the same month of the previous year).
Evaluating Forecasts
Forecast Evaluating Methods
Forecast Evaluating Methods
The Moving Average Method (See Excel)
• A moving average is a series of numbers obtained by overlapping
groups of two or more consecutive values in a time series.

• The average is “moving” because an ever-new average is

calculated by adding a more recent time series value to the
group and dropping the oldest one.
 Exchange Rate:
Japanese Yen Three Quarter Forecast Five Quarter Forecast
Against the US Moving Average Based on Moving Based on
Date Dollar (MA3) MA3 Average (MA5) MA5
Mar-80 245.52 NA NA NA NA
Jun-80 227.27 NA NA NA NA

• The simple statistical method Sep-80 219.17 230.65 NA NA NA

Dec-80 209.79 218.74 230.65 NA NA
of moving averages may Mar-81 207.73 212.23 218.74 221.89 NA
mimic some data better than Jun-81 221.45 212.99 212.23 217.08 221.89
a complicated mathematical Sep-81 233.40 220.86 212.99 218.31 217.08
function. Dec-81 223.13 225.99 220.86 219.10 218.31
Mar-82 238.09 231.54 225.99 224.76 219.10
Jun-82 244.35 235.19 231.54 232.08 224.76
Sep-82 262.49 248.31 235.19 240.29 232.08
The choice of the interval for Dec-82 253.79 253.54 248.31 244.37 240.29
Mar-83 237.39 251.22 253.54 247.22 244.37
the moving average depends
Jun-83 238.59 243.26 251.22 247.32 247.22
on the length of the underlying Sep-83 241.59 239.19 243.26 246.77 247.32
cycle or pattern in the original Dec-83 233.26 237.81 239.19 240.92 246.77

data. 245.52  227.27  219.17  209.79  207.73 1,109.47

  221.89
5 5
About Moving Average
• Useful when series is stationary
• Prefer time periods where RMSE is lowest
• Choice of interval depends upon length of the underlying cycle
• MA can mimic some data better then even sophisticated
mathematical models