Ch. 11 Forecasting Flashcards
- A process for predicting the future
- Forecasts become less accurate over time.
- Forecasts for aggregates are more accurate than forecasts for individual items.
- Forecasts are often used as a motivator or strategic objective.
- Forecasts should be developed using several different methods.
- A good forecast should consist of both a point as well as a range.
Forecasting Principles
What are the three types of Forecasting approaching?
- Judgmental
- Time Series
- Regression
A set of data points recorded over successive time periods
Projects past trends into the future
Chronologically ordered sequence of data values for one variable
Time Series
Examples of Time Series
- month unemployment rate for the period 2010 to 2019
- reports of the closing value Dow Jones Industrial average
technique that analyzes the past behavior of a time series variable to predict the future
extrapolation models
time series data set that contains no trend and the variance is constant.
stationary data
time series data set that contains a trend.
non stationary data
The degree of time lagged correlation in the time series
Autocorrelation
Four Common Time Series Accuracy Measures?
1 Mean Absolute Deviation (MAD)
2 Mean square error (MSE) (most common)
3 Mean absolute percentage error (MAPE)
4 Root Mean Square Error (RMSE)
A method for measuring the accuracy of a forecast by summing the absolute value of the differences between the actual and forecast values and dividing by the number of observations.
It is a standard error measurement that is used in time series analysis
Mean Absolute Deviation (MAD)
A method for measuring the accuracy of a forecast by summing the squared differences dived by the number of observations
It penalizes the larger errors by squaring them
Mean square error (MSE) (most common)
A method for measuring the accuracy of a forecast by summing the absolute percentage error
It is the average of absolute errors divided by actual observation values
It should not be used if there are zeros or near zeros in the actual data
Mean absolute percentage error (MAPE)
What is the simple moving average known as?
The smoothing method
A forecasting method based on averaging two or more consecutive time series data points
Simple moving average
A forecasting method based on placing a larger weight on the more recent of two or more consecutive time series observations
Weighted moving average
another averaging technique that is a forecasting method based on weighting the previous time series value using a smoothing coefficient
Exponential smoothing
a technique that involves taking the average of the averages.
Double moving average
- effective forecasting tool for time series data that exhibits a linear trend.
- it computes an estimate of the base, or expected, level of the time series, and the expected rate of increase or decrease per period
Double exponential smoothing (Holt’s Method)
The process of separating a times series into its basic components. You can use this to communicate time series to management
Decomposition
is the forecast value when the smooting coefficient for the simple exponential model is equal to one. (Also called the Naïve Model)
The forecast value when the smoothing value is one is the actual value for the previous period.
Previous actual value
when your data does not have a trend or a seasonal component what forecast methods should you use?
moving average and single expo smoothing
when your data has a seasonal component (with or without trend)
Decomposition or the Holt Winters method
when your data has a trend but does not have a seasonal component
Trend analysis or double expo smoothing
What are the four Time Series components?
Trend component
Seasonality component
Cyclic component
Irregular Component
- the long-term direction of movement in a time series.
- Trends can be detected using a scatter diagram particularly after the data has been smoothed
- Can be either upward or downward
Trend component
A graphic display of data plotted along two dimensions. In a time series analysis the X-axis is time
It is best used for detecting trends in a time series.
Scatter Diagram
Because the moving average, weighted average, and exponential smoothing techniques use some average of the previous values to forecast the future values, they consistently underestimate _________ and overestimate _________.
- underestimate the actual values if there is an upward trend in the data
- overestimate the actual values if there is a downward trend
Represents a pattern of change that is completed within one year and repeats itself regularly over the time series
Seasonality component
Type of seasonality component that tends to be on the same order of magnitude each time a given season is encountered
Additive Effects
Type of seasonality component that tends to have an increasing effect each time a given season is encountered
Multiplicative Effects
The long-term, multi-year variations around the trend of a time series generally attributed to changing business and economic conditions
Cyclic component
Components of Cyclical variation include:
- Upswing
- Contraction
- Trough
- Peak
The short term, unanticipated and nonrecurring factors in a time series. They are short term fluctuations that are neither systematic nor predictable.
Irregular Component
We can use build this type of model of a time series if data is available for one or more independent variables that account for the systematic movements in the time series.
Regression Model
A process for fitting a straight line through a time series
Regression
A method that uses regression analysis based on past time values to forecast future variable values
Autoregressive
does not have a cause and effect relationship with the time series, but the behavior of a predictor variable might be correlated with that of the time series in a way that helps us forecast future values of time series
predictor variable
a model that uses a straight diagonal trend line
Linear Trend Model
this is the Y in the regression function. Sum of the functions x and the values ( a linear function of time)
systematic variation (predictable)
the random variation in the time series not accounted for by our model (the error term of the equation)
unsystematic variation (unpredictable)
using a curved trend line to the data
Quadratic Trend Model
performance metrics recorded for numerous variables at the same point in time
Cross-sectional data
How can we gain more accurate forecasts?
by combining the forecasts from several methods into a composite forecast
If no one forecasting procedure is clearly better than the other it would be wise to?
combine the forecasts from the different procedures using a weighted average or some other method.
