Reading...
Play button
Play button
Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
22 Cards in this Set
- Front
- Back
|
List the different types of Time Series
|
1. Stationary - without a trend
2. with a trend 3. with a trend and seasonality 4. with a trend, seasonality and cycle |
|
|
When are Time Series successful?
|
1. In stable conditions
2. For short-term forecasts 3. As a base forecast 4. For screening data |
|
|
Types of Stationary series
|
1. Moving averages - 3- or 5-point moving averages (m-1 + m + m+1)/3
2.Exponential smoothing - gives more wt. to recent numbers 3. |
|
|
What is problem with even numbers in moving average?
|
The smoothed value no longer refers to a particular time period. It falls halfway between.
|
|
|
Number of points in moving average is equal to ..
|
the seasonality of the data.
|
|
|
Exponential smoothing is a way of ...
|
constructing an average which gives more weight to recent values.
|
|
|
What value does alpha usually take in exponential smoothing equations?
|
in the range of 0.1 to 0.4.
|
|
|
The forecast for any future time period in moving average is
|
the most recent smoothed value.
|
|
|
Exponential smoothing formula
|
new smoothed value = (1 - alpha)(Previous smoothed value) + alpha(Most recent actual value)
|
|
|
Forecast 3 months ahead
|
Most recent smoothed value + 3 x Trend
|
|
|
Define seasonality
|
a regular pattern of upward and downward movements which repeats itself every year or less
|
|
|
How to handle a series with trend and seasonality
|
The Holt-Winters Method, an extension of the Holt Method. Two formulae
|
|
|
Seasonality formula
|
Seasonality = actual data/smoothed data
|
|
|
Define a cycle
|
some regular repeating pattern of upward and downward movement of longer than a year
|
|
|
Most common method of dealing with trend, seasonality and cycle.
|
the decomposition method
|
|
|
Describe the Decomposition Method
|
breaks series in four distinct parts:
a. Trend b. Cycle c. Seasonality d. Random then first three reassembled to make forecast |
|
|
Trend is isolated by
|
regression analysis between the data and time
x = a + b + u x - actual data a + b = trend element u = residuals comprising seasonality, cycles, random |
|
|
Cycle is handled by
|
isolating any cycle in the data
|
|
|
Describe the Box Jenkins Method
|
allows forecasts to compensate for previous errors, as time goes by
|
|
|
ARMA is
|
a forecasting equation that combines past values of the variable with past values of the residual
|
|
|
Box Jenkins process
|
1. Pre-whiten
2. Identify 3. Estimate 4. Diagnose 5. Forecast |
|
|
Key points of Box Jenkins
|
Expensive
highly accurate for short term forecasting - 3 to 6 months |
