ch 4. Flashcards
What are the four different types of forecasting models?
- qualitative
- quantitative
- simulation
- miscellaneous
types of qualitative models?
- market survey (market research)
- expert opinion (jury of executive opinion)
- Sales force consensus estimate
- delphi method
- historical analogy
types of quantitative models?
- time series (last period, arithmetic average, moving average, weighted moving average, exponential smoothing)
- causal models (associative models)(simple regression and multiple regression)
Principles of forecasting? (4)
- forecasts are (almost) always wrong
- forecasts should include an estimate of error
- near-term forecast are more accurate than long-tern forecasts
- forecasts are more accurate for a product group than a single product
Keys factors in forecasting? (7)
- past data
- subjective or objective
- continuous or snapshot
- cost
- accuracy
- external factors
- model
forecasting process?
- Determine items to be forecasted
- determine time interval and forecast horizon
- collect data
- discover patterns/components
- consider internal/external factors
- select model
- forecast and implement
- monitor performance
- revise/replace model (if necessary)
Time series analysis
- process of discovering demand pattern with respect to time
- process of projecting the past behavior in the future
- demand is related to time
components of time series analysis?
- trend
- seasonality
- cyclical variations
- random
components of time series analysis: trend
gradual up or down movement
components of time series analysis: seasonality
any specific pattern that occurs periodically and is repeditive
components of time series analysis: cyclical variations
an up and down movement that repeats itself over a long period of time
components of time series analysis: random
fluctuations with no specific/discernible pattern
characteristics of last period demand?
- only one previous data point is used
- cannot dampen (filter) unexplained variations (actually overreacts to unexplained variations)
- responds well to demands with trend
- not appropriate for seasonal demand
characteristic of arithmetic average
- all past data are used
- appropriate for level demand (dampens unexplained variations well)
- responds to trends with a lag
- not appropriate for seasonal demand
characteristics of moving average?
- More flexible than both Last Period and Arithmetic Average models
- If n is small: behaves more like Last Period
- If n is large: behaves more like Arithmetic Average
characteristics of weighted moving average?
Same as Moving Average but more flexible because both n and weights can be controlled
what are two opposing properties of forecasting demand?
- responsiveness: ability to quikcly respond to change in demand (desirable in stuations with a trend)
- stability: ability to stabilize demand (not to overreact to random variations) (desirable in situations with steady demand and no trend)
exponentially weighted moving average: level correction
applies to level demand, with no trend or seasonal pattern
exponentially weighted moving average: level and trend corrections
applies to demand with an upward or downward trend but no seasonal pattern
exponentially weighted moving average: lever, trend, and seasonal correction
applies to situations that include both trend and seasonality
level correction dampens (filters) unexplained variations? true? false?
true
Exponentially weighted moving average: a?
value of a controls the responsiveness and stability of the model
a=0; less responsive, most stable
a=1; more responsive, least stable
seasonality
- a unique characteristic
- each period in a cycle is a season
- within each cycle, the seasonal effect is different from one period to another
- however, for any period, the seasonal effect is identical from one cycle to another
general procedure of forecastingdemand with seasonality
- Estimate It (seasonal index) for each t
- Deseasonalize the actual demand
(by dividing actual demand by It ) - Use a model to forecast the demand (using the deseasonalized data from step 2)
- Reseasonalize the forecasts
(by multiplying deseasonalized forecasts by It)
