Final Exam Midterm Review Flashcards
What is forecasting?
Prediction
- predicting when or how big of the impact
- Best guess about what will happen in the future
What can over forecasting cause?
Excessive Inventory
What can under forecasting cause?
Loss of sale
What is the only certain thing about forecasting?
The only certain thing about a forecast is that it will be wrong
What is a source of major productivity improvements?
Reducing Forecast Errors
What are the elements of good forecasting?
- Modeling
- Constant Updating (as new data comes in)
- Intuition
- Risk-taking
What is the forecast breakdown for heirarchy?
Corporate - Merger and Acquisition Decision
\/\/\/
Business - New Product Introduction
\/\/\/
Functions - Operations: capacity, inventory, process, quality, supply chain
What is an accurate statement about forecast accuracy?
As the forecast horizon increases, forecast accuracy decreases
What is considered long term forecasting? and examples.
2+ years
- Long-range marketing program
- Facility planning
What is considered medium term forecasting? and examples.
2 months - 2 years
- Production Planning
- New Product Introduction
- Capital Budgeting
What is considered short term forecasting? and examples.
0 - 3 months
- Work Force Scheduling
- Inventory Management
- Demand Forecasting
What are the two forecasting methods?
Qualitative and Quantitative
What is the application for Qualitative and Quantitative Forecasting?
Qualitative = Long-term and medium-term forecast
Quantitative = Short-term forecast
What is the description for Qualitative forecasting?
- Subjective
- Based on people’s opinions (opinion matters)
- No specific models (may lack of reliable past data)
What is the description for Quantitative forecasting?
- Objective
- Based on numeric data and models (historic data matters)
What are some method examples for Qualitative forecasting?
- Delphi Method (A panel of experts)
- Market Surveys (questionnaires to gather data on market conditions)
What are some method examples for Quantitative forecasting?
- Time Series forecasting methods (e.g. Naive, moving average, exponential smoothing)
- Casual Forecasting Methods (linear regression)
What is short-term demand forecasting focused on?
What’s the product / service demand for the next time period (week, month, year)?
What are the time series methods under Short-Term demand forecasting? How many?
4 time series methods
- Naive method
- (Simple) moving average
- Weighted moving average
- Exponential Smoothing
What is a characteristic of an unrealistic demand pattern?
If the forecast is steady (in a straight line) for a long period.
i.e. Demand doesn’t fluctuates and stays in a straight line in one area of the mean (constant of 150 over a 10 week period)
What are characteristics of a realistic demand patterns?
- Fluctuates about a constant mean
- Fluctuates and an increasing (or decreasing) trend
- Fluctuates and has “seasonal” patterns (i.e. sales of winter jackets, summer shorts)
What are the general steps (procedure) for the four forecasting methods?
Dt —> At —> Ft+1
Step 1: Based on given demand (D) information, calculate average (A) i.e., a statistic for past demand
Step 2: Use the A to inform demand (F) for the next period
What are the notations in the general steps for forecasting quantitative methods?
Dt = Demand (D) = Observed/realized demand in period (t)
At = A statistic of past demand through period t (an average of past demand)
Ft+1 = Forecasted demand for a time period t
How do you calculate error for period (t)?
et = D1 - Ft
What can be said about the four forecasting methods (similarities)?
The four forecasting methods are basically the same method, but with different weighting schemes
What is the Naive method used for?
The Naive method only uses the most recent demand to generate a forecast for the next period
- Next period’s forecast = this periods sales
What is the formula for the naive method?
At = (100%)Dt = Dt
Explain the steps to get the forecast (F) in the naive method?
Get the demand (should already be provided), carry that over to the Naive Average (A) so that the demand equals the naive average (they are the same) and then carry A down to the forecast for the next period - the previous periods Naive Average (A) is the next periods Forecast (A) so if Period 2’s A is 150, then F for period 3 is 150.
What is an alternative formula for the Naive Method?
A1 = D1, A1 = F2
What is the (Simple) Moving Average method used for?
The simple moving average forecast (with order N) uses the N (most recent period’s) demands to generate a forecast for the next period
- Managerial Decision = N
What is the formula for Simple Moving Average method?
At = EDt / N
What are the procedure for the Simple Moving average method?
Step 1: At = Dt + D t-1 + …. + Dt-N+1 / N
Step 2: At = Ft+1
What is an example for the formula for Simple Moving Average method if N = 3?
*** We can use the method “three-period” moving average
A3 = D3 + D2 + D1 / 3 —–> average of demands in periods 3,2,1
and repeat the steps down
A4 = D4 + D3 + D2 /3 and so on
The forecast (F) for the next period is the previous period average (A).
Example = If A3 = 126, then F4 = 126
What is the Weighted Moving Average method used for?
This method is based on the weighted average of the past N periods.
- Managerial decision: N & Wi
What is the formula for the Weighted Moving Average?
At = E(Wi)(Dt)
What are the procedure for the Weighted Moving Average
Step 1: 𝛴WiDt = W1Dt+W2Dt-1+…+ WNDt-N+1
Step 2: A = Ft+1
What are the steps to finding Forecast (F) in Weighted Moving average utilizing the “three-period” forecast demand, N=3?
- Find demand (should already be given).
- Multiply the most recent demand by W1added by the second most recent multiplied by W2 + the further period by W3 and so on. Using Pemdas.
- The Average (A) will be used for the forecast in the next period so if A3 = 130 then F4 = 130.
What is Simple Exponential Smoothing used for?
Exponential smoothing uses the demands observed in all previous periods to generate the forecast
- ‘a’ places weights on previous time periods
- ‘a’ is usually given a value between 0 and 1
What are the two formulas for Simple Exponential Smoothing?
At = EWiDt
wt-i = a(1-a)
What are the procedures for Simple Exponential Smoothing?
Step 1: At = aDt + (1-a)Ft
(current average = a Current Demand + (1-a) Current Forecast)
Step 2: At = Ft+1
How to use the formula in steps for Simple Exponential Smoothing?
Recall Ft = At-1 (Current forecast comes from last average)
Ft = At−1= αDt−1+(1−α) Ft−1
Ft−1 = At−2 = αDt−2 + (1−α) Ft−2
Ft−2 = At−3 = αDt−3 + (1−α) Ft−3
Any Ft or At loops in all previous periods demands
What are the steps to finding the Forecast (F) in the Simple Exponential Smoothing?
Using demand and forecast for the current period, multiply the alpha that’s given by the current period demand and add that to 1 minus the alpha multiplied by the current forecast to get the Average (A) for example 107. Then that current period average (A) (example 107) will be used for the forecast for the next period. So if A1 = 107, then F2 = 107.
What are the two true statements using intuition when considering Simple Exponential Smoothing?
1 - Exponential smoothing uses the demands observed in all previous periods to generate the forecast for the following periods.
2 - The weights associated with each previous periods appear to be exponentially decreasing.
What is the weight formula for Simple Exponential Smoothing?
Wt-1 = a(1-a)^i
aE |0,1|
What do time series methods rely on?
They rely on historical data to understand future pattern?
What is the highlight for the Naive Method?
Only the last period is important in predicting demand
What is the highlight for the Simple Moving Average?
- The last N periods are of equal importance
- Managerial Decision: value of N
What is the highlight for the Weighted Moving Average?
- The last N periods are unequally important
- Managerial Decision:
a. Value of N
b. Value of weights (W) for each past demand
What are the highlights with the Simple Exponential Smoothing?
- All periods are important but the relative importance declines smoothly over time in an exponential fashion.
- Managerial Decision: value of a (smoothing constant)
What is true of the weight scheme in the Naive Method?
The most recent period has all the weights
Why concern forecast accuracy?
Every model is wrong, but which one is less wrong?
Why should firms estimate forecast accuracy?
- To monitor erratic demand observations or “outliers”
- To determine when the forecasting method is no longer tracking actual demand
- To determine the parameter values that provide the forecast with the least error
What is the forecast error (e) formula? Give an example.
Forecast error equals demand minus forecast
et = Dt - Ft
e2 = D2 - F2
What are two true statements regarding forecast errors for each period?
- It does not involve At (average) directly
- Errors can be positive (under-forecasting) or negative (over-forecasting)
What is the purpose for the Absolute Error Statistics?
Does not account for the relative magnitude of underlying demand values
What are the three forecast error methods under Absolute Error Statistics?
- MFE (Mean Forecast Error)
- MAD (Mean Absolute Deviation)
- MSE (Mean Squared Error)
What is the purpose for Relative Error Statistics?
Accounts for the relative magnitude of underlying demand values
What forecast error method is under the Relative Error Statistics?
MAPE (Mean Absolute Percentage Error)
What is the procedure for the Mean Forecast Error (MFE)?
Step 1: Calculate forecast error (e) for each period
Step 2: Calculate MFE by taking the average of the computed forecast errors
What are the steps in determining the Mean Forecast Error (MFE)?
- Find the forecast error (e) by subtracting demand (D) by forecast (F)
- Find the average of e by adding (and subtracting) all the values of e (e2, e3, e4, etc.) and dividing it by the number of periods (n) that have an error (so if period 1 does not have an error do not include that in the total so if there are 6 periods and only 4 have errors then only divide the sum total of e by 4)
What are the procedures for finding the Mean Absolute Deviation (MAD)?
Step 1: Calculate forecast error and absolute error for each period
Step 2: Calculate MAD by taking the average of the computed absolute errors
What are the steps in determining the Mean Absolute Deviation (MAD)?
- Find the forecast error (e) first
- Find the absolute error for the forecast error by only taking the positive number of each value in error (so -25 would be just 25)
- Add up all the absolute error and divide it by the total number being added
What is the procedure for finding Mean Squared Error (MSE)?
Step 1: Calculate squared error for each period
Step 2: Calculate the MSE by taking the average of the computed squared errors
What are the steps in determining the Mean Squared Error (MSE)?
- Find the forecast error (e)
- Square the value for e (so 25 squared equals 652)
- Sum the values of the squared error and divided it by the total number being added
What is the procedure for finding the Mean Absolute Percentage Error (MAPE)?
Step 1: Calculate absolute error percentage for each period
Step 2: Calculate MAPE by taking the average of the computed absolute error
What are the steps in determining the Mean Absolute Error Percentage (MAPE)?
- Find the forecast error (e)
- Take the absolute value of the forecast error (e) and divide it by the demand to get the percentage
- Add up all the percentages (without the percentage sign or else it’ll mess up the calculations)
- Divide the sum by the total number being added to get the percentage
What is the effect for weight when the alpha (a) is larger or smaller in the systemic change in the mean?
A smaller alpha = the most recent demand is lighter - is less responsive to recent changes
A larger alpha = the most recent demand is heavier - traces the trend faster
What is the effect of N in the Moving Average for the systemic change in the mean demand?
Smaller N traces the trend faster
Larger N is less responsive to recent changes
What parameter for a method is better in handling the systematic change in the mean demand?
- Moving Average
- Exponential Smoothing
Which parameter for a method is better in handling a random spike in demand?
- Moving Average
- Exponential Smoothing
What effect does N have in the Moving Average with random spikes in demand?
- Smaller N traces the random spike closely
- Larger N has a smoother forecast
What happens with alpha (a) in exponential smoothing for random spikes in demand?
- Larger a traces the random spikes closely
- Smaller a has a smoother forecast
What are the two exponential smoothing methods?
- (Simple) Exponential Smoothing
- Double Exponential Smoothing
What is the difference between Simple and Double exponential smoothing?
- Simple: Does not account for the trend explicitly
- Double: Accounts for the trend explicitly
What is the atheoretical approach and the methods when it comes to time series models?
- Assumptions: past data is best predictor for the future
- Models:
1. Naive Method
2. Moving Average
3. Weight Moving Average
4. Exponential Smoothing
5. Double Exponential Smoothing