AN 300 FINAL EXAM UKY QUESTIONS & ANSWERS
Time-Series forecasting - Answers :- A time series is a record of sequential
observations of an item of interest over time (e.g., demand/sales).
- Time-series forecasting is a predictive analytics technique that uses historical values
of data on an item of interest to:
- Uncover underlying data patterns in sequential observation of an item of interest
overtime
- Project the pattern into the future
- Predict the outcome of the item of interest
Time series forecast: 1. Level 2. Trend 3. seasonality
Data Patterns - Answers :• Horizontal/level • Trend• Seasonality• Cycle , Random
Horizontal/Level, Trend, Seasonality, Cycle terms - Answers :Horizontal/Level: A
constant average value over time.
Trend: A gradual upward or downward shift in values over time.
Seasonality: A recurring pattern that occurs at set periods within a larger time frame.
Cycle: An alternating pattern of points lying above and below an underlying pattern (as
opposed to random fluctuation) across multiple years.
Business application - Answers :• Marketing- Generate sales forecasts of different
brands of products
for production planning. • Management
- Predict market growth rate for strategic planning.• Operations
- Generate product demand forecasts for material requirements planning.
Naive Method - Answers :Naïve forecast for the next period = Actual value of the last
period.
- Require only one historical data value.
- Adapt more readily to sudden shifts in the data pattern.
- For data with any pattern
- Track recent movement of data
- Simple moving average of order 1
Naive Method example - Answers :Naive forecast for July will be the actual demand for
June
Smoothing methods - Answers :For short-range (e.g. weekly) forecasts of time-series
data with a horizontal pattern.
- Need a series of historical data values and other method depends parameters
- Historical moving/Simple average
- Simple moving average of order K
- Exponential smoothing with alpha and a starting forecast
"Smooth" out the random component of the time series. Slow to adapt to sudden shifts
in the underlying data pattern.
, Historical moving average - Answers :Historical moving average forecast for the next
period = Average of all prior actual historical data.
- The average of all prior actual data points in the time series.
- Also known as simple average method.
- Requires no other input data except the data points in the time series.
Historical mining average method example - Answers :The average demand "January to
June" for all months divided by the amount of months = average
Simple moving average - Answers :Simple moving average forecast of order "k" for the
next period = Average of the most recent "k" historical actual values.
- It requires the number of time periods used in the moving average (k) be given.
- A time period of one results in a naïve forecast.
- Choose a small number of time periods to track movement in the
most recent data points.
- Choose a large number of time periods to smooth out the random movement of the
data points.
- Use small k to track recent data movement, Use Large K to smooth out random
components in the data
Simple moving average method - Answers :if average demand is in July well add up
April, may and June then divide by 3
Exponential smoothing method - Answers :Exponential smoothing forecast of the next
period =
smoothing constant x historical value of the last period +
(1 - smoothing constant) x forecasted value for the last period.
- A weighted average of all prior historical actual values.
- Require a smoothing constant (a) and a "starting forecast" value to get the method
going.
- 0 <a<1
- The higher the value of a, the more weight is given to the recent data.
Exponential smoothing method example - Answers :0.1 x Actual for February + (1- 0.1)
x Forecast for February
= 0.1 x 700 + 0.9 x 650 = 655
Forecast for May = 0.1 x Actual for April + 0.9 x Forecast for April = 0.1 x 800 + 0.9 x
670.5 = 683.45
Linear regression models - Answers :For data with trend and/or seasonal patterns.
- Linear regression for trend - Linear regression for seasonality - Linear regression for
seasonality with trend
- Need a series of historical data values
• A mathematical equation that relates a dependent variable with one or more
independent variables.
Time-Series forecasting - Answers :- A time series is a record of sequential
observations of an item of interest over time (e.g., demand/sales).
- Time-series forecasting is a predictive analytics technique that uses historical values
of data on an item of interest to:
- Uncover underlying data patterns in sequential observation of an item of interest
overtime
- Project the pattern into the future
- Predict the outcome of the item of interest
Time series forecast: 1. Level 2. Trend 3. seasonality
Data Patterns - Answers :• Horizontal/level • Trend• Seasonality• Cycle , Random
Horizontal/Level, Trend, Seasonality, Cycle terms - Answers :Horizontal/Level: A
constant average value over time.
Trend: A gradual upward or downward shift in values over time.
Seasonality: A recurring pattern that occurs at set periods within a larger time frame.
Cycle: An alternating pattern of points lying above and below an underlying pattern (as
opposed to random fluctuation) across multiple years.
Business application - Answers :• Marketing- Generate sales forecasts of different
brands of products
for production planning. • Management
- Predict market growth rate for strategic planning.• Operations
- Generate product demand forecasts for material requirements planning.
Naive Method - Answers :Naïve forecast for the next period = Actual value of the last
period.
- Require only one historical data value.
- Adapt more readily to sudden shifts in the data pattern.
- For data with any pattern
- Track recent movement of data
- Simple moving average of order 1
Naive Method example - Answers :Naive forecast for July will be the actual demand for
June
Smoothing methods - Answers :For short-range (e.g. weekly) forecasts of time-series
data with a horizontal pattern.
- Need a series of historical data values and other method depends parameters
- Historical moving/Simple average
- Simple moving average of order K
- Exponential smoothing with alpha and a starting forecast
"Smooth" out the random component of the time series. Slow to adapt to sudden shifts
in the underlying data pattern.
, Historical moving average - Answers :Historical moving average forecast for the next
period = Average of all prior actual historical data.
- The average of all prior actual data points in the time series.
- Also known as simple average method.
- Requires no other input data except the data points in the time series.
Historical mining average method example - Answers :The average demand "January to
June" for all months divided by the amount of months = average
Simple moving average - Answers :Simple moving average forecast of order "k" for the
next period = Average of the most recent "k" historical actual values.
- It requires the number of time periods used in the moving average (k) be given.
- A time period of one results in a naïve forecast.
- Choose a small number of time periods to track movement in the
most recent data points.
- Choose a large number of time periods to smooth out the random movement of the
data points.
- Use small k to track recent data movement, Use Large K to smooth out random
components in the data
Simple moving average method - Answers :if average demand is in July well add up
April, may and June then divide by 3
Exponential smoothing method - Answers :Exponential smoothing forecast of the next
period =
smoothing constant x historical value of the last period +
(1 - smoothing constant) x forecasted value for the last period.
- A weighted average of all prior historical actual values.
- Require a smoothing constant (a) and a "starting forecast" value to get the method
going.
- 0 <a<1
- The higher the value of a, the more weight is given to the recent data.
Exponential smoothing method example - Answers :0.1 x Actual for February + (1- 0.1)
x Forecast for February
= 0.1 x 700 + 0.9 x 650 = 655
Forecast for May = 0.1 x Actual for April + 0.9 x Forecast for April = 0.1 x 800 + 0.9 x
670.5 = 683.45
Linear regression models - Answers :For data with trend and/or seasonal patterns.
- Linear regression for trend - Linear regression for seasonality - Linear regression for
seasonality with trend
- Need a series of historical data values
• A mathematical equation that relates a dependent variable with one or more
independent variables.
