This system utilizes historical data to determine the best algorithm and parameter combination to predict future values of time-series data.

**ALGORITHMS UTILIZED**

**Moving Average**

- The simplest approach of all
- Assumes the data has no trend component
- Calculates the predicted value by averaging the n most recent data points
- Each historical value carries the same weight

**Exponential Smoothing**

- Assumes the data has no trend component
- Diminishes the weight of older data based on the value of a parameter alpha (0 ≤ alpha ≤ 1)
- As alpha increases, it makes the predicted value more sensitive to each new data point
- Stable models use a lower value of alpha in order to make the prediction less responsive to changes

**Exponential Smoothing with a Linear Trend**

- Similar to exponential smoothing except it assumes that the data does contain a linearly-based trend component
- Uses alpha in the same way as exponential smoothing
- Also uses a beta value (0 ≤ beta ≤ 1) to adjust the sensitivity of the linear trend’s slope to new data

**Least Squares**

- Assumes that the model is based solely on a linear trend
- It minimizes the total of squared vertical differences between the data points and the regression line

**Winters Method**

- A derivation of exponential smoothing with a linear trend, except it uses a seasonality factor to track any cyclical trends
- Requires two complete cycles of historical data
- It incorporates the same alpha and beta values in the exponential smoothing with linear trend model
- Also uses a gamma value (0 ≤ gamma ≤ 1) to adjust the sensitivity to cyclical trends
- A lower value of gamma will dampen the cyclical portion of the demand function

**REFERENCES**

- Sipper, Daniel & Bulfin, Jr., Robert L.,
*Production Planning, Control, and Integration*(New York, New York: McGraw-Hill, 1997) - Lewis, William E.,
*Software Testing and Continuous Quality Improvement*(New York, New York: CRC Press LLC, 2000) - Jacobs, F. Robert & Chase, Richard B.,
*Operations and Supply Management: The Core*(New York, New York: McGraw-Hill, 2008) - Hopp, Wallace J. & Spearman, Mark L.,
*Factory Physics*(New York, New York: McGraw-Hill, 2001) - Sipper, Daniel & Bulfin, Jr., Robert L.,
*Production Planning, Control, and Integration*(New York, New York: McGraw-Hill, 1997)

We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.