Forecasting Methods

You have 15 forecasting methods for use in forecasting profiles that are based on Bayesian machine learning.

You can use one or a combination of these forecasting methods while configuring your forecasting profile. You also have three naive forecasting methods that are used when the other forecasting methods can't produce an acceptable forecast in a forecasting profile that's based on Bayesian machine learning.

The following table lists the 15 forecasting methods and three naive forecasting methods:

Forecasting Method Enabled in Predefined Forecasting Profiles Uses Short Causal Factors Uses Long Causal Factors Uses Multiplicative Causal Factors Uses Nonseasonal Causal Factors Supports Method Decomposition Supports Causal Decomposition Calculates MAPE, MAD, and Bias

Auto Regressive External Inputs (X)

No

No No No Yes Yes Yes Yes

Auto Regressive Integrated External (V)

No

No No No Yes Yes Yes Yes

Auto Regressive Logistic (A)

No

No Yes No No Yes Yes Yes

Causal Winters (B)

No

Yes No No No Yes No Yes

Combined Transformation (E)

No

No Yes No No Yes Yes Yes

Croston for Intermittent (F)

Yes

No No No No No No Yes

Dual Group Multiplicative (D)

No

No No Yes No Yes Yes Yes

Holt (H)

Yes

No No No No Yes Yes Yes

Logistic (G)

No

Yes No No No Yes Yes Yes

Modified Ridge Regression (M)

Yes

No Yes No No Yes Yes Yes

Multiplicative Monte Carlo Intermittent (K)

No

No Yes No No No No Yes

Multiplicative Monte Carlo Regression (C)

No

No Yes No No Yes Yes Yes

Regression (R)

Yes

Yes No No No Yes Yes Yes

Regression for Intermittent (J)

Yes

Yes No No No No No Yes

Transformation Regression (L)

Yes

Yes No No No Yes Yes Yes
Naive (N) Yes No No No No No No No
Holt Naive (T) Yes No No No No No No No
Moving Average Naive (O) Yes No No No No No No No

The six forecasting methods that are enabled in predefined forecasting profiles provide the best results in most business scenarios that are commonly found. You can use the remaining forecasting methods in user-defined forecasting profiles.

To enable you to analyze the generated forecasts, the letters representing the used forecasting methods and their weights are stored in these measures:

Note: The measures for method weights are populated only when you select the Include details of forecast methods check box on the Parameters tab in the Run Plan dialog box for a demand or demand and supply plan.

Measure

Description

Applicable Work Areas

Bookings Forecasting Methods

Contains the letters for the forecasting methods that are used for generating the bookings forecast for an item.

  • Demand Management

  • Demand and Supply Planning

  • Sales and Operations Planning

Bookings Forecasting Methods Weight

Contains the weight (percentage) assigned to a forecasting method during the bookings forecast generation.

  • Demand Management

  • Demand and Supply Planning

Consumption Forecasting Methods

Contains the letters for the forecasting methods that are used for generating the consumption forecast for an item.

Replenishment Planning

Shipments Forecasting Methods

Contains the letters for the forecasting methods that are used for generating the shipments forecast for an item.

  • Demand Management

  • Demand and Supply Planning

  • Replenishment Planning

  • Sales and Operations Planning

Shipments Forecasting Methods Weight

Contains the weight (percentage) assigned to a forecasting method during the shipments forecast generation.

Note: In a replenishment or sales and operations plan, this measure won't return any value because causal decomposition isn't supported in such plans.
  • Demand Management

  • Demand and Supply Planning

  • Replenishment Planning
  • Sales and Operations Planning

For information about forecasting methods, refer to the white paper titled "Oracle Demand Management Cloud Analytical Methods (Forecasting Methods)" that's available in Document ID 2582285.1 on My Oracle Support.

Naive Forecasting Methods

If none of the enabled forecasting methods in the predefined or user-defined forecasting profile can generate an acceptable forecast at any level of the forecast tree, then the following steps are taken:

  1. If the Holt (H) forecasting method wasn't enabled, and the number of data points in the history is sufficient, then a simplified version of the forecasting method called Holt Naive (T) is used at the highest level of the forecast tree. The generated forecast is used if acceptable.
  2. If the Holt Naive (T) forecasting method doesn't generate an acceptable forecast, then the following steps are taken on the basis of the value of the EnableNaiveForecast forecasting parameter:
    1. If the value of the forecasting parameter is zero, no forecast is generated.
    2. If the value of the forecasting parameter is 1, the Naive (N) forecasting method is used at the highest level of the forecast tree. A simple forecast is generated on the basis of the average of the bookings, consumption or shipments history.
      Note: In predefined forecasting profiles, the value of the EnableNaiveForecast forecasting parameter is 1. Therefore, in predefined forecasting profiles, the Naive (N) forecasting method is used.
    3. If the value of the forecasting parameter is greater than 1, then the Moving Average Naive (O) forecasting method is used at the highest level of the forecast tree. A simple moving average of the bookings, consumption, or shipments history is used for generation of the forecast. The value of the forecasting parameter determines the number of buckets that are used, and the bucket is specified in the Forecasting Time Level field on the Demand Tab on the Plan Options page for the demand, demand and supply, or replenishment plan.

Method decomposition and causal decomposition aren't available for naive forecasting methods.