Browse Prior Art Database

A Method and System of Macro Demand Forecast

IP.com Disclosure Number: IPCOM000236458D
Publication Date: 2014-Apr-29
Document File: 8 page(s) / 161K

Publishing Venue

The IP.com Prior Art Database

Abstract

This article discloses a method and system to address the demand forecast problem, especially for the situation where the data series is short. The method takes product's generation dimension into consideration within each product category (e.g., high-end, mid-range, and entry as the high level categories, 1st, 2nd and 3rd generation of product within each category). Demands of one product category are composed by the sum of demands of different product generations within the category at a certain time period. Demands of the new generation of product at its entry phase is predicted by either the derived predicted demands from the whole category demand minus old generations' demands or by the Grey forecast model based on the limited existing demands (at least 4 data points). Under the situation where two predicted demands are both available, a comparison step against their historical accuracy is designed to make a choice from two values. The key advantage of invention is that new product demand can be forecasted more accurately with the converged capabilities of two different forecast technologies. A corresponding system is also described in this article to implement the method automatically.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 53% of the total text.

Page 01 of 8

A Method and System of Macro Demand Forecast

Product demand forecast from manufacturer or sales channel perspective is addressed in this disclosure. CN1815498A/US20060178927, Demand forecast system and method. It relates to supply chain management, to a demand forecast system and method with an adjustment mechanism based on a Grey forecast model. Drawbacks of known solutions are as follows. They either consider products with rich of data points or products with less data points separately. They do not combine known technology capabilities all together to advance forecast precision. They do not leverage composite relationship between product category and its product generations.

The core idea is to forecast for one product category and take different product generations into consideration. It combines ARIMA time series forecast method for the long period of data and Grey System theory and minus calculation based on ARIMA results for the short period of data (at least 4 data points). It uses a comparison module to decide the forecast method for the short period of data.

The key advantage of invention is that new product demand can be forecasted more accurately with the converged capabilities of two different forecast technologies.

The core ideas are listed here.

1. Forecasting one specific product demand under its entire product segmentation context;

2. Using ARIMA time series forecast method for existing products and entire product segmentation with rich data entries

3. Using derived variable from ARIMA predictions for new product with limited data entries

4. Using Grey forecast model for new product with limited data entries (at least 4 data points to predict the 5thpoint)

5. Using a comparison module to decide the forecast between derived variable from ARIMA and prediction from Grey forecast model

Fig. 1 shows the computational steps for the disclosed method.

1



Page 02 of 8


T : >=40

Fig. 1 Process of Method

The detailed computational steps are explained in the following two situations.

2



Page 03 of 8

3



Page 04 of 8

Fig. 2 Time series of demands for product segment and different generations

For the situation I, for example, for June, 2010, there are 4 prediction points

1. entire power high-end segment

     Data is starting from 2004/12, 66 (over 40, ARIMA usage threshold TARIMA; customizable in the system) actual data points indicate to use ARIMA model

2. P5 processor based high-end sub-segment

Similar to prediction point 1, data is starting from 2004/12, 66 (over 40, ARIMA usage threshold TARIMA; customizable in the system)

actual data points indicate to use ARIMA model

3. P6 processor based high-end sub-segment

Data is starting from 2008/7, 23 actual data points indicate to use the prediction mode for scarce data

a. If more than one prediction points are under scarce data prediction mode, only use derived variable method when other models don't work

b. 23 (over 4, Grey forecast model usage threshold TGrey)...