Industrial Engineering: Forecasting

Method in Forecasting

In prevous post we learn some method in forecasting , now lets see how every method work ! .

1. Naive

In Naive method we just take prevous data to forecasting a next period , this the example :

Suppose D_oct = 90 units then demand prediction for November will be = D_oct = 90 units.

From above example we get formula Demand forecast for next period is same with the current demand:

D_t+1 = D_t

2. Moving Average

· Simple MA

In simple MA method we take m period prevous data and sum it and then divine with m , this the example :

from the example above , we get this formula for Simple MA :

D_t+m = D_t + D_t+1 + D_t+2 + … + D_t+m-1

___________________________ m

To be considered:

First forecasting value will appear in m+1 period.

m must be determined well in order to get valid result.

· Weighted MA

Each historical data has its own weight based on the confidence value.

D_t+m = w_tD_t + w_t+1D_t+1 + … w_t+m-1D_t+m-1

w_t + w_t+1 + … w_t+m-1= 1

for example :

MA with Trend

Trend in the historical data is involved in the demand forecasting.

D_t+n = a_0t + a_1t (n)

a_0t = 2MA_t^[1]- MA_t^[2]

a_1t = (2 / (n-1)) (MA_t^[1]- MA_t^[2])

here the example for MA with trend :

3. Exponential Smoothing

· Simple ES

It’s use for short predict so we just need a little data and new data . Exponential Smoothing divine in two method , simple and double

D_t+1 = αD_t + (1- α) Ď_t-1

Forecast for period 13 will be 108.52 = 109

· Double Exponential Smoothing

Used when historical data contains trend.

D_t+n = a_0t + a_1t (n)

a_0t = 2D_t- D_t^[2]

a_1t = (α/β) (D_t- D_t^[2])

Forecast for period 13 will be 169.83 + 5.07(1) = 174.9

Forecast for period 20 will be 169.83 + 5.07(8) = 210.6

And now for Validation which one is better , we use this 2 formula . But i preffer use MSE , its more accurate i think


	ok thank for all and for my teacher Muhammad Ridwan Andi Purnomo, ST, MSc, PhD thats for today , see you in a next post

Industrial Engineering