Demand Forecasting I Time Series Analysis

Demand Forecasting I
Time Series Analysis
Chris Caplice
ESD.260/15.770/1.260 Logistics Systems
Sept 2006
Agenda “Predictions are usually difficult
– especially for the future”
Yogi Berra
Problem and Background

Four Fundamental Approaches
Time Series Methods
MIT Center for Transportation & Logistics – ESD.260 2 © Chris Caplice, MIT
Demand Processes
Demand Forecasting
Predict what will happen in the future
Typically involves statistical, causal or other model
Conducted on a routine basis (monthly, weekly, etc.)
Demand Planning
Develop plans for creating or affecting future demand
Results in marketing & sales plans – builds unconstrained
forecast
Conducted on a routine basis (monthly, quarterly, etc.)
Demand Management
Make decisions in order to balance supply and demand within
the forecasting/planning cycle
Includes forecasting and planning processes
Conducted on an on-going basis as supply and demand changes
Includes yield management, real-time demand shifting, forecast
consumption tracking, etc.
Four Fundamental Approaches
Subjective Objective
Judgmental Time Series
Sales force surveys “Black Box” Approach
Delphi techniques
Uses past to predict the
Jury of experts future
Experimental Causal / Relational
Customer surveys
Econometric Models
Focus group sessions
Leading Indicators
Test Marketing
Simulation Input-Output Models
Often times, you will need to use a combination of approaches

Cost of Forecasting vs Inaccuracy
Å Overly Naïve Models Æ Å Good Region Æ Å Excessive Causal Models Æ
Total Cost
Cost
Cost of Errors Cost of Forecasting

In Forecast
Forecast Accuracy
Time Series
The typical problem:
Generate the large number of short-term, SKU level, locally
disaggregated demand forecasts required for production, logistics,
and sales to operate successfully.
Predominant use is for:
Forecasting product demand of . . .
Mature products over a . . .
Short time horizon (weeks, months, quarters, year) . . .
Using models to assist in the forecast where . . .
Demand of items is independent
Special situations are treated differently
New product introduction
Old product retirement
Short life-cycle products
Erratic and sparse demand
Time Series
Method of using past occurrences to model the future
Assumes some regular & recurring basis over time
Basic Components
Demand rate
Level (a) a
Value where demand hovers around
Trend (b)
time
Persistent movement in one direction
Typically linear but can be exponential, quadratic, etc.
Demand rate
Seasonal Variations (F)
Movement that is periodic to the calendar
b
Hourly, daily, weekly, monthly, quarterly, etc.
Cyclical Movements (C)
Periodic movement not tied to calendar time
Random Fluctuations (e or ε)
Irregular and unpredictable variations, noise S
Demand rate
Combine components to model demand in period t
Multiplicative: xt = (b)(F)(C)(e) time
Additive: xt = a + b(t) + Ft + Ct + et
Mixed: Combination xt = a + b(Ft)t + et
Time Series
Simple Procedure
1. Select an appropriate underlying model of
the demand pattern over time
2. Estimate and calibrate values for the model
parameters
3. Forecast future demand with the models
and parameters selected
4. Review model performance and adjust
parameters and model accordingly
Time Series: Example
What is the forecast for period t+τ
made at the end of period t, ̂xt,t+τ?
116
115
114
Demand
113
112
111 140
110 120
109 100
108 80
Volume
1
10
19
28
37
46
55
64
73
82
91
10
60
40
Week 20
0
0 20 40 60 80 100 120
Week
So, what can I say?

Time Series
How important is the history? Two
extreme assumptions . . . .
Cumulative Forecast Naïve Forecast
All history matters equally Most recent dictates next
Pure stationary demand Random Walk, Last is Next
Underlying Model: Underlying Model:

x t = a + e t x t = xt-1 + e t
where: where:
e t ~ iid (μ=0 , σ2=V[e]) e t ~ iid (μ=0 , σ2=V[e])
Forecasting Model: Forecasting Model:
∑
t
x t ,t +1 =
x
i =1 i x t ,t +1 = xt
t
Time Series
Moving Average
Only include the last M observations
Compromise between cumulative and naïve
Cumulative model (M=n)
Naïve model (M=1)
Assumes that some step (S) occurred
Underlying Model:
So, some questions
x t = a + e t How do we find M?
where: What trade-offs are
e t ~ iid (μ=0 , σ2=V[e]) involved?
How responsive are the
Forecasting Model: three models?
∑
t
xi
x t ,t +1 = i =t +1− M
M
Moving Average Forecasts
116.00
115.00
114.00
113.00
112.00
111.00
110.00
109.00
108.00
0
73
82
91
10
19
28
37
46
55
64
1
10
ActDemand Naïve MA3
MA10 MA20 Cumulative
Time Series: Exponential Smoothing
Why should past observations all be weighted the same?
Value of observation degrades over time
Introduce smoothing constant (α)
Underlying Model:
x t = a + e t
where:
e t ~ iid (μ=0 , σ2=V[e])
Recall that
Forecasting Model: et = xt - ̂xt-1,t
̂xt,t+1 = ̂xt-1,t + αet (0<α<1)
or
̂xt,t+1=αxt + (1-α)̂xt-1,t
̂xt,t+1 = αxt + (1-α) ̂xt-1,t

but recalling that ̂xt-1,t = αxt-1 + (1-α) ̂xt-2,t-1
̂xt,t+1 = αxt + (1-α)(αxt-1 + (1-α) ̂xt-2,t-1)
̂xt,t+1 = αxt + α(1-α)xt-1 + (1-α)2 ̂xt-2,t-1
̂xt,t+1 = αxt + α(1-α)xt-1 + α(1-α)2xt-2 + (1-α)3 ̂xt-3,t-2
̂xt,t+1 = α(1-α)0xt + α(1-α)1xt-1 + α(1-α)2xt-2 + α(1-α)3xt-3...
Pattern of Decline in Weight
1
Effective Weight
0.8
alpha=.1
0.6
alpha=.5
0.4
alpha=.9
0.2
0
0 1 2 3 4 5
Age of Data Point
Time Series: Non-Stationary Models
MA with Trend Data
30
25
20
Demand
15
10
0
1 2 3 4 5 6 7 8 9 10 11 12
Period
Demand MA(3) MA(6)
Note that MA and standard Exp Smoothing will just lag a trend
They only look at history to find the stationary level
Need to capture the ‘trend’ or ‘seasonality’ factors
Time Series: Level & Trended Data
Similar to exponential smoothing
Holt’s Method - smoothing constants for level (a) and
trend (b) terms
Underlying Model:
x t = a + bt + e t
Demand rate
where: e t ~ iid (μ=0 , σ2=V[e])
a
time
Forecasting Model:
Demand rate
̂xt,t+τ = ̂at + τ ̂bt b
time
Where: ̂at = αxt + (1-α)( ̂at-1 + ̂bt-1)
̂bt = β( ̂at - ̂at-1) + (1- β) ̂bt-1
Time Series: Level & Trended Data
This is a linear
Forecast follows weighted combination
on this line
of level at A and B
Forecast
at t+1
aˆt +1
aˆt = α HW xt + (1 − α HW )(aˆt −1 + bˆt −1 )
xt A
A B
Demand or Forecast
ât bˆt = β HW (aˆt − aˆt −1 ) + (1 − β HW )bˆt −1

C D
B (aˆt − aˆt −1 )
C
bˆt −1
aˆt −1 D
(aˆt −1 + bˆt −1 )
This is a linear weighted
combination of slopes
0 at C and D
t-1 t t+1
Time
Source: Atul Agarwal MLOG’05 Objective is to forecast t+1 & beyond
Time Series: Level & Seasonal Data
Multiplicative model using exponential smoothing

Introduces seasonal term, F, that covers P periods.
Underlying Model:
Demand rate
x t = aFt + e t a
where: e t ~ iid (μ=0 , σ2=V[e])

time
Forecasting Model: S
Demand rate
̂xt,t+τ = ̂at ̂Ft+τ-P
Where : ̂at = α(xt/ ̂Ft-P) + (1- α) ̂at-1

time
̂Ft = γ(xt/ a
̂ t) + (1- γ) ̂Ft-P
An Example where P=4, t=12, τ=1:

̂x12,13 = ̂a12 ̂F9
̂a12 = α(x12/ ̂F8) + (1- α) ̂a12
̂F12 = γ (x12/ ̂a12) + (1- γ) ̂F8
Time Series: Level, Seasonal, & Trended Data
Expand exponential model to include seasonality
Winter’s Method – Similar to Holt’s Method with added term
Seasonality is multiplicative
Underlying Model:
x t = (a+bt) Ft + e t
Demand rate
where: e t ~ iid (μ=0 , σ2=V[e]) a
Forecasting Model: time
̂xt,t+τ =( ̂at +τ ̂bt) ̂Ft+τ-P
Demand rate
b
Where : ̂at = α(xt/ ̂Ft-P) + (1- α)( ̂at-1+ ̂bt-1)

time
̂bt = β( ̂at - ̂at-1) + (1- β) ̂bt-1
̂Ft = γ(xt/ ̂at) + (1- γ) ̂Ft-P S
Demand rate
time
Comments on Time Series Models
Most of the work is bookkeeping
Initialization procedures can be arbitrary
Adding seasonality greatly complicates calculations
Most of the value comes from sharing with users
Provide insights into explaining abnormalities
Assist in initial formulations and models
Picking appropriate smoothing factors
Level (α)
Stationary: ranges from 0.01 to 0.30 (0.1 reasonable)
Trend/Season: ranges from 0.02 to 0.51 (0.19 reasonable)
Trend (β)
Ranges from 0.005 to 0.176 (0.053 reasonable)
Seasonality (γ)
Ranges from 0.05 to 0.50 (0.10 reasonable)
Questions, Comments,
Suggestions?

Demand Forecasting I Time Series Analysis

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Demand Forecasting I Time Series Analysis

Uploaded by

Copyright:

Available Formats

Demand Forecasting I

Time Series Analysis

Problem and Background

Often times, you will need to use a combination of approaches

Cost of Errors Cost of Forecasting

So, what can I say?

Underlying Model: Underlying Model:

̂xt,t+1 = αxt + (1-α) ̂xt-1,t

̂xt,t+1 = αxt + (1-α)(αxt-1 + (1-α) ̂xt-2,t-1)

̂xt,t+1 = αxt + α(1-α)xt-1 + (1-α)2 ̂xt-2,t-1

̂xt,t+1 = αxt + α(1-α)xt-1 + α(1-α)2xt-2 + (1-α)3 ̂xt-3,t-2

̂xt,t+1 = α(1-α)0xt + α(1-α)1xt-1 + α(1-α)2xt-2 + α(1-α)3xt-3...

Pattern of Decline in Weight

Demand MA(3) MA(6)

Where: ̂at = αxt + (1-α)( ̂at-1 + ̂bt-1)

̂bt = β( ̂at - ̂at-1) + (1- β) ̂bt-1

ât bˆt = β HW (aˆt − aˆt −1 ) + (1 − β HW )bˆt −1

Multiplicative model using exponential smoothing

where: e t ~ iid (μ=0 , σ2=V[e])

Where : ̂at = α(xt/ ̂Ft-P) + (1- α) ̂at-1

An Example where P=4, t=12, τ=1:

Forecasting Model: time

̂xt,t+τ =( ̂at +τ ̂bt) ̂Ft+τ-P

Where : ̂at = α(xt/ ̂Ft-P) + (1- α)( ̂at-1+ ̂bt-1)

You might also like