CTL.

SC1x -Supply Chain & Logistics Fundamentals

Time Series Analysis

MIT Center for

Transportation & Logistics
 Demand Forecast
Now
Past Future

Identifying patterns in the past

help us to forecast the future.

CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 6
 Agenda
Time Series Components
Cumulative Forecasts
Nave Forecasts
Moving Average Forecasts

CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 7
 Time Series Components

CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 8
 Time Series Components
Level (a)

Demand rate
n Value where demand hovers around (mean) a
n Captures scale of the time series
n With no other pattern present its a constant value
time

Trend (b)
n Rate of growth or decline

Demand rate
n Persistent movement in one direction
n Typically linear but can be exponential, quadratic, etc.
b

Seasonal Variations (F) time

n Repeated cycle around a known and fixed period

n Hourly, daily, weekly, monthly, quarterly, etc.
F
Can be caused by natural or man-made forces

Demand rate
n

Random Fluctuations (e or )
n Remainder of variability after other components time
n Irregular and unpredictable variations, noise

CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 9
 Time Series Components
Cyclical Movements (C)
n Periodic movement not of a fixed period
n Duration can be of different lengths
n Most often tied to longer term business cycles or economic conditions
80
70
60
50
40
70 30
Institute of Supply Management (ISM)

20
65
Purchasing Manager Index ( PMI)

Expansion
60
55
50

Contracting
45
40
35
30
Oct-06 Apr-07 Oct-07 Apr-08 Oct-08 Apr-09 Oct-09 Apr-10 Oct-10
Data source: Federal Reserve of St. Louis.
CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 10
http://research.stlouisfed.org/fred2/series/NAPM
 Time Series Models
Components can be combined in different ways:
Note we can transform the multiplicative
n Multiplicative: xt = bFtCtet xt = bFtCtet to: ln(xt)=ln(b)+ln(Ft)+ln(Ct)+ln(et)
n Additive: xt = a + bt + Ft + Ct + et
n Mixed: xt = (a + bt)Ft + Ct + et Model depends on how
xt = a + btFt + Ct + et seasonality impacts
xt = aFt + bt + Ct + et trend and/or level?

We will focus on four models

n Level Model: xt = a + et
n Trend Model: xt = a + bt + et
n Mix Level-Seasonality Model: xt = aFt + et
n Mix Level-Trend-Seasonality Model: xt = (a + bt)Ft + et

Notation:
xt = Actual demand in period t t = time period (0, 1, 2,n)
a = Level component b = linear trend
Ft = Seasonal index appropriate for period t Ct = Cyclical index for period t
et = Error independent random variable (=0) and constant 2
CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 11
 Cumulative vs. Nave Forecasts

CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 12
 Time Series Models
Predominant use of Time Series is for forecasting product demand of . . .
Mature products at the SKU level over a . . .
Short time horizon (weeks, months, quarters, year) . . .
Where demand of items is independent.
So, components used are level, trend, seasonality, and error.

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

CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 13
 Time Series Analysis
Critical assumption: How important is the history?
Two extreme assumptions: Very Important or Not at All
Cumulative Forecast Nave Forecast
n All history matters equally n Most recent dictates next
n Pure stationary demand n Random Walk, Last is Next

Underlying Model: Underlying Model:

xt = a + e t xt = xt-1 + e t

where: where:
et ~ iid (=0 , 2=V[e]) e t ~ iid (=0 , 2=V[e])
Forecasting Model: Forecasting Model:
t
x xt ,t+1 = xt
i=1 i
xt ,t+1 =
t
xt ,t+ = Forecast made at end of period t for demand in period t+ , for =1,2,3 ...
xt = Actual demand for period t
CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 14
 Cumulative vs. Nave Forecasts
Suppose we are at time=10 and want to find forecast for time=11; x^10,11
t xt 110
1 109 108 Cumulative Forecast:
2 92 106 10
104 x 995
3 98 i=1 i
x10,11 = = = 99.5
4 96 102 10 10
5 104 100

6 98 98

7 109 96
94 Nave Forecast:
8 99
9 94 92 x10,11 = x10 = 96
90
10 96
1 2 3 4 5 6 7 8 9 10 11

Lets look at next period forecasts for cumulative and nave models . . .
Cumul Nave 110
t xt x^ t,t+1 x^ t,t+1
108
Note:
1 109 109 109 106
104 Cumulative model is
2 92 100.5 92
3 98 99.7 98 102 calm while the
4 96 98.8 96 100 Nave model is
5 104 99.8 104 98 nervous.
6 98 99.5 98 96 Nave model is more
7 109 100.9 109 94 responsive than the
8 99 100.6 99 92 cumulative model.
9 94 99.9 94 90
1 2 3 4 5 6 7 8 9 10 11
10 96 99.5 96
CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 15
 Moving Average Forecast

CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 16
 Underlying Model:
Time Series Models xt = a + e t
Moving Average where:
n Only include the last M observations
n Compromise between cumulative et ~ iid (=0 , 2=V[e])
and nave Forecasting Model: t
xi
xt ,t+1 = i=t+1M
M
xt ,t+1 110

108

106
t xt Nave M2 M4 M6 Cum
104
1 109 109 109.0
2 92 92 100.5 100.5 102
3 98 98 95.0 99.7 100
4 96 96 97.0 98.8 98.8 98
5 104 104 100.0 97.5 99.8
6 98 98 101.0 99.0 99.5 99.5 96

7 109 109 103.5 101.8 99.5 100.9 94

8 99 99 104.0 102.5 100.7 100.6 92
9 94 94 96.5 100.0 100.0 99.9
90
10 96 96 95.0 99.5 100.0 99.5
1 2 3 4 5 6 7 8 9 10 11

Actual Nave M2 M4 M6

CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 17
 Moving Average Models
t
xi
xt ,t+1 = i=t+1M

Moving Average Model is a general model M

n Cumulative model (M=t)
n Nave model (M=1)

How big should M be?

n Too small? Overly responsive to noise, very nervous
n Too big? Averages out noise, misses step changes in demand
n Often use practical values of M (4, 6, 12, etc.)

Note that Moving Average models always lag!

n Assumes stationary demand
n The larger the M, the longer the lag

CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Time Series Analysis 18