New Product

OT prod
Product Capacity RT prod
Demand estimation
Out
Forecasting sourcing
Production
plan
Machine

Rush
order Disaggretion Quality &
Maintenance
Material and
Inventory
Scheduling

production

Use
 Forecasting

2014 25
2015 32
2016 24
2017 28
2018 26
2019 27
2020 ?

F=27 Simple Avg of all data

F=26.25 Avg of last four years
F=28 Increase in last 2 data
F=30 Other factor
F=26 Increase/decrease
F=27 Avg of last 3 data
F=26.833 Weighted Avg
F=26 Remove 32 and avg.

𝐹 =𝑎+𝜀
A= level or constant

𝜀 = 𝑛𝑜𝑖𝑠𝑒 𝑜𝑟 𝑖𝑛ℎ𝑒𝑟𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑛𝑐𝑒

𝑥1+𝑥2+𝑥3+⋯…+𝑥𝑛
1. Simple Avg= 𝑛
𝑤1𝑥1+𝑤2𝑥2+⋯….+𝑤𝑛𝑥𝑛
2. Weighted Avg Method= 𝑤1+𝑤2+⋯…+𝑤𝑛
3. Kth period moving Avg Method

Time series forecasting Model

Causal Models (Cause and effect analysis)

Exponential Smoothing Model

𝐹𝑡+1 =∝ 𝐷𝑡 + (1−∝)𝐹𝑡
∝= 𝑠𝑚𝑜𝑜𝑡ℎ𝑖𝑛𝑔 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡
𝐹7 =∝ 𝐷6 + (1−∝)𝐹6
𝐹6 =∝ 𝐷5 + (1−∝)𝐹5
𝐹2 =∝ 𝐷1 + (1−∝)𝐹1
Let’s Assume ∝= .2, ;F1=27 (simple Avg), 25, 32, 24, 28, 26, 27

F2=26.6;
 𝐹3 =∝ 𝐷2 + (1−∝)𝐹2
F3=0.2*32+(1-.2)*26.6=27.68; F4=26.94, F5=27.15; F6=26.92,F7=26.94

𝐹7 =∝ 𝐷6 + (1−∝)𝐹6
𝐹7 =∝ 𝐷6 + (1−∝)(∝ 𝐷5 + (1−∝)𝐹5 )
𝐹7 =∝ 𝐷6 + (1−∝) ∝ 𝐷5 +(1-∝)^2𝐹5 )

𝐹7 =∝ 𝐷6 + (1−∝) ∝ 𝐷5 +∝ (1−∝)2 𝐷4 + ⋯ … +∝ (1−∝)5 𝐷1 + (1−∝)6 𝐹1

=0.2D6+.16d5+….+

Infinite GP series,

=∝ +(1−∝) ∝ +∝ (1−∝)2 + ⋯ … +∝ (1−∝)5 + (1−∝)6 … + 𝑖𝑛𝑖𝑓𝑖𝑛𝑖𝑡𝑦 − − − −𝑑𝑒𝑛𝑜𝑚𝑖𝑛𝑎𝑡𝑜𝑟

a/1-r

∝/(1 − (1−∝)=1

∝ should satisfy two condition

∝> 0, ∝< 1
0 <∝< 1
∝= 2/(𝑛 + 1)
26, 28, 29, 32, 34, 35, 40

Simple Avg method=F8=32

Linear Regression Model

Suppose we have following data:

1 26
2 28
3 29
4 31
5 32
6 35
7

Chart Title
40

35

30

25

20

15

10

0
0 1 2 3 4 5 6 7

∑6𝑒=1 𝑒 2 minimise

Y=a+bt

y-a-bt

𝑒𝑡 = 𝑦𝑡 − 𝑎 − 𝑏𝑡
Minimize sum of square error

∑ 𝑒𝑡 2 = ∑(𝑦𝑡 − 𝑎 − 𝑏𝑡)2

Partially differ the above equation wrt a

−2 ∑(𝑦𝑡 − 𝑎 − 𝑏𝑡) =0……(1)

Partially diff w r t b

−2 ∑ 𝑡(𝑦𝑡 − 𝑎 − 𝑏𝑡) = 0……(2)

Expand above 2 equ

Expand above 2 equ

∑ 𝑦 = 𝑛𝑎 + 𝑏 ∑ 𝑡……(1a)

∑ 𝑌𝑡 = 𝑎 ∑ 𝑡 + 𝑏 ∑ 𝑡 2 …..(2a)

Yt t Y*t t^2
26 1 26 1
28 2 56 4
29 3 87 9
31 4 124 16
32 5 160 25
35 6 210 36
Summation=181 21 663 91

181=6a+21b

663=21a+91b

a=24.266; b=1.685

F7=a+bt

F7=24.266+1.68*7=36.02.

Linear Regression Model

In case of constant data

Y=C

𝑒 2 = ∑(𝑦 − 𝑐)2

∑𝑦
𝑐=
𝑛
HOLT’s Model

𝐹𝑡+1 = 𝑎𝑡 + 𝑏𝑡 forecast for the next period

𝑎𝑡 =∝ 𝐷𝑡 + (1−∝)(𝑎𝑡−1 + 𝑏𝑡−1 )
𝑏𝑡 = 𝛽(𝑎𝑡 − 𝑎𝑡−1 ) + (1 − 𝛽)𝑏𝑡−1
Data: 26, 28, 29, 31, 32, 35; ∝= .2; 𝛽 = 0.3

A1=D1=26
B1=35-26/6-1=1.8

F2=a1+b1= 26+1.8=27.8

𝑎2 =∝ 𝐷2 + (1−∝)(𝑎1 + 𝑏1 )=0.2*28+0.8*27.8=27.84

𝑏2 = 𝛽(𝑎2 − 𝑎1 ) + (1 − 𝛽)𝑏1 =.3*(27.84-26)+.7*1.8=1.81

F3=a2+b2=27.84+1.81=29.652.

Period t Dt at bt Ft
1 26 26 1.8
2 28 27.84 1.812 27.8
3 29 29.52 1.77 29.67
4 31 31.23 1.75 31.29
5 32 32.729 1.69 32.99
6 35 34.592 1.72 34.49
7 36.32

Seasonality Model

Year 1 Y2 Y3
Q1 53 58 62
Q2 22 25 27
Q3 37 40 44
Q4 45 50 56
157 173 189

Seasonality Index

Year 1 Y2 Y3 Avg
Q1 53/157=.34 58/173=.34 62/189=.33 0.336
Q2 22/157=0.14 25/173=.14 27/189=.14 0.14
Q3 37/157=0.23 40/173=.23 44/189=.23 0.23
Q4 45/157=0.28 50/173=.29 56/189=.30 0.29
1 1 1

Total Demand forgiven 3 years is 157, 173, 189

∑ 𝑦 = 𝑛𝑎 + 𝑏 ∑ 𝑡……(1a)

∑ 𝑌𝑡 = 𝑎 ∑ 𝑡 + 𝑏 ∑ 𝑡 2 …..(2a)

Yt t Y*t t^2
157 1 157 1
173 2 346 4
189 3 567 9
519 6 1070 14
a = 141; b=16

y=a+bt= F4=141+(16*4)=205

Year 1 Y2 Y3 Avg Y4
Q1 53/157=.34 58/173=.34 62/189=.33 0.336 205*0.336=69
Q2 22/157=0.14 25/173=.14 27/189=.14 0.14 205*.14=29
Q3 37/157=0.23 40/173=.23 44/189=.23 0.23 205*.23=47
Q4 45/157=0.28 50/173=.29 56/189=.30 0.29 205*.29=59
1 1 1

Causal Model

𝒚 = 𝒂 + 𝒃𝒙
∑ 𝑦 = 𝑛𝑎 + 𝑏 ∑ 𝑥……(1a)

∑ 𝑌𝑡 = 𝑎 ∑ 𝑥 + 𝑏 ∑ 𝑥 2 …..(2a)

𝒚 = 𝒂 + 𝒃𝒙𝟏 + 𝒄𝒙𝟐

𝒚 = 𝒂𝒙𝒃

Goodness of Forecasting Model

1) Mean Absolute deviation(MAD)

2) Mean squared deviation
3) Mean absolute percentage deviation
4) Tracking Signal

Data: 30, 32, 31, 30

Simple Avg Method=30.75

Absolute deviation=(|𝟑𝟎 − 𝟑𝟎. 𝟕𝟓| + |𝟑𝟐 − 𝟑𝟎. 𝟕𝟓| + |𝟑𝟏 − 𝟑𝟎. 𝟕𝟓| + |𝟑𝟎 − 𝟑𝟎. 𝟕𝟓|) = 𝟑
MAD=3/4

𝒆 𝒕 = 𝑫𝒕 − 𝒇𝒕
𝒏
𝟏
𝑴𝑨𝑫 = ∑|𝒆𝒊 |
𝒏
𝒊
𝑺𝑫 = (𝟑𝟎 − 𝟑𝟎. 𝟕𝟓)𝟐 + (𝟑𝟐 − 𝟑𝟎. 𝟕𝟓)𝟐 + (𝟑𝟏 − 𝟑𝟎. 𝟕𝟓)𝟐 + (𝟑𝟎 − 𝟑𝟎. 𝟕𝟓)𝟐 = 𝟐. 𝟕𝟓

MSD=2.75/4=
𝟏 𝟐
𝑴𝑺𝑫 = 𝒏 ∑𝒏𝒊 𝒆𝒊
 𝒏
𝟏 |𝒆𝒊 |
𝑴𝑨%𝑫 = ∑ ( ) ∗ 𝟏𝟎𝟎
𝒏 𝑫𝒊
𝒊
|𝟑𝟎−𝟑𝟎.𝟕𝟓| |𝟑𝟐−𝟑𝟎.𝟕𝟓| |𝟑𝟏−𝟑𝟎.𝟕𝟓| |𝟑𝟎−𝟑𝟎.𝟕𝟓|
=(( 𝟑𝟎
) ∗ 𝟏𝟎𝟎 + ( 𝟑𝟐
) ∗ 𝟏𝟎𝟎 + ( 𝟑𝟏
) ∗ 𝟏𝟎𝟎 + ( 𝟑𝟏
) ∗ 𝟏𝟎𝟎) =

Data: 30, 32, 31, 30

2) Now apply 2 moving avg method
F=30.5
MAD=3/4
MSD=3/4

Tracking Signal= SFE/MAD

𝒆 𝒕 = 𝑫𝒕 − 𝒇𝒕
SFE=∑𝒏𝒊 𝒆𝒊