CHAPTER

FORECASTING

Prepared By
Md. Limonur Rahman Lingkon
Lecturer
Department of Industrial & Production Engineering (IPE)
Rajshahi University of Engineering & Technology (RUET)
 FORECASTING

Forecasting is an estimate of an event which will happen in future. The forecast value is not a
deterministic quantity. The events may be-
• Demand of Product
• Rainfall at a Particular place
• Population of a country
• Growth of Technology etc.
Forecasting is required for-
• Materials Planning
• Scheduling Type of Production System
• Basis for coordination of plans for activities etc.

1
 BUSINESS FORECASTS TYPE

In Business, forecasts may be classified into 3 types. Such-

Technology Forecast :
The characteristics of Technology forecasts-
1. Level of technical performance.
2. Rate of technological advancement.
Economic Forecast :
The characteristics of Economic forecasts-
1. Level of business growth.
2. Level of employment.
3. Level of inflation etc.
Demand Forecast :
Expected level of demand for goods and services.
2
 FACTOR AFFECTING FORECAST (DEMAND)

• Business Cycle
• Random Variation
• Customer’s Plan
• Product Life Cycle
• Competition effort and price
• Customer’s confidence and attitude
• Quality
• Credit Policy
• Design of goods or service
• Reputation for service
• Sales effort
• Advertising
3
 TYPES OF DEMAND IN DECISION MAKING

1. Marketing 3. Finance
• Demand forecasting of products • Cash Flows
• Forecast of Market share • Rates of Expenses
• Forecasting trend in prices • Revenues

2. Production 4. Personnel
• Materials Requirement • Number of workers in each
• Trend in material and labor costs category
• Maintenance Requirements • Labor turn over
• Plant Capacity • Absenteeism

4
 SOURCE OF DATA FOR FORECASTING

• The quantity and type of data needed in • Must be relevant data

developing forecasts can vary a great deal
from one situation to another. • May be obtained from
• Company’s record
• Published records
• Some Forecast Techniques require only data
series that is to be forecasted. Such as: • Journal
Decomposition series, Exponential • Survey’s
smoothing etc. • Newspaper
• Etc.
• Some, like multiple regression methods • Suitable time period
require a data series for each variable must be ensured
included in the model of forecast.

5
 DEMAND PATTERN

Forecasting is based on the pattern of events in the Representation of Historical Pattern

past. A pattern may solely exist as a function of
time. Such a pattern can be identified directly from
historical data. It is important to understand the
most common demand patterns:
• Historical Pattern Forecast
• Seasonal Demand Pattern Variable

• Cyclical Pattern
• Trend

1. Historical Pattern (Stationary Pattern) :

This exists when there is no trend in data and
Time
when the mean value does not change over time.
6
 DEMAND PATTERN

2. Seasonal Demand Pattern: This demand exists 3. Cyclical Demand Pattern: In this type of pattern, the
when the series fluctuates according to some length of a single cycle is longer than a year. The cycle does not
seasonal factor. The season may be months, quarters, repeat at constant intervals of time. The best examples are the
weeks, sales of different products etc. prices of some metals, gross national product etc.

Representation of Seasonal Pattern Representation of Cyclical Pattern

Forecast
Forecast
Variable
Variable

Time
Time
7
 DEMAND PATTERN

4. Trend Pattern : This type of pattern exists when there is an increase or

decrease in the value of the variable over time. The examples are sales of many
products, stock prices, business and economic indicators.

Forecast
Variable

Time
8
 FORECASTING MODELS
The forecasting techniques can be classified into qualitative techniques and quantitative techniques. These are
presented below:
1. Qualitative Forecasting Techniques : They use subjective approaches. These are useful where no data is
available and are useful for new products. They are-
a. Delphi type method
b. Market Surveys
2. Quantitative Forecasting Techniques : They are based on historical data. These are more accurate and
computers can be used to speed up the process. They are-
a. Simple Moving Average
b. Single Exponential Smoothing
c. Double Moving Average
d. Double Exponential Smoothing
e. Simple Regression
f. Semi-average Method
g. Multiple Regression 9
 SELECTION OF FORECASTING TECHNIQUES

The selection of forecasting techniques depends on following

three factors. Such as-
1. The characteristics of the decision making situation, which
includes- 3. Present Situation, includes-
• Time Horizon • Item that is being forecast
• Level of Detail • Amount of Historical data
• Number of Items • Time Allowed for preparing
• Control versus Planning forecast

2. The characteristics of the forecasting methods, includes-

• Time horizon- Number of periods
• Pattern of Data
• Type of Model – Casual, Time Series, trend
• Cost
• Accuracy
• Ease of application 10
 MEASURES OF FORECASTING ACCURACY

• •

11
 MEASURES OF FORECASTING ACCURACY

• •

12
 MEASURES OF FORECASTING ACCURACY

13
 MATHEMATICAL ANALYSIS

1 150 165 -15 15 225 -10.000 10.000

2 160 165 -5 5 25 -3.125 3.125
3 165 165 0 0 0 0.000 0.000
4 175 165 10 10 100 5.710 5.710
5 180 165 15 15 225 8.330 8.330
5 45 575 27.165

MAD = (45/5) = 9; MSE = (575/5) = 115; MAPE = (27.165/5) = 5.433% MFE = (5/5) = 1

14
SIMPLE MOVING AVERAGE METHOD

16
 MATHEMATICAL ANALYSIS (3 MONTHS MOVING AVERAGE CALCULATION)

Time Demand for Moving

t Month t Average
M(t)
1 95
2 100
3 87 94.00
4 123 103.33 94.00 29.00
5 90 100.00 103.33 -13.00
6 96 103.00 100.00 -4.00
7 75 87.00 103.00 -28.00
8 78 83.00 87.00 -9.00
9 106 86.33 83.00 23.00
10 104 96.00 86.33 17.67
11 89 99.67 96.00 -7.00
12 83 92.00 99.67 -16.67 16
 MATHEMATICAL ANALYSIS (3 MONTHS MOVING AVERAGE CALCULATION)

Comments:

1. In most cases, this method is applied to forecast for only one period into the future.
2. The forecaster must wait until demand entries are available for making the first forecast.
3. This method is applicable only for horizontal demand data pattern.
4. This method is required to store the first n values, which consumes considerable amount of
computer storage space if the demands of several products are required to be forecasted.
5. Greater smoothing effect can be obtained by including more observations in the calculation of
the moving average.

17
 WEIGHTED MOVING AVERAGE METHOD

Equal weights are assigned to all periods in the computation of the simple moving average. The weighted
moving average assigns more weight to some demand values (usually the more recent ones) than to others.

Table: Three Months Weighted Moving Averages

Time in Moving
Months (t) Average M(t)
1 120
2 130
3 110 118
4 140 129 118
5 110 119 129
6 130 126 119

18
 DOUBLE MOVING AVERAGE METHOD
•
Time Demand F(t) e(t)
Month D(t)
(t)
1 60
2 70
3 85 71.66
4 60 71.66
5 88 77.66 73.66
6 68 72.00 73.77 85.66 -17.66
7 106 87.33 79.00 68.44 37.56
8 75 83.00 80.77 104.00 -29.00
9 86 89.00 86.44 87.44 -1.44
10 124 95.00 89.00 94.11 29.89
11 122 110.66 98.22 107.00 15.00
12 87 111.00 105.55 135.55 -48.55 19
 DOUBLE MOVING AVERAGE METHOD

Comments:
1. This method is applicable to trend data
patterns.
2. This forecaster must have 2n data
points to find a forecast and thus
substantial data storage is needed.
3. This method can be used to forecast for
any number of period into the future
easily.

20
 SIMPLE (SINGLE) EXPONENTIAL SMOOTHING METHOD

21
 MATHEMATICAL ANALYSIS
•

Week

Jan. Week 1 450 400 50 10 410

Jan. Week 2 460 410 50 10 420
Jan. Week 3 465 420 45 9 429
Jan. Week 4 434 429 5 1 430
Feb. Week 1 420 430 -10 -2 428
Feb. Week 2 498 428 70 14 442
Feb. Week 3 462 442 20 4 446
22
 ADJUSTED EXPONENTIAL SMOOTHING METHOD
•

MONTHS
DEMAND JANUARY FEBRUARY
(WEEKLY) 1 2 3 4 1 2 3 4
650 600 550 650 625 675 700 710
23
 MATHEMATICAL ANALYSIS

Week
Jan. Week 1 600 650 605 1 606
Jan. Week 2 605 600 605.4 0.88 606.28
Jan. Week 3 605.4 550 600.65 -0.246 600.404
Jan. Week 4 600.65 650 605.36 0.742 606.102
Feb. Week 1 605.36 625 607.99 1.12 609.11
Feb. Week 2 607.99 675 615.7 2.44 618.14
Feb. Week 3 615.7 700 626.33 4.08 630.41
24
Feb. Week 4 626.33 710 738.37 5.67 744.04
 LINEAR REGRESSION

Regression means dependence and it involves estimating

the value of a dependent variable Y, from independent
variable X. In Simple regression , one independent Y
variable will be used whereas two or more independent
variables will be used for multiple regression. The simple
linear regression can be represented as following form:
Y = a + bX
Where, Y = a + bX
Y = Dependent Variable
X = Independent Variable b
a = Intercept
b = Slope
a

X
25
 LINEAR REGRESSION

26
 MATHEMATICAL ANALYSIS

A firm believes that its annual profit depends on its expenditures for research. The information for the
preceding six years is given below. Estimate the profit when the expenditure is 6 units.

Expenditure for Research

Year Annual Profit (Y)
(X)
1 2 20
2 3 25
3 5 34
4 4 30
5 11 40
6 5 31
Solution: 7 6 ?

Year Expenditure for Research (X) Annual Profit (Y)

1 2 20 40 4
2 3 25 75 9
3 5 34 170 25
4 4 30 120 16
5 11 40 440 121
6 5 31 155 25

30 180 1000 200 27

 MATHEMATICAL ANALYSIS

28
 MATHEMATICAL ANALYSIS

Alpha company has the following sales pattern. Compute the sales forecast for the year 10.

Year 1 2 3 4 5 6 7 8 9
Sales (In
6 8 11 23 29 34 40 45 56
Lakhs)

Solution:
Sales (In
Year
Lakhs) x = X-5 xY
X
Y
1 6 -4 -24 16
2 8 -3 -24 9
3 11 -2 -22 4
4 23 -1 -23 1
5 29 0 0 0
6 34 1 34 1
7 40 2 80 4
8 45 3 135 9
9 56 4 224 16
Total 252 0 380 60 29
 SEMI AVERAGE METHOD
This method is sometimes employed when a line appears to be
an inadequate explanation of the trend.
Problem: The details of sales turnover of a fertilizer company
for the period 2004-2010 are given below. Compute the
estimated sales for the year 2011.

Year 2004 2005 2006 2007 2008 2009 2010

Sales
(in 39 48 65 78 95 91 112
Crore)

Solution:
Semi-average
Year Sales (in Crore) Semi-total
Trend Value
2004 39
2005 48 152 50.67
2006 65
2007 78
2008 95
2009 91 298 99.34
2010 112 30
 DELPHI METHOD
• It is a forecasting technique applied to subjective
nature of demand values.
• Technology forecasting is another example where Precautions-
there is no quantitative data based on which future • Panel members must be unknown to each other
technology can be predicted.
• Panel members should not discuss between
• Knowledge of past history of related Product is them before providing opinion
necessary for this method. • Questionnaire should be unambiguous
Procedure- • Opinions should be the matter that is sought for
• Knowledgeable persons are asked to provide • Similarity should be checked
estimates
• Different opinions are asked
• Conclude a consensus
• Iterative process converges to a specific value or
range of values
31
THANK YOU