CHAPTER 02: DATA
PATTERNS AND CHOICE
OF FORECASTING
TECHNIQUES
INSTRUCTOR: ZUHAIB KHOKHAR
Outline
Data Patterns 1. Mean Absolute Deviation (RMS)
(MAD) Root Percent Mean
Types of Data Patterns 2.
2. Mean Square Error Square (R%MS)
1. Horizontal Pattern
(MSE) Forecasting Methodologies
2. Trend Pattern
3. Mean Absolute
1. Quantitative Models
3. Seasonal Pattern Percentage Error
(MAPE) 2. Qualitative Models
4. Cyclical Pattern
4. Mean Percentage Error 3. Technological
5. Autocorrelated Pattern
(MPE) Approaches
Prediction Error
Evaluating Reliability
Errors in Forecasting
1. Root Mean Square
Data Patterns
Simple observations of the data will show the way that data have behaved over time.
Data pattern may suggest the existence of a relationship between two or more
variables.
Data Patterns
Univariate forecasting techniques depend on past patterns, whereas multivariate
forecasting utilizes past relationships to predict the future.
Qualitative forecasts depend on the subjectivity introduced by the forecaster and
his/her intuition rather than on manipulation of past historical data.
Types of Data Patterns
5.
1. Horizontal 3. Seasonal 4. Cyclical
2. Trend Pattern Autocorrelated
Pattern Pattern Pattern
Pattern
Types of Data Patterns
1. Horizontal Pattern:
Horizontal pattern is also known as “Stationary Pattern”.
When there is no trend in the data pattern the forecaster is dealing with a horizontal
data pattern. This implies that the observations do not tend to increase or decrease in
any systematic way. In statistical terms we refer to this as a stationary pattern.
Types of Data Patterns
1. Horizontal Pattern:
Types of Data Patterns
2. Trend Pattern:
Trend pattern is also known as “Secular Trend”.
The trend pattern is a long-term growth movement of a time series. The trend may be
upward, downward, or steady. It can be pictured as a smooth linear or nonlinear graph.
Types of Data Patterns
3. Seasonal Pattern:
A seasonal pattern in the data is observed when a predictable and repetitive movement
is observed around a trend line within a period of one year or less.
Types of Data Patterns
3. Seasonal Pattern:
Types of Data Patterns
4. Cyclical Pattern:
A cyclical pattern occurs with business and economic expansions and contractions.
Although there are similarities to a seasonal pattern, cyclical movements vary in
length, usually lasting longer than 1 year.
In forecasting, cyclical patterns are most difficult to predict because cycles do not
repeat themselves at constant intervals of time. Hence, it is impossible to distinguish
between cyclical behavior and smoothly evolving trend.
Types of Data Patterns
4. Cyclical Pattern:
Types of Data Patterns
5. Autocorrelated Pattern:
“Data in one period are related to their values in the previous period”.
This pattern simply shows that the data in one time period are related to their value in
previous periods. When faced with an autocorrelated data pattern, the forecaster
knows that there is an automatic correlation between observations in a series. What
this means is that, if there is a high positive autocorrelation, the value in June is
positively related to the value in May.
Prediction Error
Prediction error is error in forecasting.
It measures the average error that can be expected over time.
We can measure forecast accuracy by summarizing the forecast errors in different
ways.
Prediction Error
Where,
Error in Forecasting
1. Mean Absolute Deviation (MAD)
2. Mean Square Error (MSE)
3. Mean Absolute Percentage Error (MAPE)
4. Mean Percentage Error (MPE)
Error in Forecasting
1. Mean Absolute Deviation (MAD):
Mean Absolute Deviation (MAD) measures the accuracy of the prediction by
averaging the alleged error (the absolute value of each error). MAD is useful when
measuring prediction errors in the same unit as the original series.
Error in Forecasting
1. Mean Absolute Deviation (MAD):
Where,
Error in Forecasting
2. Mean Square Error (MSE):
Mean Square Error (MSE) is also known as Mean Square Deviation (MSD).
Mean squared error (MSE) measures the amount of error in statistical models. It
assesses the average squared difference between the observed and predicted values.
When a model has no error, the MSE equals zero. As model error increases, its value
increases.
Error in Forecasting
2. Mean Square Error (MSE):
Error in Forecasting
3. Mean Absolute Percentage Error (MAPE):
Forecasters also depend on mean absolute percentage error (MAPE) as a measure of
accuracy of a forecast. This measure is similar to MAD except that it is expressed in
percentage terms. The advantage of the measure is that it takes into account the
relative size of the error term to actual units of observation.
Error in Forecasting
3. Mean Absolute Percentage Error (MAPE):
Error in Forecasting
4. Mean Percentage Error (MPE):
There are instances when a forecasting method consistently forecasts low or high. To
determine whether such a bias is present in a technique, forecasters depend on the
mean percentage error as a measure of accuracy.
With no bias detected in the technique, the MPE is expected to be zero.
Error in Forecasting
4. Mean Percentage Error (MPE):
Error in Forecasting
Question: From the following data by using
Naïve Method:
Find:
1. Mean Absolute Deviation (MAD)
2. Mean Square Error (MSE)
3. Mean Absolute Percentage Error (MAPE)
4. Mean Percentage Error (MPE)
Error in Forecasting
Solution:
Error in Forecasting
1. Mean Absolute Deviation (MAD):
Error in Forecasting
2. Mean Square Error (MSE):
Error in Forecasting
3. Mean Absolute Percentage Error (MAPE):
Error in Forecasting
4. Mean Percentage Error (MPE):
Evaluating Reliability
Management is also interested in evaluating the reliability of a forecast with actual
performance.
2. Root Percent
1. Root Mean
Mean Square (R
Square (RMS)
%MS)
Evaluating Reliability
1. Root Mean Square (RMS):
Root Mean Square (RMS) is also called as Root Mean Square Error (RMSE).
Root mean square error or root mean square deviation is one of the most commonly
used measures for evaluating the quality of predictions.
Low RMS = Good (Best Fit Model)
High RMS = Bad
Evaluating Reliability
1. Root Mean Square (RMS):
RMSE could be more than 1. That's just mean that your model couldn't find a solution
or (optimized very well). It means that the squared root of squared differences
between your predictions and observed values is 11. Naturally lower values indicate a
better fit of your model.
Evaluating Reliability
1. Root Mean Square (RMS):
Evaluating Reliability
2. Root Percent Mean Square (R%MS):
The root percent mean square (R%MS) is the average error that is measured in
percentage terms.
Its interpretation is same like Root Mean Square (RMS).
Evaluating Reliability
1. Root Mean Square (RMS):
Evaluating Reliability
2. Root Percent Mean Square (R%MS):
Forecasting Methodologies
1. Quantitative 3. Technological
Models Approaches
2. Qualitative
Models
Forecasting Methodologies
1. Quantitative Models:
Quantitative Models are also known as “Statistical Models”.
Quantitative models are the objective approaches to forecasting.
The quantitative forecasting techniques include time series and regression approaches.
Forecasting Methodologies
1. Quantitative Models:
Where,
Forecasting Methodologies
2. Qualitative Models:
The qualitative methods, while lacking the rigor of the quantitative approach, do
provide good forecasts when experienced forecasters are employed.
Qualitative models are non-statistical models or jument models.
Forecasting Methodologies
3. Technological Approaches:
The techniques used in this category combine the quantitative and the qualitative
approaches so that a long-term forecast is made.
The objectives of the model are to respond to technological, societal, political, and
economic changes in order to make a forecast.
