Download as rtf, pdf, or txt
Download as rtf, pdf, or txt
You are on page 1of 9

The Hong Kong Polytechnic University

Department of Industrial & Systems


Engineering

ISE430 Laboratory 1 Report

Student Name (ID)


Donald Montana (12105218D)

Submitted to
Dr. C.K. Kwong

Prepared on
16th October, 2015
Part a Results of the laboratory exercise
Dendrogram

Factor Analysis

Communalities

Initial Extraction

Quality 1.000 .658


Performance 1.000 .696
Userfriendliness 1.000 .344
Comfort 1.000 .694
Attractiveness 1.000 .631

Total Variance Explained

Component Initial Eigenvalues Extraction Sums of Squared


Loadings

Total % of Variance Cumulative % Total % of Variance

1 2.097 41.947 41.947 2.097 41.947


2 .926 18.517 60.464 .926 18.517
3 .818 16.361 76.826
4 .634 12.689 89.515
5 .524 10.485 100.000
Total Variance Explained

Comp Extracti Rotation Sums of


onent on Squared Loadings
Sums of
Square
d
Loading
s

Cumula Total % of Cumula


tive % Varianc tive %
e

1 41.947 1.551 31.021 31.021


2 60.464 1.472 29.443 60.464
3
4
5

Rotated Component Matrixa

Component

1 2

. .089
Quality 80
6
. .157
Perfor
82
mance
0
Userfri . .391
endlin 43
ess 8
. .826
Comfo
10
rt
4
. .777
Attracti
16
veness
3

Component Matrixa

Component

1 2

Quality .650 -.486


Performance .706 -.445
Userfriendliness .586 -.014
Comfort .640 .533
Attractiveness .650 .456

Component Transformation Matrix

Component 1 2

1 .731 .683
2 -.683 .731

Component Score Covariance Matrix

Component 1 2

1 1.000 .000
2 .000 1.000

Component Score Coefficient Matrix

Component

1 2

Quality .585 -.172


Performance .574 -.121
Userfriendliness .214 .180
Comfort -.170 .629
Attractiveness -.110 .572

Perceptual Map
Cluster Map

Part b Comments on the results


Dendrogram

For the analysis of the dendrogram, we simply divide the distance into an interval of 5, from
5 to 25 in order to conduct the segmentation based on the result. The corresponding segment
of each interval is presented below. To simplify the presentation, we assign a stage number to
each attributes.

Customer Stage
Q3Quality 1
Q3Performance 2
Q3USerFriendliness 3
Q3ComfortToCarry 4
Q3Attractiveness 5
Q3Price 6
Q3Barnd 7

Distance Segments
5 Segment 1: 1-2
Segment 2: 3-6
Others: 5, 7, 4
10 Segment 1: 1-2
Segment 2: 3-6-5-7
Segment 3: 4
15 Segment 1: 1-2
Segment 2: 3-6-5-7
Segment 3: 4
20 Segment 1: 1-2
Segment 2: 3-6-5-7-4
25 Segment 1: 1-2-3-6-5-7-4

From the segmentation results above, we can see the followings


When distance is set to 5, there would be totally 5 segments. Quality and Performance
would be clustered as segment 1. User-friendliness and Price would be clustered as
segment 2. This suggested that customers are interested in these two segments, while the
other are individual segments.
When distance is set to 10 and 15, the result obtained is the same. Quality and
Performance would be clustered as segment 1. User-friendliness, Price, Attractiveness
and Brand would be clustered as segment 2, leaving Comfort to carry as an individual
segment 3. In this case, we can see that the first segment is still the same as distance 5
while the second segment would be much larger as there are 4 attributes clustered
together. Customer would be interested in User-friendliness, Price, Attractiveness and
Brand, all four attributes in the second segment.
When distance is set to 20, there would be 2 segments left, including the first segment
merging Quality and Performance and the second segment merging User-friendliness,
Price, Attractiveness Brand and Comfort to carry.
When distance is set to 25, there would be only 1 segment left, which merging all 7
attributes together to form a segment. This is not quite desirable since it shows nothing
about the market segment for further analysis as there is only 1 very large segment.

Perceptual Map

From the perceptual map, it identifies the position of different product brands by selecting
two dominant attributes. In this case, X-axis represents VAR00001 and Y-axis represents
VAR00002. VAR00001 and VAR00002 are determined by equation:

F1 = a1Qj + a2Pj + a3Uj + a4Cj + a5Aj

F2 = b1Qj + b2Pj + b3Uj + b4Cj + b5Aj

respectively, with Q, P, U, C, A representing quality, performance, user-friendliness, comfort


and attractiveness respectively. a1 to a5 and b1 to b5 are coefficients of the related attributes
as presented below.
Component

1 2

Quality .585 -.172


Performance .574 -.121
Userfriendliness .214 .180
Comfort -.170 .629
Attractiveness -.110 .572

For VAR00001, we can see that the dominant attributes for this variable are Quality and
Performance, while the dominant attributes for VAR00002 are Comfort and Attractiveness.
The larger the coefficient of the attribute is, the more the related to the corresponding
variable. Therefore, when analyzing the perceptual map, we can simply use the dominant
attributes to explain the result.
From the perceptual map, we can see the followings
Samsung, Lenovo and Dell can be grouped into a cluster as they all sharing the same
propertiesBoth low Quality and Performance, and Comfort and Attractiveness. This
means that they have relatively low competitiveness.
Fujitsu and HP can be grouped into a cluster as they both have low Comfort and
Attractiveness while high Quality and Performance. Customer interested in Quality and
Performance and have no opinion about Comfort and Attractiveness would have a high
preference in these two brand.
Sony is an individual cluster showing both high Quality and Performance, and
Comfort and Attractiveness. It has the highest competitiveness since it shows superior
performance in VAR00001 and VAR00002.

From the information above, it indicates the opportunity for the product positioning. Since
the three clusters showed in the perceptual map positioned in top, left-bottom corner and
right-bottom corner, there would be a great opportunity to position a new product in the
middle part of the perceptual map which means to have so-so Quality and Performance, and
Comfort and Attractiveness. This would be very attractive to customers who do not desire a
high performance in either VAR00001 or VAR00002 while it should not be low. Mediate
performance in these two aspects would be their preferred choices.

Cluster Map

From the cluster map, there were totally 4 clusters which are indicated in the figure in section
1 with different colors. The REGR factor score 1 and score 2 refer to profile 1 and 2
respectively representing different attributes. We can conclude the followings
Cluster 1 is quite dispersed. It is not particularly similar to any profile, neither profile
1 and 2.
Cluster 2 is extremely similar to both profile 1 and 2.
Cluster 3 is very far from both profile 1 and 2.
Cluster 4 is quite similar to profile 1 while it is neither far from nor similar to profile
2.
From these interpretations, we can see that Cluster 2 has the highest performance in all
attributes including Quality, Performance, User-friendliness, Comfort and Attractiveness. In
contrast, Cluster 3 has the lowest performance in all attributes. Cluster 1 has mediate
performance in the 5 attributes. For cluster 4, it has high performance in Quality and
Performance, while so-so in the other attributes.

Overall Comments

In this laboratory, there were totally 3 methods of cluster presented, including the
dendrogram, perceptual map and cluster map. Dendrogram is very useful in determining the
suitable number of segments. It is a hierarchical and agglomerative process. By deciding the
distant value with all the required factors taken into considerations (such as cost and profit),
we can simply draw a vertical or horizontal line in order to determine the number of segments
as well as their corresponding attributes. Perceptual map is derived from a set of data, usually
from customer, about the rating of existing products in the market along various attributes. It
is the perceptions of similarities between brands and products. Cluster map is a non-
hierarchical process. It requires to pre-specify the number of clusters which later assign each
observation to a cluster then the distance between each point in that cluster and the mean
distance value will be calculated. Points will be assigned to a cluster in order to obtain a
minimum sum of distance between each point and the mean value.

In fact, all three approaches adopt the distance calculations. It is suggested that to perform the
hierarchical approach to obtain the number of cluster as the input for the K-mean clustering
approach which actually form the clusters. The computation of dendrogram may be very slow
depending on the number of attributes and K-mean clustering is a faster and reliable approach
in obtaining the cluster analysis.

You might also like