Data Mining of Market Segments Based on the Dominant Movement Patterns of Tourists

1
Data Mining of Market Segments Based on the
Dominant Movement Patterns of Tourists

• Jianhong (Cecilia) Xia, Vic Ciesielski, Colin
  Arrowsmith, Graeme Wright and Katrina Spilsbury
• Department of Spatial Sciences
• Curtin University of Technology
• c.xia_at_curtin.edu.au
• Tel [phone]
• WALIS International Forum 2008
• March 14, 2008

2
Objectives

• This paper applies existing data mining methods
  using the Expectation-Maximisation (EM) algorithm
  to identify tourist segments, based on the
  dominant or statistically significant movement
  patterns observed in tourist behaviours

3
Methodology
4
Models of tourist movement
the movement pattern of tourist i.
tourist i being at attraction at
destination 1 and moving to
attraction at destination 2, At the end of
his or her trip tourist i may be found at
n the last destination visited by tourist i in
the study.


5
General Log-linear models

• To test associations and interactions between
  more than two categorical variables
• To test the association between attractions and
  to indicate which combinations of attractions are
  dominant.

6
Rationale for associations between attractions

• Close in space
• Pairs of attractions might have been packaged by
  operators
• Tourist desire or interests
• Park manager have developed promotion policies
• Economic reasons to optimise their trips

7
Tourism Market Segmentation

• a process of dividing a market into homogeneous
  subgroups. Tourists in the same group are similar
  to each other, and different from other groups
• The objectives of market segmentation
• Determination of segmentation variables
• Tree clustering, Two-way Joining, K-means
  clustering and the EM clustering

8
EM clustering algorithm

• Advantage of EM clustering algorithm
• categorical and numerical variables
• computing the number of clusters
• tending to be a small number of significant
  clusters and a number of others that are not
  interesting or useful
• Each tourist is assigned into a certain cluster
  with a certain probability, which govern the set
  of attribute values of observations (tourists) in
  the cluster.
• The EM algorithm has been implemented in the
  Waikato Environment for Knowledge Analysis (Weka).

9
Selection of target markets

• To identify the socio-demographic characteristics
  of tourists and their travel behaviours for
  significant spatial movement patterns.
• The segment with the heaviest users is selected
  as the target segment

10
Study area Phillip Island
11
Phillip Island
12
Survey Design

• Questionnaire design
• Questionnaire sampling

13
(No Transcript)
14

Identification of dominant movement patterns of
tourists on Phillip Island

• General Log-linear Analysis tool (GENLOG) in
  Statistical Package for the Social Sciences
  (SPSS).
• To identify Statistically significant attraction
  associations in each of the seven movement
  patterns groups

Table 1 Parameter estimates for two-attraction
movement patterns (n381)
15
Target markets for package DG.
16
Interesting results

• The maximum numbers of segments identified from
  the sample data are four.
• The key variables identified to distinguish
  tourist segments are Type of tourists, Age,
  Lifecycle and Time to visit Phillip Island.
• The tourists who travelled with different
  movement patterns could have similar
  characteristics
• Tourists who travelled with certain movement
  patterns had unique features.

17
Research limitation and future study

• Research limitation
• Sample size
• Interpretation of the results from EM algorithm
• Future study
• Tour package could be developed for market
  promotion
• Tourist transaction data could also be applied to
  tourist market segmentation using EM algorithm

18
Data Mining of Market Segments Based on the
Dominant Movement Patterns of Tourists

• Thank you
• Any questions please?
• Jianhong
  (Cecilia) Xia