ENTER2017 eTourism Conference presentation

1. Predicting Tourism Demand
Using Big Data – Issues and
Challenges
Haiyan Song
School of Hotel and Tourism Management
The Hong Kong Polytechnic University

2. How to use big data in tourism forecasting
 Predictive analytics gives marketing professionals more
insight into customer preferences, which can be used to
understand tourists better and improve tourism
firms/destination competitiveness
 Data mining techniques can be exploited to help
tourism forecasting with big data
 Differentiate big data from traditional tourism
forecasting techniques

3. How to use big data in tourism forecasting
 Making sure that tourism business benefits are derived from big
data. It is important to set specific business goals rather than
just dealing with the big data itself.
 Visualizing the big data with a view to understanding the nature
and characteristics of the data.
 Structuring the big data according to the requirements of
traditional forecasting techniques

4. How to use big data in tourism forecasting
 In order to apply traditional forecasting method to big data, we
have to simplify the structured big data
 One of the solutions is to shrink the big data and get the most important
information in a suitable format that can be easily applied to the traditional
forecasting models.
 Mixed frequency model with big data
 Using a novel forecasting method with the mixed-data sampling (MIDAS)
approach to fully utilize the high frequency search engine data
 Factor model and forecasting combination
 The best way to forecast the low frequency series (such as tourism
demand) using high frequency data is to combine the shrinkage method
with the mixed frequency models.

5. The framework of tourism forecasting with big data.
Traditional
Time Series
Model
Big Data
TransactionsWeb & Social IoTSensing Machine
Mixed
Frequency
Model
Shrinking Method
Factor Models
LASSO
Elastic Net
LARS…
Combine Forecasting
Data Storage
Model Selection
Large Structured Datasets
Information CriterionModel Fit Model Stability
Data
Processing
Results and Implications
Model
Optimization
Methodology
Data Shrinking
MF-VAR Model
MIDAS Model Time Series and
Statistical Techniques
Artificial Intelligence
Methods

6. Challenges
 Can the patterns that emerged from big data analysis or forecasting be
generalized?
 The second challenge is to find the right talents capable of both
working with the new forecasting technologies and interpreting the
data to find meaningful business insights.
 Third is to overcome the obstacle of data access and connectivity,
which requires the right platforms to aggregate and manage the big
data.
 The fourth problem is how to find new ways of leveraging big data in
tourism forecasting.

7. Thank You!