With
the use of neural network based models, tourism demand
forecasting softwares provide ongoing, dynamic and adaptive
information management. By Dr Kaustubh Chokshi*,
CEO of UK-based Intelligent Business Systems.
Before
September 11, 2001, no software system on earth could
have predicted the forthcoming global slump in tourism.
And immediately after that cataclysmic date, none was
required. The same could be said for the tsunami that
hit Asian coastlines in December 2004, albeit with largely
regional implications.
On
an on-going basis, the tourism industry is affected by
a wide range of fluctuating parameters of lesser proportions,
all of which have a positive or negative impact on movements
of tourists and their choice of destination. However,
the degree of impact varies and is not easily quantifiable.
Serious social conflicts, wars and economic crises are
obvious influencers, but there could be a host of factors
such as the price of oil or currency exchange rates that
also have a bearing.
Wouldn''t
anyone operating in the hospitality industry or tourism
trade give their right arm to have a system that could
take into consideration all these factors and come up
with reasonably accurate prediction of tourist traffic
for the approaching tourist season?
Obviously
it would be impossible to develop such a system using
traditional logic-based programs, given the non-linear
nature of the interaction parameters and the possible
outcomes. On the other hand, Artificial Intelligence technology
and neural network architecture eminently fits the bill,
and would be able to deliver an effective demand forecasting
model that is as accurate as one could hope to get.
In
fact, scientists from Intelligent Business Systems (the
organisation that I head) worked along with researchers
from a leading university in the UK to develop just such
a tourism demand forecasting model for a prominent client.
While I am not at liberty to divulge the specifics of
the project, some of the concepts and considerations that
went in to the building of the model should make for interesting
reading.
First,
any tourism demand forecasting model needs to monitor
electronic newswire services, analyse the textual information
they contain and extract events of relevance. Based on
the extracted factual knowledge, the system should be
able to predict potential variations in tourism demand
in the country or region directly affected by the events,
as well as in neighbouring regions. The realisation of
such a system requires the deployment of various techniques
from the field of Artificial Intelligence, including Natural
Language Processing (text analysis and classification),
and Time-series Processing and Forecasting.
Large
amounts of ever-increasing unstructured electronic data
exist in the form of databases, internet web pages, and
newswire archives. This information has to be classified
and organised, thus allowing consequent qualitative retrieval
of the data as a potentially mineable resource of information.
Quantitative
analysis of digital newswire information to derive useful
classification of events has of course not been commercially
available. If at all anything existed, it used techniques
such as machine learning and decision trees, which were
found to be adequate. Now, artificial neural networks
have been successfully applied to the task of such text
classification. This is due in part to recent advances
in the appropriate scientific theories that underlie such
approaches, and also the sheer increase in computational
power that is available for data crunching.
Making
actual "semantic" sense of the textual
information is of great commercial and scientific interest.
Semantic sense implies a more intelligent application
of natural language processing techniques to the actual
understanding of text rather than brute-force analysis
and categorisation. Artificial neural networks have proved
themselves to be the most accurate tool in this regard.
As mentioned in previous articles in the series for domain-b,
artificial neural networks have a computational approach
to data analysis that more closely resembles the powerful
abilities of the human brain. Neural networks importantly
have the ability to learn existing information and adapt
to new information. This makes neural networks more powerful
than many machine-learning techniques such as those used
within search engines like Google and Yahoo.
With
the use of neural network based models, tourism demand
forecasting software is able to provide an ongoing, dynamic
and adaptive management of information; over time, further
analysis can provide forecasts based on ongoing levels
of stability / instability and other parameters at worldwide
tourism destinations.
A
neural network model can be optimised for non-linear data
and time-series prediction. It integrates the various
parameters culled from the data with traditional exogenous
values; these values are usually employed in econometric
tourism demand forecasting. The neural network learns
the rules of these traditional exogenous values, as well
as the impact of the current factors, making for robust
and accurate forecasting.
The
tourism industry is one that is most affected by the uncertainties
of the prevailing social, political and economic conditions
and the constantly changing geopolitical landscape. Obviously,
if tourist traffic can be confidently predicted with a
high degree of accuracy, one can plan for the influx (or
lack of it) far more efficiently and cost-effectively.
Organisations (or governments) that are better equipped
for such predictions are more likely to succeed in the
emerging global markets, whether in tourism or otherwise.
And once again, it''s Artificial Intelligence that will
make the difference.
*
Dr Kaustubh Chokshi is CEO of Intelligent Business Systems
(IBS), a UK-based AI Enterprise Solutions company, which
has recently expanded into India. Dr Chokshi has a PhD
in Artificial Intelligence from the University of Sunderland,
UK.
|