Data mining is a method to find relationships between data. In a broader sense: the discovery of knowledge in databases. It is a tool for automated discovery of the interesting non-obvious patterns hidden in a database that have a high potential for contributing to the bottom line (Peelen and Beltman, 2013).
Peelen and Beltman mention several data mining techniques which can all be useful in the hotel industry:
- neural networks
- evolutionary computation
- association rules
- decision trees
- case-based reasoning
However this blog will only focus on neural networks and give an example of how these networks can be applied with online-bookings. Neural networks are characterized by one input and one output layer. However, it is possible to apply additional layers. The ones between the input and output layer are referred to as hidden levels and there is no value given to them. In order to determine the final model, the weights and values of the nodes at the hidden levels will have to be calculated. They will be adjusted so that the input values correspond with the desired output values (Peelen and Beltman, 2013).
Nowadays, online bookings mostly include the whole service and policies offered by the hotel to the customers. Hotels provide a range of services to customers, which are mostly not limited to the simple reservation. Therefore, instruments should be more complete and sophisticated to pursue customer loyalty and customer take over (Corazzaa, 2014).
The software used to support customer bookings must route the customer’s preferences. The software should offer continuous service, a multi-language environment and safety of transactions. However, it should also have Definition Paths based on the characterization of the customer, which are a set of questions asked to customers in order to assess his or her booking preferences. This can be done through online questions or analyzing historical data of customers. The collection of this data enables the hotel to offer specific services of the customers’ choice (Corazzaa, 2014).
By using a neural network, hotels are able to offer a suitable function that receives as inputs the state of the hotel (for example the current number of rooms available, the number of free rooms tomorrow, etc.) along with the requests of the customer, and provides as output a solution, or alternative solution if the requests are not satisfactory. These functions and values strongly depend on the hotel structure. Therefore, it might be useful to use multiple layers in the neural network to map known input datasets into known output datasets, since there might be unknown functions with an unknown capability (Corazzaa, 2014).
For the greatest effectiveness it is important to keep updating the Definition Paths and keep the neural networks up to date. Since neural networks are already time consuming, using this data mining tool might not be convenient for every hotel.
Corazzaa M., Fasanoc G., Masonc F. (2014). An Artificial Neural Network-based technique for on-line hotel booking. Procedia Economics and Finance, Volume 15, Pages 45-55
Peelen, E. & Beltman, R. (2013). Customer Relationship Management (Second Edition). Amsterdam: Pearson Education Benelux BV