Discovering Similar Cities Using Text Mining: A Recommendation Application for Turkey

The purpose of this study is to show that it is possible to benefit from the use of text mining to capture alternative cities that are targeted by a person on touristic journeys. The first process has been to collect the texts containing the descriptions of 100 cities and to convert them into a dataset form for text mining. Secondly, K-Means and the density-based spatial clustering of applications with noise (DBSCAN) algorithms have been used and compared to obtain similar cities. Multi-layer perceptron, Naïve-Bayes, K-Nearest Neighbor, Decision Tree and Support Vector Machine (SVM) algorithms have been used to classify these cities. Since Multi-Layer Perceptron yields over 70%, it has been determined to be the most successful algorithm for this purpose. The SOM (Self Organizing Map) algorithm has been used to obtain more consistent and accurate results of the distribution, and the clusters have been finalized. In analyzing of the application for Turkey, 28 cities in Turkey and 72 other cities have been evaluated and it has been possible to present the cities which have been similar to Turkey as alternatives. For this purpose, the obtained results from text mining have been visualized through a mobile application. The results of the analysis for the mobile application have been recorded in a database and presented to the user on the Android platform using the Windows Communication Foundation (WCF) web service methods.