Geographical data generation for testing location-based services

With the rapid development of mobile cloud, mobile services have been an integrated part of people's daily life and work. Location-Based Services (LBS) provide information services, like navigation and recommendation, for given physical locations gathered through mobile network. It is critical enabling technique for more and more mobile applications. Hence, the correctness and completeness of LBS are important for mobile services. However, due to its openness and complexity, LBS testing faces new challenges. Taking reverse-GeoCoding service as an example, the paper investigates methods for geographic test data generation. Based on simulated-annealing method, two algorithms are proposed for the definition of fitness functions. One is based on the assumption of intensive area of valid positions, with spatial coverage as the optimization objective. The other is based on the assumption of LBS fault models, with the optimization objective to enhance defects detection probabilities. To guide effective search process, the energy in the annealing algorithm is defined by the predicted probability using Naïve Bayes classifier. Experiments are exercised on real mobile service platforms, which show encouraging results that the proposed methods can enhance test effectiveness and defect detection capabilities with tolerable time cost. © 2016, Science Press. All right reserved.