Geographical Test Data Generation by Simulated-Annealing

Location-based services (LBS) have been widely used in many applications like navigation and recommendation, especially for mobile systems. Testing of LBS services is hard due to the difficulties to cover location areas and to evaluate the correctness of location information for LBS query with a given address. To cope with the difficulties, the paper proposed a testing framework to search geographical test data using simulated-annealing algorithm, and to validate the query results by comparing across different LBS platforms. The heuristic search is based on defect clustering assumption. That is, defects intend to cluster in certain areas. Hence, areas with detected defects deserve more test cases in the follow-up testing. A Bayes classifier is used to predict defect probability of geographical areas, and to guide position data generation in the annealing algorithm. Experiments on real LBS platforms showed that with acceptable cost, the proposed method can considerably enhance test effectiveness.