一种位置服务测试数据生成技术研究  

Geographical Data Generation for Testing Location-Based Services

作　　者：侯可佳[1,2] 黄军[1,2] 白晓颖[1,2] 周立柱[1,2] 

机构地区：[1]清华大学计算机科学与技术系,北京100084 [2]清华大学信息科学与技术国家实验室,北京100084

出　　处：《计算机学报》2016年第11期2161-2174,共14页Chinese Journal of Computers

基　　金：国家自然科学基金(61472193);北京市自然科学基金(4132062);国家"九七三"重点基础研究发展规划项目基金(2011CB302505);国家"八六三"高技术研究发展计划项目基金(2013AA01A215)资助~~

摘　　要：随着移动互联网的高速发展,移动服务越来越成为人们生活、工作中不可或缺的一部分,绝大部分移动服务都要以用户的位置信息为基础,位置服务(Location-Based Services,LBS)则通过调用GPS等服务在确定用户位置的基础上,向用户提供导航、服务推荐等功能.位置服务的基础性使得其正确性和完整性至关重要,同时其开放性和复杂性对测试提出了更高的要求.文中以反地理编码服务为例,探讨了位置空间测试数据的生成方法.基于模拟退火算法(Simulated Annealing,SA)的优化过程,提出了两种适应度函数定义方法,一种是基于有效点密集区假设,以提高位置空间覆盖率为优化目标;另一种是基于LBS故障模式假设,以提高潜在故障的检测能力为优化目标.研究引入了贝叶斯分类器,将预测值作为探索点的能量值,指导搜索过程.研究选取了实际应用的位置服务平台进行了实验验证,实验表明,与简单随机算法相比,该方法在可容忍的时间代价下,可以有效提高测试效率和错误检测能力.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.

关 键 词：移动平台服务 位置服务 朴素贝叶斯理论 数据生成 模拟退火算法 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]