Smartphone-based Data Collection with Stunner Using Crowdsourcing: Lessons Learnt while Cleaning the Data

摘要

The increasing popularity of smartphones makes them popular tools for various big data collecting crowdsourcing campaigns, but there are still many open questions about the proper methodology of these campaigns. Beyond this, despite the growing popularity of this type of research, there are familiar difficulties and challenges in handling a wide range of uploads, maintaining the quality of the datasets, cleaning the data sets containing noisy, incorrect data, motivating the participants, and providing support for data collecting regardless of the remoteness of the device. In order to collect information about the Network Address Translation (NAT) related environment of mobile phones, we utilized a crowdsourcing approach. We collected more than 70 million data records from over 100 countries measuring the NAT characteristics of more than 1300 carriers and over 35000 WiFi environments during the three year project. Here, we introduce our data collecting architecture, some of the most prominent problems we have encountered since its launch, some of the solutions and proposed solutions to handle difficulties.