Girji. Metacode extensibility in Girji

Abstract

Online web services that store fitness and health related data is evolving to provide useful services to end-users. Examples of this include RunKeeper, Fitbit and MS HealthVault. These services interconnect creating an ecosystem of online web services. The services they provide are mainly targeted for the consumer marked. However, professional sport clubs may potentially benefit from integrating these systems in their daily activities. This, however, requires a different set of data analytic than commonly provided by these services.

A problem is connecting this type of ecosystems with statistical analytic tools like R, Matlab, and Excel for doing statistical analytics and machine learning. There is privacy concerns related to sensitive data and misuse. This thesis explores this problem creating an extensible system for doing analytic with statistical analytic tools on online data archives.

The system integrates the web services Fitbit and RunKeeper and creates a runtime support for analytics written in R and Python. The system encapsulates the burdens of privacy concerns and authentication for interacting with web services. The system implements operational consent to give athletes high level of control how data is used in analytics over long periods of time. This is provided through metacode abstraction and eligibility checking. Metacode extends the runtime dynamically by being able to run user provided code that for instance can check for privacy violations upon data accesses. The evaluation showed that Fitbit and RunKeeper as data archives for analytics have some constraints in concern of latency and rate limits. With caching and preemptive crawling the web services can become useful data sources for professional sport clubs to integrate with.