Online Computation of Up-To-Date Summaries in the Swiss Feed Database
Samuele Zoppi · 2011
This thesis develops the up-to-date summary, an automatic approach to aggregate histories of nutrient measurements in the Swiss Feed Database. Since measurements are taken irregularly and are sparse in the time, a simple aggregation over the entire history is not representative of the real world state. We fight this challenge by detecting trends in history of measurements with a set of data fitting functions: uniform fitting function, linear regression and kernel regression. The experimental evaluation proves the scalability of our approach to aggregate the measurements of good and bad quality data. Further, this thesis contributes to the development of the Feed Database with the integration of the up-to-date summaries into web application and with the import of the raw temporal data.