Parameter tuning: Exposing the gap between data curation and effective data analytics

摘要

The “big data” movement promises to deliver betterrndecisions in all aspects of our lives from business to sciencernhealth, and government by using computational techniquesrnto identify patterns from large historical collections of data.rnAlthough a unified view from curation to analysis has beenrnproposed, current research appears to have polarized intorntwo separate groups: those curating large datasets and thoserndeveloping computational methods to identify patterns inrnlarge datasets. The case study presented here demonstratesrnthe enormous impact that parameter tuning can have on thernresulting accuracy, precision, and recall of a computationalrnmodel that is generated from data. It also illustrates thernvastness of the parameter space that must be searched inrnorder to produce optimal models and curated in order tornavoid redundant experiments. This highlights the need forrnresearch that focuses on the gap between collection andrnanalytics if we are to realize the potential of big data.