In open source software development projects , a lot of data generated during each phase of software development . specially on bug tracking system where different users reported bugs, and diffferent issues. These data containts a lot of noise , uncertainity and trustworthiness issues. Now question is how to handle these challenges and if these are not handeled at a right time time then the perfromance of classifiers, or models can be degraded.
The truthfullness of data is a measure concern in open source software evolution. And unecrtainity is a infection avaalable in data then how to handle it? Can we take a subset data and checket if uncertinity as a factor exist or and the treatment , we are giving is working correctly......
We are not talking about the volume and handling those large data by using cloud/hadoop /mapreduce/radoop/creating multple nodes /division of data....
There is a corpus of knowledge on "proof of programme". You are addressing the dual problem, "Proof of data".
There are several ways:
-provenance and quality
you drink pure source water, which you know has been bottled close to the source
-post-processing and sanitising
If you have properties on your data vectors like (x,y) with x+y must be 2, then you can detect errors or inaccuracies and "correct"/modify to meet the constraint
Again, physics and intrinsic models of what the data are about (semantics) will help dramatically.
There is no such thing as data in the void. There is data with a physical meaning, and that is this physical meaning which you check/enforce.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Analysis
- Data Analysis
- Dataset
- Big Data