Keywords: Machine Learning, Algorithms, Data Science, Linked Data, Association Analysis, Cluster Analysis
This talk will present two perspectives that make data science interesting. The first will be to focus on the concept of Surprise Discovery in Data (SDD). The second will be to focus on some surprising (atypical) applications of some fairly typical machine learning algorithms. In the first case, SDD is like Outlier Detection 3.0. If we assume that first-generation outlier detection (1.0) is distance-based, and second-generation outlier detection (2.0) is density-based, then outlier detection 3.0 is pattern-based, where patterns can be as diverse as the data themselves: images, text, voice, high-dimensional streams, linked-data graphs, and more. SDD could therefore be called “Interestingness Detection” -- focused on finding the surprising, unexpected, and interesting points, patterns, and behaviors in big data. The talk will then continue with a discussion of several short examples of interesting (atypical) applications of typically uninteresting algorithms in different domains. I will not list the algorithms here, but hopefully you will be surprised by some of their applications.