Romeu Moura

Hands-on section showing mining techniques for the social web (as a graph): Use them to visualize human interactions at a higher level, be it on public social networks (like facebook) or Enterprise private social networks (like yammer). while the examples will be social, the techniques exposed are usable in any graph datastore, the exact techniques I shall focus on will be: * Extraction * Finding frequent patterns * (Un)Supervised pattern learning * Constructing decision trees * Entity resolution Everything will be illustrated in code, All code will be open-source and pushed to github.

Learn to use Bandit Testing for fun and profit. Measure the impact of your software changes in your community, not by guessing or by asking them but by watching their behaviour. Use that data to decide whether to keep or throw away your changes. While I shall endeavour to keep this talk non-dev friendly there will be hands-on code.

Method and tools to drive your open source project with a continuous loop of building measuring and learning, using the constant feedback of your community, experimenting constantly, acting over measured behaviour instead of gut feeling, being ready to accept that you are frequently wrong and should stop investing your time in features nobody wants. The whole idea is relevant for projects that are highly innovating, if you did not took the road less traveled: this talk may not apply to you. There will be a mix of strategy, methodology, organization and architecting but no code, the talk should be accessible for every audience.