Big Data Analytics in Business, Services Centers, and Healthcare

We describe our research based on extensive interactions with Silicon Valley and other firms, including AOL, SAP, Cisco, IBM, healthcare and energy.
I will focus particularly on Hierarchical Bayesian models, including Bayesian Kalman filtering, in addressing the following problems:
- Online Advertising: Attribution modeling – Assigning credit for user commercial actions to ad exposures. I will describe both aggregate methods and disaggregate methods which address incorrect A/B testing approaches. I will describe Big data and Sparsity issues.
- Energy Analytics for building energy optimization: I will indicate similarities in approaches based on Bayesian Kalman filtering
- Topic modeling, Information Extraction and Retrieval in Service Centers
o The key issue is Big Data resulting in cognitive overload
o We exploit the Generalized Dirichlet and other models which better capture causal models for text, and lead to significantly enhanced likelihood results with orders of magnitude speedup
- We consider combined state spaces of numeric and text to better predict wellness and interventions in healthcare

Speaker: Ram Akella, UC Santa Cruz