Parallel and online machine learning algorithms for cloud-based predictive analytics engine

Cloud-based predictive analytics service allows businesses to improve their performance with a minimal investment in infrastructure. Moreover, it enables collaboration and simplifies decision making. However, there are technical obstacles that prevent the large-scale deployment of such service to support millions of users. In fact, a cloud analytics platform runs on a distributed clustercomputing system. Hence, it is crucial to develop new algorithms that are optimized for parallel processing and efficiently use the cloud computing resources. Therefore, the goal of this research project is to develop such algorithms to enable fast and scalable predictive analytics. Precisely, the focus will be on developing new techniques for training the machine learning models by processing the data in parallel or by sequentially building the model in an online manner. These algorithms should be efficient in terms of computing costs and communication overhead. The new algorithms and a prototype application will be implemented in Haskell.

Faculty Supervisor:

Vijay Bhargava

Student:

Partner:

D&B Cloud Innovation Center

Discipline:

Engineering

Sector:

Finance and Insurance

University:

The University of British Columbia

Program: