Alan Rezazadeh
PhD Southern Alberta Institute of Technology Calgary, Alberta, Canada
Title: A New Paradigm for Successful Implementation of Big Data Predictive Analytics and Forecasting
Biography
Biography: Alan Rezazadeh
Abstract
The main objective of this presentation is to discuss a new paradigm for successful implementation of big data analytics. Big data analytics may contain multiple sources of corporate, proprietary vendors, and public data portals (e.g. governments) to support complex business processes. Amalgamation of the required data for predictive analytics may present challenges for data transfers, integrations, updates and quality issues such as missing and inaccurate readings. Adding to this complexity managing outliers, due to natural disasters such as forest fires may cause abnormal measurements and readings.
Predictive analytics may require tens of years of business process consumable data comprised of complex modelling and integration prior to analysis. The availability of large amount of required data demands effective and efficient data models for enabling business operations utilizing near real-time predictive analytics. The ultimate goal is avoiding retrieval of any unnecessary data in order to achieve the desired performance.
This presentation discusses two big data analytics initiatives, integrating multiple public and corporate data sources supporting decision makers for planning and monitoring business operations. One of the lessons learned was utilizing data virtualization for an effective implementation of data integration layer within a big data analytics initiative. Data virtualization methodology provided significant improvements during extraction, transformation, analysis as well as run-time performance, by eliminating duplication and loading only relevant data. The authors will address the lessons learned for effective design of analytics output to be understandable and consumable by business users to support their operational requirements.