Microsoft senior data scientist
微软西雅图总部资深数据科学家。 Leading the machine learning V-team including data scientists and data engineers to build big data machine learning system for Azure infrastructure forecasting and insight generation.
Microsoft cloud demand forecast influences the decision of an annual capital expenditure of multiple billion dollars and is crucial for Microsoft cloud business. We are building a big data machine learning system to scalably solve various demand forecast problems, improve forecast accuracy in situations where the traditional time series approach does not perform well, generate business insights and substantially improve automation. The system includes a unified data system built via data mining and various machine learning models. This talk will cover both data layer and various models of the model layer. We will demonstrate how the machine learning approach solves the forecast problem of extremely short history, how the deep learning models improves forecast accuracy compared to the classical approach for time series data, and how the system generates business insights and serves as a platform for business decision-making.
This tutorial will target industry, government practitioners and researchers with interests in building machine learning systems for demand forecast and insight generation. This talk will cover:
1. Big industry forecast complexity and the rationale of building machine learning systems.