Improving customer service with predictive models for IBM London Software Lab

点击翻译按钮获取中文摘要

IBM Canada London Software Lab aims to provide excellent, professional services and products to customers around the world for their IoT (Internet of Things) products. However, they find it challenging when trying to help their clients with performance trend prediction and hardware sizing recommendation. Specifically, their customers have transactions and hardware that differ greatly compared to the company's internal benchmark testing. There are no existing utilities that can accurately model these differences. For this reason, IBM Canada London Software Lab urgently needs novel machine learning technologies to build predictive models, and create an analytic system to assist customers with hardware sizing, performance trending and future data prediction tasks. My graduate students and I have proposed advanced machine learning solutions to make use of the client data of IBM London Lab and build accurate predictive models. Specifically, we plan to address technical challenges in feature selection, model evaluation and optimization using several state-of-the-art machine learning algorithms. Our solution will potentially help to improve the customer service of IBM London Lab. As we work with IBM London to provide better IoT hardware sizing and performance trending services to their clients in USA/Canada, we will generate direct and indirectly economic and social benefits for Canadians.