Haengju Lee

An economic model for green cloud

A novel economic model for cloud-based services is presented that: (i) transparently presents energy demands (of services) to the customers in a simple abstract form, called green point, which is understandable to any general user; (ii) provides economic incentives (through dynamic discounts) as motivations for customers to select greener configuration; and (iii) offers service prices to customers such that the profit of cloud vendor is maximized while providing the discounts. Price is differentiated for different classes of customers (e.g. gold, silver, and bronze) and dynamic based on posterior distribution on resource demand considering both current demand and willingness toward green configuration. The model enables a paradigm shift in cloud service offering that provides higher transparency and control knobs to users for greener configuration. Preliminary results indicate higher profit using the proposed model compared to static pricing in existing pay-per-use service offerings.

Image categorization for marketing purposes

Images meant for marketing and promotional purposes (i.e. coupons) represent a basic component in incentivizing customers to visit shopping outlets and purchase discounted commodities. They also help department stores in attracting more customers and potentially, speeding up their cash flow. While coupons are available from various sources - print, web, etc. categorizing these monetary instruments is a benefit to the users. We are interested in an automatic categorizer system that aggregates these coupons from different sources (web, digital coupons, paper coupons, etc) and assigns a type to each of these coupons in an efficient manner. While there are several dimensions to this problem, in this paper we study the problem of accurately categorizing/classifying the coupons. We propose and evaluate four different techniques for categorizing the coupons namely, word-based model, n-gram-based model, externally weighing model, weight decaying model which take advantage of known machine learning algorithms. We evaluate these techniques and they achieve high accuracies in the range of 73.1% to 93.2%. We provide various examples of accuracy optimizations that can be performed and show a progressive increase in categorization accuracy for our test dataset.