Transactions on Machine Learning and Data Mining (ISSN: 1865-6781)
Volume 10 - Number 2 - October 2017 - Pages 57-65
Simulation based Approach for Special Discounting Threshold
Shashank Shekhar, Kaushik Shankar, Sushanta Kumar Mishra, Sam Pritchett, and Parag Chitalia
VMware Inc., Palo Alto, USA
Abstract
There are many existing solutions available that provide a recommendation
to increase the threshold of the minimum order size to avail special
discounts (SPF) by minimizing the loss in order conversion. The goal of such
solutions is to maximize the earning and maximize quote to order conversion.
However, these solutions are primarily based on the relationship between order
size and quote to order conversion and utilize historical conversion rates at different
price points (List Price-Discount). This paper illustrates the use of Single
Item Single Period Newsvendor problem and Extreme Gradient Tree Boosting
methods to simulate a case for increasing order thresholds and calculating the
corresponding loss in order conversion. The boosting method was applied on
the training data and simulated and cross-validated across three test data sets to
test three different price change scenarios. We computed the expected quotes at
the end of the period and expected conversion at the end of the period to maximize
profit.
Keywords:Extreme Gradient Tree Boosting, SPF, List Price, B2B, Single Item Single Period Newsvendor problem, quotes, conversion
Download Paper
(379 KB)