PERFORMANCE ASSESSMENT OF DIFFERENT CLASSIFICATION METHODS FOR COUPON MARKETING IN E-COMMERCE

Abstract

E-Commerce environment represents typical commercial transactions that take place virtually online. UK Online Shopping and E-Commerce Statistics provided by Nasdaq estimates that by the year 2040, 95% of all purchases around the world will be performed through e-Commerce. In 2021, there will be 2.1 billion digital buyers worldwide, up from 1.66 billion in 2016 (Spiralytics). In 2019, the 31 billion digital coupons were redeemed up from 16 billion in 2014; 77% of consumers spend 10-50$ more than anticipated when redeeming mobile coupons (Invesp). These statistics and expected trends in the future motivated us to investigate a performance assessment of different classification methods for digital coupon marketing. For this purpose, we used available data provided by the Data Mining Cup 2015. We compared three methods for decision tree generation (C4.5, C5.0, Random Forest), Naive Bayes, Support Vector Machine, and Logistic Regression. We tested their performance within different data samples created from the initial dataset. The best accuracy was provided by the C5.0 algorithm (91,3%) and Support Vector Machine (91,5%). These machine learning methods also had the highest success rate for the other metrics.

Keywords

• digital coupons
• e-commerce
• classification
• decision tree
• naive bayes