Evaluating the Progressive Performance of Machine Learning Techniques on E-commerce Data

Abstract

The web in current years has been a big tendency, which helps researchers to make it a source of information and essential in the various fields of a commercial area that represents the e-commerce. Machine learning, a branch of artificial intelligence, plays a vital role in creating a great experience at e-commerce companies. Machine learning techniques follow different efficient ways to extract knowledge from huge amount of data. A reliable products analysis of the performance of any e-commerce company is critical. However, due to its various global infrastructures, many likely products get grouped in a different way. So, the analysis of quality of product always depends on the accuracy of the products classification. The better the classification, the more insights can be generating good category of products. Classification is a fundamental problem in machine learning. The main motto of this study is to compare the performance analysis (basis of accuracy). Different machine learning (supervised) methods are used to classify the products. In this paper we compare different machine learning techniques (Nonlinear and rule-based) to classify the products. These approaches have been tested with data from the Kaggle Otto Group Product Classification dataset. The performances of algorithms are measured in two cases, i.e., dataset before feature selection (before preprocessing) and dataset set after feature selection (after preprocessing) and compared in terms of accuracy. The experimental result shows that the overall performance of nonlinear machine learning (KNN) techniques is better than rule-based (C5.0) techniques. The result shows that among the individual classifiers implemented, k-nearest neighbor is having highest accuracy of around 88%. An extensive study is given to explain the efficiency of different classifiers.