Analysis & Prediction of Sales Data in SAP-ERP System using Clustering Algorithms

TL;DR

This paper applies clustering algorithms to analyze and predict sales data in SAP-ERP systems, revealing patterns that can enhance sales strategies and revenue, especially using partition methods like K-Means and EM.

Contribution

The study demonstrates the effectiveness of partition clustering algorithms over density-based and hierarchical methods for sales data analysis in an ERP context.

Findings

Partition methods like K-Means and EM outperform other clustering techniques.

Clustering reveals patterns useful for sales improvement.

Analysis confirms suitability of specific algorithms for sales data.

Abstract

Clustering is an important data mining technique where we will be interested in maximizing intracluster distance and also minimizing intercluster distance. We have utilized clustering techniques for detecting deviation in product sales and also to identify and compare sales over a particular period of time. Clustering is suited to group items that seem to fall naturally together, when there is no specified class for any new item. We have utilizedannual sales data of a steel major to analyze Sales Volume & Value with respect to dependent attributes like products, customers and quantities sold. The demand for steel products is cyclical and depends on many factors like customer profile, price,Discounts and tax issues. In this paper, we have analyzed sales data with clustering algorithms like K-Means&EMwhichrevealed many interesting patternsuseful for improving sales revenue and achieving higher sales volume. Our study confirms that partition methods like K-Means & EM algorithms are better suited to analyze our sales data in comparison to Density based methods like DBSCAN & OPTICS or Hierarchical methods like COBWEB.