Customer Behaviour Prediction Using Web Usage Mining

Abstract

Web usage mining is the use of data mining technologies to detect and
service the needs of web-based applications by identifying and mining interesting
usage patterns from web data. It entails first capturing client behaviour and flow on a
website, then mining this data for behavioural patterns. It is a critical component of
the ecommerce business that allows websites to review previously collected web
traffic statistics. By finding the patterns in this data, e-commerce companies may
improve their performance and recommend better items and services to customers.
The system is set up to track various analytics data and record web shopping/buying
behaviours in order to generate future prediction statistics. The system looks for user
budget tracking, comparing it to prior years, user bounce rates (the amount of people
who leave the payment page and come back), and other site usage indicators. In this
paper, We present a Heuristic-based Distributed Miner (HDM) design to obtain
consumer common behaviour patterns in real time, thereby solving the Web Usage
Mining (WUM) problem. Also provided is an improved k-means clustering (IKM)
algorithm for clustering web sites based on similarity function. Two assessment
measures are used to evaluate the proposed IKM method to the standard k-means
algorithm for cluster web sites: Sum of squared error (J) and Execution time (in mille
seconds). The experimental results shows that the proposed algorithms shows high
performance compared with previous algorithm.