Community Mining for Predicting the Purchasing Behaviour of Customer in Shopping Dataset

Abstract

Evolution of technology has resulted in a significant growth in each and every field starting from business to research. This growth has caused a great impact in growing the business revenue. Multiple E-commerce sectors have been evolving day by day which are working day and night to reach the peak. At present the research is being carried out in every sector to determine the frequent customers, analyze their behaviour in terms their purchase and various other factors. The main aim is to get hold of the loyal customer by satisfying their needs. So every ecommerce sectors are moving in the direction to determine the best model to identify such key players. Key player is the one who is found to make the frequent purchase which helps to increases the revenue of retailers. The historical data has to be used to perform the analysis of key players in a shopping data set. Handling such huge data requires best tools and techniques and one such domain is Data Mining. In this proposed work an efficient model is developed by applying the data mining techniques such as preprocessing, feature selection, community build and mining. The dataset collected has to be cleaned in order to reduce the computation and to eliminate the error rate for which preprocessing is carried out using the algorithm regex and mean weighted average vector. The preprocessed data has to be reduced in terms of its dimension for which the feature selection is applied. The algorithms PCA with accuracy 89%, recursive feature elimination with accuracy 67% and Karl Pearson Correlation is also used and compared. The community is built for the extracted features using dependency algorithm. Community mining techniques has been applied to find the degreeness, closeness, betweeness property of the nodes is used to find the key player among the communities.