Customer Behavior Analysis using Weighted Selective Naive Bayesian
R.Siva subramanian1, D.Prabha2
1R.Siva subramanian*, Research Scholar, Anna University, Chennai, India.
2Dr.D.Prabha, Associate Professor, Department of Computer science and Engineering, Sri Krishna College of Engineering and Technology, Coimbatore, India.
Manuscript received on January 14, 2020. | Revised Manuscript received on January 21, 2020. | Manuscript published on February 10, 2020. | PP: 1110-1115 | Volume-9 Issue-4, February 2020. | Retrieval Number: D1636029420/2020©BEIESP | DOI: 10.35940/ijitee.D1636.029420
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The predominant success of each enterprise relies upon the efficient analysis of customer behavior and a deep understanding of customers’ needs. Performing better customer analysis provides an effective analysis of potential customers, better decision-making, improved business processes, the measure of customers churn and enhance customer retention. Efficient customer analysis can be performed using the Naive Bayes (NB), which is a simple classifier used for predicting customer behavior. However, the performance prediction of Naive Bayes is strongly affected in some real-time datasets which involve the presence of correlated attributes and this creates a breach of the assumption made by the Naive Bayes model on the dataset. To enhance the performance prediction of the NB and to eliminate correlated variables, this research suggests a simple WSNB (weighted Selective Naive Bayesian) method that uses the C4.5 DT for selecting the attributes with high information values. Then the selected weighted attributes are further used with the Naive Bayes for improving performance prediction. The Experimental approach is tested with a bank client dataset, which indicates that WSNB makes better predictions than the standard Naive Bayes. Also, WSNB reduces the running time of the classifier by eliminating the correlated attributes, which in effect minimize the size of learning and testing data.
Keywords: Customer Analysis, Feature Selection, Decision Tree (DT), Machine Learning, Naive Bayes (NB), Prediction, WSNB
Scope of the Article: Machine Learning