SMOTE Approach for Predicting the Success of Bank Telemarketing

At present classification of imbalanced data is a major concern. Machine learning algorithm fails to achieve expected result when data is imbalance. Major challenge of imbalance dataset is that sometimes minority classes are useful but machine learning algorithms are tended to be biased towards the majority classes and ignore the minority classes. To achieve the better result from our machine we need to train our machine using relevant data in such a way that makes those machines more talented and can give the accurate decision by itself for an unknown result. Here we propose SMOTE approach for predicting the success of bank telemarketing. Data collected form direct marketing campaigns of a banking institution. We analyzed 150 features related with bank client and product. Feature selection process was explored in the modeling phase and performed with dataset. Our dataset includes 45211 instances and 17 attributes. This large set of banking transaction data is imbalanced. In this paper we will construct a model to make these dataset balance using Synthetic Minority Oversampling Technique (SMOTE) technique and analyse the performance using Naive Bayes algorithm. Our model will help to find best strategies for the improvement of next marketing campaign.