A New Method for Missing Data Imputation in Banking Datasets Based on Biclustering

Online banking (Internet banking) has emerged as one of the mostprofitable e-commerce applications over the last decade and thusdata analysis and data mining techniques are extensively used toenhance decision making in financial institutions and banks. One ofthe main challenges in data mining for e-banking is the existence ofmissing values. A new method is proposed in this paper to imputemissing values based on the cross-relationship between informationstored in the banking databases. Given that all banking informationis not stored in a single table and there are useful data in othertables, it is demonstrated how missing values of city attribute incustomers table can be estimated using the information stored intransactions table. First a pivot table is generated based ontransactions table and then a biclustering algorithm is applied togroup customers. Finally, missing city values of the customers areimputed using existing ones. The experimental results show that theproposed method has better performance than classic imputationmethods and can be easily employed in other similar cases forimputing missing attributes in banking datasets.

کلیدواژه ها:

Banking ، Data Mining ، Missing Values ، Imputation ، Biclustering