Evaluation of classification algorithms for banking customer’s behavior under Apache Spark Data Processing System

Many different classification algorithms could be used in order to analyze, classify or predict data. These algorithms differ in their performance and results. Therefore, in order to select the best approach, a comparison studies required to present the most appropriate approach to be used in a certain domain. This paper presents a comparative study between two classification techniques namely, Naive Bayes (NB) and the Support Vector Machine (SVM), of the Machine Learning Library (MLlib) under the Apache Spark Data processing System. The comparison is conducted after applying the two classifiers on a dataset consisting of customer’s personal and behavioral information in Santander Bank in Spain. The dataset contains: a training set of more than 13 million records and a testing set of about 1 million records. To properly apply these two classifiers on the dataset, a preprocessing step was performed to clean and prepare data to be used. Experimental results show that Naive Bayes overcomes Support Vector Machine in term of precision, recall and F-measure.