Performance Enhancement of Logistic Regression for Big Data on Spark