Cost Sensitive Improved Levenberg Marquardt Algorithm for Imbalanced Data

Training of neural network is one of the most important aspects of neural network learning. Second order algorithm Levenberg Marquardt, is widely used for training purpose. But drawback related to this algorithm is large memory size, network size and network type. This work uses modified Levenberg Marquardt algorithm which removes memory size problem of traditional Levenberg Marquardt algorithm. Imbalance is major hurdle while training neural network. Cost sensitive approach is proven to be efficient technique to deal with this problem in case of neural network. This work utilizes the same approach to alleviate the effect of imbalanced dataset on neural network classifier.