Novel usage of wavelets as basis in RDNN for telecommunication applications

The paper explores the use of wavelet basis recurrent dynamic neural network (RDNN) to improve the estimation of reliability growth of communication network´s software. The presented RDNN handles noise contaminated data and provides enhanced speed and performance of the system as compared with alternate approaches. Nonlinearity of the system is represented by proper selection of the wavelet function. The integrated defect tracking model parameters and the data are fed to the RDNN and the network is trained for optimizing the defect tracking model performance. The designed system will assist the service release management in obtaining more effective risk reduction. Using wavelet basis for RDNN requires only 10 iterations compared to 200 iterations with hump as basis function. It also reduces the maximum percentage error from 88% to 7.69% in the expected outputs. This also improves the telecommunication system defect tracking and network deployment