Objectives: The effective detection system can help to detect breast cancer in the early stage. Focus of this paper is to improve Ultra Wideband (UWB) based early breast cancer detection efficiency. Methods/Statistical Analysis: Hence, a set of 125 data samples was created through forward scattering by placing two home-made UWB antennas between breast phantoms with and without tumor. Samples were divided into 5 groups with 25 samples each. K-fold cross validation method was used to train, test and validate the data set based on Feed Forward Back Propagation Neural Network (FFBPNN). Findings: The developed system detect breast tumor with average accuracy of (1) 100%, 78.17%, 70.66%, 92.46% and 86.86% and (2) 100%, 85.17%, 87.89%, 94.11% and 83,86 for existence, x, y, z and size respectively for (1) FFBPNN and (2) k-fold cross validation based FFBPNN. K-fold cross validation based FFBPNN showed in average 90.23% detection efficiency in terms of tumor existence, location (x, y and z) and size) which is nearly 5% improved compared to conventional FFBPNN (85.43%). The detected tumor is visualized in 2D and 3D environment. Application/Improvements: Hence, K-fold cross validation FFBPNN can be a future candidate as better and faster breast cancer detection system and saves precious lives.