Facial based human age estimation has become popular now a days due to tremendous increase in real time applications. Age estimation process comes with various challenges such as variation in lighting conditions, poses and facial expression. Performance of age estimation is evaluated with the help of measure ‘Mean Absolute Error’ (MAE). Main purpose of this work is improving facial based age estimation accuracy by building a Neural Network (NN). Beauty of NN is it learns nonlinear features of input data efficiently. For building NN model input images undergoes preprocessing, feature extraction, training and testing phases. NN model is built by providing training using various training algorithms. For prediction of age estimation neural network have shown superior performance for trainlm algorithm as compared to trainscg and traingdm algorithm. Experimentation are performed on partitioned training images and evaluated on testing images. We achieved low mean absolute error of 4.53 for 70 percent training images dataset.