Complex Pattern Recognition using Feedback Algebraic Activation Function
P. Dhivya, A. Bazilabanu · 2021
Various researches on the Wisconsin breast cancer dataset in the prediction of benign and malignant were proposed in this work. The various classification models were used in the literature survey in which logistic regression give the better accuracy of 95.41%. But in machine learning model, the linear relationship between the features may lead to over fitting in that model. The neural network is a fully connected layer to find and understand the nonlinear combinations of the features in the dataset. To reduce the over fitting and improves the accuracy of the model, the Neural Network with Feedback Algebraic Activation (FAA) function is introduced and applied to the hidden layer in the network . In this, the error will be added as a feedback to the network and uses the keras callbacks and keras tuner for selecting the optimal parameters and produces the better results with the accuracy of 95.91% compared with all other activation functions in the neural network.