Phishing attacks became most common cyber security attack in the digital platforms. To identify these types of attacks using URL analysis, the proposed research has identified the important features and then to design a customized neural network, it has enhanced Bayesian optimizer by adding layers and estimator values to find the best parameters that suits for the given dataset. Previous works has focused on implementation using either traditional or pre-trained approaches which has achieved least accuracy with approximate values of 65.25%, so to find the solution for this problem, the proposed research defined an algorithm which defines the objective function to maximize the target values of the estimators. It also implements stratified validation to find the values at iteration and picks the one with best value as output. The model has parameterize all the estimators associated with the neural network and found that it works with an accuracy of 99.5% for training data and with 99.7% for validation data.