The most prevalent type of cancer worldwide is liver cancer. Early detection followed by treatment of patients with liver cancer can improve their chances of surviving the disease. Liver cancer data was first gathered from the Kaggle website. In the gathered dataset, there are some irrelevant and wrong data presented. The pre-processing step is used to clean up the data and then the data is split and utilized for training and validation. An Artificial Neural Network (ANN) technique can automate the prediction of liver cancer. When building an ANN, data from approximately 80% of the original set for training the model. The last 20% data is then utilized to validate the model. Accuracy is high throughout the training phase; however, it is low during the validation phase. The accuracy rates obtained during training and validation were 98% and 82%, respectively. To avoid the problem of overfitting, L2 and dropout regularization are included in the ANN model. After applying the two-parameters, the accuracy of the training and validation stages is 99%. The accuracy and loss of a model are measured to determine its quality. The essential factors of the research are outlier removal, ANN model design, and overfitting rejection.