Abstract: Programmed cell death is called apoptosis and when this process breaks down, cancer begins to form. The chance of survival and successful treatment greatly increases with the early detection of cancer. So in this paper we are proposing an artificial neural network based prediction algorithm for the early detection of cancer. In our proposed work, we divide the available cancer data records into two group called train data records and test data records. Based on a prediction algorithm, we train three different artificial neural network models to predict the abnormalities in the patient data by employing sliding window method. Later on, we use this trained networks to test data records and the results are analyzed using various metrics parameters such as sensitivity, specificity, accuracy, precision and area under curve.

Keywords: Artificial neural networks; Sensitivity; Specificity; Accuracy; Precision.