الفهرس | Only 14 pages are availabe for public view |
Abstract Deep learning (DL) algorithms are used to predict many serious diseases. So that a large amount of healthcare data can be analyzed quickly. Cancer is one such disease, accounting for one out of every six deaths globally. Many academics have used predictive frameworks like Machine Learning (ML) and DL to forecast cancer illness, as well as the likelihood of recurrence, progression, and patient survival estimates. All stakeholders are currently concerned about the accuracy of cancer prediction. To increase cancer prediction performance, a framework among three DL frameworks with excellent accuracy and low prediction times was chosen in this thesis. We propose a binary version of the continuous AC-parametric whale optimization (BACP-WOA-S) technique since this prediction demands a quick and high-precision optimizer. This version is built on the sigmoid transfer functions. These functions were used to define a minimally optimal subset of features and to increase classification accuracy. These frameworks used Feed-Forward Neural Network (FFNN). They have the following forms: the first consists of 4 layers (Pre-processing, Feature selection, Un Optimized Deep learning, Prediction). The second consists of 3 layers (Pre-processing, optimized deep learning, prediction). The third combines the previous frameworks (preprocessing, Feature selection, Optimized deep learning, and prediction layers). Using comparison research, we compared various frameworks. The third frame outperformed the others under all situations, with an average accuracy of 100%, whereas the first and second frameworks achieved 94.97% and 93.12% accuracy, respectively. |