Probabilistic Artificial Neural Networks for Malignant Melanoma Prognosis

摘要: Artificial Neural networks (ANNs) have found applications in a wide variety of medical problems and proved successful for non-linear regression classification. This paper details novel flexible probabilistic ANN model the prediction conditional survival probability malignant melanoma patients. Hazard density functions are also estimated. The is trained using log-likelihood function, generalisation has been addressed. Unrestricted by assumptions that unrealistic or parametric forms difficult to justify, thereby attains advantage over traditional statistical models. Furthermore, an estimate variance-covariance matrix obtained asymptotic Fisher information matrix. Implemented Excel® spreadsheet, model’s user-friendly design further adds its flexibility, with much potential use statisticians as well researchers.