Combining Machine Learning Mechanism With Statistical Models For Predicting The Survival Of Lung Cancer

Abstract

Machine learning is a great artificial intelligence technology used in cancer prediction and diagnosis. Also, with the development of personalized and medical applications, there has recently been a rising tendency to machine learning (ML) techniques for prognostic predictions. However, developing accurate prediction systems of cancer mortality in medical environments remains a challenge.A regression model is being constructed to forecast lung cancer individuals' survival rates in months. It has previously been demonstrated that predictive analytics function well for brief overall survival of lower than 6 months; nevertheless, model performance decreases as greater survival times are attempted to be predicted. To estimate survival rates, regression systems are proposed in conjunction using a classification framework for this research. The Surveillance, Epidemiological, and End Results (SEER) dataset were used to gather a collection of de-identified lung cancer survivor data. To evaluate lung cancer information from the SEER study to construct effective lung cancer survival predictive design. Several qualities were removed/modified/split as a consequence of specially planned pre-processing processes, and two of the 11 resultant characteristics were shown to offer substantial predictive potential.ANOVA is used to choose a subset of components for the analyses. For classifying, a confusion matrix is used, and for regression, the Root Mean Square Error (RMSE) is used.For classifications, Random Forests (RF) were utilized, whereas, for regression, conventional Regression Analysis, Gradient Boosted Machines (GBM), and Random Forests have been used. The regression findings indicate that RF performs better for survival periods of and months on RMSE and , correspondingly, whereas GBM was quiet for survival periods of months on RMSE . Results comparison charts show that its regression analysis functions stronger for median survival durations than the RMSE estimates can reflect.