An Improved CHI2 Feature Selection Based a Two-Stage Prediction of Comorbid Cancer Patient Survivability

There are theoretical and practical ramifications to modelling cancer patients' survival with concurrent illnesses. Cancer is one of the leading causes of mortality worldwide. Stomach, liver, thyroid, lungs, and skin cancers are a few of the more common types. The early identification and prevention of these malignancies are important goals. Recent investigations have found that some patients suffer cancer-related co-morbidities. Studies show that comorbid conditions worsen the prognosis of cancer patients. There are several methods that might have led to this finding. With hazard ratios ranging from 1.1 to 5.8, the majority of studies discovered that cancer patients with comorbidity had a poorer 5-year survival rate than those without. Just a few research have examined the effects of certain chronic conditions. There is no proof that comorbidity causes more aggressive cancers. Our research indicates that forecasting survival is a two-stage issue. Predicting a patient's five-year survival rate is the initial step. In the second phase, those whose expected outcome is “death” are told how long they have left to live. Male and female concurrent cancer cases were identified and categorised using the SEER database (Stomach, Lung, Liver, Thyroid and Skin Cancers). The dataset was handled throughout the classification phase using CHI2-based feature selection. These two techniques addressed the issues that an inconsistent data set raised.