Prognostic And Diagnostic Prediction Models Of Invasive Cancer In Women And Its Detection Accuracy Via Developing Machine Learning Techniques

Abstract

Cancer is a common disease, and it takes several human lives every year. It is generally classified on the basis of the tissue from where it originates, such as carcinoma or lymphoma. However, a single tumour may possess several types of heterogeneity. Cancer management is a tedious task due to its heterogeneity. Cancer causes >50% of the total deaths in developing countries due to poor diagnosis. Breast Cancer (BC) is the most heterogeneous cancer and massive molecular data is available on it; however, classification of the information is imperative for its management. Machine learning is an evolving tool that can be used to classify heterogeneous breast cancer datasets.

 Objective: Proper classification of tumour diversity and diagnosis of BC using ML approaches can improve the chances of survival; effective prognosis can help clinicians recommend the right treatment.

 Methods: Machine learning (ML) classifiers used for classification in the study are Classification and Regression Trees Naive Bayes Classifier, Artificial Neural Networks, Support–Vector Machines and Logistic Regression.

 Results: ML is the most efficient tool for classifying the diversity of breast cancer datasets existing with heterogeneity in risk factors. The experimental findings show that the logistic regression model gives the highest accuracy (96.60%), with a lower error rate than other models.

 Conclusion: ML appears to be a powerful and practical tool for the categorisation of breast cancer. Massive data concerning breast cancer is available, and crucial features can be extracted through ML. This research paper presents a review of machine learning tools used in the extraction of vital elements of tumour categorisation proposing a prognosis.