An Efficient Cancer Detection Model Using ML and Transfer Learning Techniques

Cancer, a pervasive and challenging medical condition, demands innovative approaches for early detection and intervention. This paper explores the complex domain of breast cancer detection, recognizing its crucial impact on improving patient outcomes. Two machine learning algorithms, namely SVM and RF, along with three transfer learning strategies employing distinct pre-trained CNN architectures (ResNet50, VGG16, and InceptionV3), were employed to augment the accuracy of cancer detection. The proposed algorithms are implemented with a Kaggle breast cancer dataset with three labels namely benign, malignant and normal. The transfer learning models are trained on the dataset, with evaluations performed using a validation set. For comparison, conventional ML classifiers and also applied for the cancer detection. The proposed transfer learning showcases a comprehensive approach to cancer detection by leveraging the strengths of multiple pre-trained CNN architectures. The proposed transfer learning given accuracy of 98.9% for breast cancer detection. The results of the experiments revealed that transfer learning outperformed the conventional methods for cancer detection.