Scaling Approach Over Learning Layer of Deep Learning Model to Reduce the FALSE Error in Binary Classification

Objective: Worldwide Breast cancer is a main health issue for the women, every year thousands of deaths occurring due to late detection and diagnosis. Computer-aided diagnostic detection systems are becoming a useful tool for early cancer cell identification. The proposed work uses the Deep Learning, Convolutional Neural Network (CNN), model to identify the cancerous cells in medical imaging accurately. Method: The proposed work balances the Learning Layers (LL) of CNN model to find the finest binary classification to detect whether the given image is Malignant or Benign. Experiment is carried out on gradual scaling of LL from 1 to 5 with learning cycles 10,15 and 20 respectively for each stage of scale to reduce the False Positive (FP), False Negative(FN) errors in each layer. Result: The performance of the scaled LL is evaluated using the metric like residual error loss, training accuracy in training stage. The model is evaluated with precision, recall, accuracy and $\mathbf{f 1}$-score in the testing stage. By leveraging the power of deep learning, the system achieved an impressive accuracy rate of 97% in differentiating between the two common types of cancer cells in CNN-LL5.