Precise and well-timed diagnosis of the fatal types of cancers such as lung and colon cancers plays a crucial role to advance patient outcome and decreasing death rates. With the upcoming display of Machine Learning techniques, there is an unexpected opportunity to improve the diagnostic accuracy and successfulness of cancer prediction from medical imaging. The proposed work utilizes Bayesian Optimization based deep learning model (BO-DL) using Convolutional Neural Network in lung and colon cancer diagnosis using histopathological images as input. The proposed method is compared with existing methods using various performance metrics. Results reveal that the accuracy of diagnosis for lung and colon cancer is better compared to existing methods.