The Lung and colon cancers are the top causes of death and illness, so early detection is very important. Conventional testing methods have flaws, such as the need for high-end tools and the chance of mistakes between observers. Histopathology is the diagnostic gold standard because it offers cell-level pictures of tissue, but it is labor-intensive and requires highly trained pathologists. This paper introduces a novel approach for classifying histopathological lung images in order to detect cancer using transfer learning techniques. Utilizing EfficientNetB0 neural network architecture models to analyze input images is the methodology. These models extract pertinent image features that are then transmitted to a multi-class Support Vector Machine (SVM) for classification. Training and evaluation are performed with the LC25000 histopathology images dataset. This enables the extraction of informative features from histopathological images of the lungs. The suggested method is compared with a number of other classification methods, such as Artificial Neural Networks (ANN), Random Forest, K-Nearest Neighbours, Adaboost, and XGBoost, to see how well it works compared to current techniques. The results indicate that the EfficientNetB0 and SVM combination is effective. Notably, the model obtains an accuracy of 100% during training, 99.02% accuracy during validation, and 99.33% accuracy during testing. Additional indicators of performance, such as F1-score, Cohen kappa score, recall, and precision, demonstrate that the proposed method outperforms existing approaches. This study has the potential to considerably improve the early diagnosis and treatment of lung cancer, a crucial area of medicine.