Osteosarcoma is a rare, high-grade bone sarcoma that contributes to 28% of primary bone cancers in adolescents, as well as up to 56% in children and teens. Histology continues to be the preferred method for assessing a disease's stage, but additional advancements are required because early detection and expensive treatment are still inadequate. In recent years, automated analysis approaches have become increasingly popular in the field of osteosarcoma cancer diagnosis for microscopic image assessment. Researchers have utilised Deep Learning techniques, mostly convolutional neural networks (CNNs), in the diagnosis of osteosarcoma. The aim of the research is to develop a novel approach for osteosarcoma diagnostic prediction models. The effectiveness of ensemble learning and transfer learning techniques is investigated and evaluated. Transfer learning techniques and customCNN are employed on a publicly accessible histopathological image data set for classifying osteosarcoma as a viable, non-tumour, or non-viable tumour. Furthermore, fine-tuning transfer learning approaches are applied along with ensemble learning classifiers. The ensemble model performs better as compared to other pretrained and customCNN models. The outcome achieved surpasses existing approaches, which makes it suitable for clinical settings and as a load-reducing tool for medical professionals.