Blood cell cancer prediction becomes more accurate when young white blood cells are separated from red blood cells. The only method to predict a blood disorder involves taking photographs of the skin, then calculating, shading, and measuring the images. This study introduces a new deep learning (DL) model that can accurately classify and predict blood cell cancer. Our method performs very well for finding malignant cases by using a modified EfficientNetV2B3 with either a Multilayer Perceptron (MLP) classifier or a Support Vector Machine (SVM). The SVM-based model outperforms the MLP-based model, with metrics like an accuracy of 99.2%, a recall of 99.2%, an F1-score of 99.2%, and an area under the receiver operating curve (AUC) of 99.5%. These results demonstrate that our DL model could be a useful tool for doctors to use in making accurate cancer diagnoses, and the study adds to the growing body of research on DL in cancer prediction.