International Journal of Engineering Technology and Management Sciences Website: ijetms.in Issue: 2 Volume No.7 March - April – 2023 DOI:10.46647/ijetms.2023.v07i02.011 ISSN: 2581-4621 @2023, IJETMS | Impact Factor Value: 5.672 | Page 87 A Novel Brain Tumor Classification Model Using Machine Learning Techniques Saikat Sundar Pal 1 , Prithwish Raymahapatra 2, Soumyadeep Paul 3 , Sajal Dolui 4, Dr. Avijit Kumar Chaudhuri 5 , Sulekha Das 6 1,2,3,4 UG - Computer Science and Engineering, Techno Engineering College Banipur, Habra, Kolkata 5,6 Assistant Professor, Computer Science and Engineering, Techno Engineering College Banipur, Habra, Kolkata 1 0000-0002-1042-622X, 2 0000-0001-7147-9491, 3 0000-0003-2188-4469, 4 0000-0002-1005-9765, 5 0000-0002-5310-3180, 6 0000-0002-6641-3268 Abstract The objective of this research work is to classify brain tumor images into 4 different classes by using Convolutional Neural Network (CNN) algorithm i.e. a deep learning method with VGG16 architecture. The four classes are pituitary, glioma, meningioma, and no tumor. The dataset used for this research is a publicly available MRI Image dataset of brain tumor with 7023 images. The methodology followed in this project includes data pre-processing, model building, and evaluation. The dataset is pre-processed by resizing the images to 64x64 and normalizing the pixel values. The VGG16 architecture is used to build the CNN model, and it is trained on the pre-processed data for 10 epochs with a batch size of 64. The model is evaluated using the area under the operating characteristic curve (AUC) metric of the receiver. The results of this project show that the CNN model with VGG16 architecture achieves an AUC of 0.92 for classifying brain tumor images into four different classes. The model performs best for classifying meningioma with AUC of 0.90, followed by pituitary with AUC of 0.91, glioma with AUC of 0.93, and no tumor with AUC of 0.89. In conclusion, the CNN model with VGG16 architecture is an effective approach for classifying brain tumor images into multiple classes. The model achieves high accuracy in identifying different types of brain tumors, which could potentially aid in early diagnosis and treatment of brain tumors. Keywords - CNN, VGG16, AUC, Brain Tumor Introduction: Brain tumor image classification using convolutional neural network (CNN) is a challenging task in medical image analysis. In this project, we will use the VGG16 model, which is a widely used deep learning architecture for image classification tasks. The goal of this project is to classify brain tumor images into four different classes: pituitary, glioma, meningioma, and no tumor. We will use the Area Under the Operating Characteristic Curve (AUC) of the receiver as the evaluation metric for our model. AUC, or Area Under the Curve, is an important metric used in machine learning(ML) to evaluate the performances of binary classification models. It measures the performance of the given model by distinguishing between positive and negative samples. In a problem that is binary classified, the model predicts a probability score for each sample, and the AUC represents the probability that the model will rank a positive sample that is randomly chosen higher than a negative sample that is also randomly chosen. The AUC ranges from 0 to 1, where an AUC of 0.5 indicates that the model is not better than random guessing, while an AUC whose value is 1, indicates a perfect model. AUC is a popular metric because it is more robust than accuracy when dealing with imbalanced datasets. It also provides a comprehensive evaluation of the model's performance, taking into account both false positive and false negative rates. Furthermore, AUC is useful for comparing the performance of different models, as it is independent of the classification threshold used to make predictions. This allows for a fair comparison of models even if they have different thresholds.