The most dangerous and fatal illness is a brain tumor. Generally, disparate imaging methods, such as CT, MRI, and PET, are commonly used to assess a brain tumor. To analyze these tumors, magnetic resonance imaging is a commonly used imaging modality. In this work, n1 = 130 subjects of low-grade gliomas (benign), n2  = 200 subjects of glioblastoma multiforme (malignant), and n3  = 196 normal subjects are used. The deep convolutional neural network is used to identify brain tumors. The significant advantage of CNN is it can take in features from the given information automatically. The convolutional layers are utilized to convolve the inputs with weights to acquire a feature vector. The performance was determined using the CNN model, namely AlexNet. The effect of various optimization techniques to improve the result is also observed. Three optimizers were used such as ADAM, SGDM, and RMSprop, and detection accuracy of 98.1%, 92.5%, and 83.0% is achieved. The features are extracted using CNN, and tumor detection is done by using four supervised machine learning classifiers. The classifiers used are SVM, KNN classifier, Naïve Bayes classifier, and discriminant analysis. The accuracy acquired from the classifiers is 96.2%, 94.3%, 75.0%, and 96.2%, respectively. The proposed deep learning-based methods produce outstanding performance than conventional methods.