Identifying the exact location of intracranial tumors is a task expected from a radiologist in order to help the neurosurgeon in surgical planning and/or the radiation oncologist in ensuring targeted radiation therapy. Since an MRI image consists of many 3D slices, that could take several hours before the radiologist can finish manually segmenting the entire tumor. The process is dependent on the experience and subjective decision-making of the radiologist. Moreover, manual segmentation is vulnerable to inter-observer variability as assessment is subject to variation of interpretation between different radiologists. It is also prone to intra-observer variability as it is likely that the same radiologist may have a slightly different impression of the same MRI image which he read previously. Thus, there is a need for an automated intracranial tumor segmentation clinical decision support tool that is devoid of subjective interpretation as this could lessen the time needed to correctly identify the exact location of the intracranial tumors and can also help the radiologist in the decision-making process. The main intention of this study is to perform intracranial tumor analysis which consisted of two major steps: (1) tumor classification, and (2) tumor segmentation applied to a publicly available Brain MRI Dataset. After preprocessing, classification models were constructed to determine if a given intracranial MRI image has tumor or not. Once tumor is detected, the segmentation model will then determine the exact location of the tumor by overlaying the predicted tumor mask on the Cranial MRI image. Various deep learning techniques were used for classification and segmentation: ResNet50, and EfficientNetB1 for classification while U-Net and ResUNet for segmentation. The best performing classification model was achieved by using ResNet50 (96% average sensitivity, 99% average specificity, 99% average precision, 97% average F1-score, 0.963 Matthews Correlation Coefficient) while the best performing segmentation model was achieved using ResUNet (91% average Dice similarity score, 85% average Intersection over Union, 93% average Tversky Index, 95% average sensitivity, and 99% average specificity).