Acute Lymphoblastic Leukemia (ALL) is a highly malignant disorder that explicitly affects White Blood Cells (WBC) produced from lymphoid cells. Due to its potentially fatal nature, the prompt diagnosis of ALL is of utmost importance. At present, the domain of medical sciences is profoundly influenced by automated computer-assisted methodologies that are based on Artificial Intelligence (AI) and Deep Learning (DL). These methodologies serve as indispensable instruments for physicians in the expeditious identification of illnesses and also in reducing their burden. This paper proposed a transfer learning-based methodology where the last 4 convolution layers of VGG19 are fined-tuned. Next, the best hyperparameters of the fine-tuned VGG19 are evaluated using the Sine Cosine Algorithm (SCA). Further, a mixed dataset is proposed by combining a private dataset comprising 1000 smear images with a public dataset comprising 108 images. The overall scheme was experimented on the mixed dataset that achieved an accuracy of 98.18%. Moreover, the comparative studies establish the superiority of the proposed ALL scheme over comparing methods.