In this age of social media, traditional news & media outlets are nearly outdated. When disasters strike, the influx of social media data rises significantly, making it vital to extract important information. Prompt response to tweets during emergencies is essential for saving lives and resources. This study explores the effectiveness of different classification algorithms in identifying and categorizing disaster-related tweets. We compare the performance of traditional machine learning models such as logistic regression, naive-bayes classifier, support vector machines, and complement-bayes classifier with the Recurrent Neural Networks and BERT Bidirectional Encoder Representations from Transformers classifiers. We evaluate these algorithms on their ability to classify tweets into two categories, i.e., whether a disaster has occurred or not. The classification performance is assessed using standard metrics like accuracy, precision, recall, and F1-score. Predominantly superior performance of BERT in relation to the traditional ML algorithms can be easily seen. Findings from this study provide valuable insights for developing robust and efficient systems for real-time disaster tweet classification, ultimately aiding emergency response efforts and saving lives.