The social media generates large volume of data through tweets and text messages during and after any disaster. The analysis and classification of the obtained data at the time of disaster is essential for conveying the information to the appropriate rescue personnel. In this paper, an automated text classification system is proposed in order to classify the data effectively. The classification of the tweets related to disaster is a challenging task as the texts are not correctly written or do not convey the exact meaning since people send the text messages in panic situations. A manual vocabulary has been created by considering the nature of the disaster data. The created vocabulary is used for splitting the tweets into various categories. In the categorized data, popular statistical feature selection methods like, term frequency, Chi Square are used in combination with Support Vector Machine (SVM) and Naïve Bayes algorithms to classify the data. The results reveal that SVM performs better than Multinomial and Bernoulli Naïve Bayes for all the classes of disaster related data.