Natural Language Processing (NLP) is one of the greatest familiar field of Artificial Intelligence. Bangla text classification is become a significant problem due to the availability of textual documents. So, it's high time to solve this problem for better information organization and knowledge management. This study identifies the type of Bangla sentence from a human story: exclamatory or interrogative text document. Several machine learning approaches are applying to the dataset and identify the sentence from the text document. Regulating data is collected from various platforms such as story, Bengali's blog, conversation, etc. To get the output, we need to go through hard work as data cleaning, preprocessing, tokenized by the Countervector, etc. After processing data for the feeding model, we conduct various classification approaches to predict the sentence and provide high accuracy for deciding the result. Among all the approaches, Random Forest and XGBoost provide the highest accuracy of 96.39% in parallel. This study also provides the advantage of sentimental analysis by detecting the type of document in further implementation.