In any product development cycle, costs can increase when a project takes longer than anticipated. Because accurately estimating the completion date of a project is not easy. The risk remains large even in Agile Scrum, where the project is planned and run in short iterations. Machine learning can be essential in planning and estimating the project schedule to estimate user story efforts. This paper is an effort in that direction, where the effectiveness of Extreme Learning Machines (ELM) in the domain of predicting the effort estimate of user stories (multi-class text classification domain) is studied and compared with some of the existing techniques like Support Vector Machine (SVM) and Logistic Regression (LR). In this paper, the focus is on highlighting the performance of ELM in the field of multi-class text classification; results from other models are studied and analysed. Some standard techniques are investigated to improve the accuracy of models, such as feature selection and parameter tuning.