Software Development Lifecycle (SDLC) is a process for developing high-quality software products that satisfy client needs. Software requirement detection and classification is one of the challenging and tedious tasks in SDLC as most of the time it is done manually. Thus, many software projects fail because of issues related to requirements. In this paper, we proposed a supervised machine learning (SML) approach to automate the detection and classification process of software requirements from Amharic textual documents such as meeting minutes, interview notes, requirements specifications, user guides, reports, and memos. Automating software requirements detection and classification process reduces development costs, time, personnel, and risk of software project failures. The datasets were prepared from five different software development companies in Addis Ababa, Ethiopia. We splitted the dataset into 80% training set and 20% testing set and both sets passed through three text preprocessing steps (tokenization, normalization, stemming), and then feature extraction is done using TF-IDF and word2vec. Then we tunned the five most common ML classifier algorithms (SVM, NB, KNN, LR, and DT) and we trained them. Then to compare different algorithms we used 10-fold cross-validation and we did experimentation for each model 10 times. Finally based on the 10-fold CV average value of accuracy, precision, recall, and f1- score we compared the performance of algorithms. An SVM model is the best performing model for classification with 95% accuracy, 87% precision, 89% f1-measure, and 88% recall. Keywords— Functional requirements, Non-functional requirements, Natural Language Processing, Machine Learning, Amharic requirement specification.