The software project is managed by software managers who familiar with the project management knowledge areas. These software managers know everything intelligent means discerning or looking around and well informed by their professions. However, in the world especially in developing countries particularly in Ethiopia, the majority of software projects in software companies fail due to a lack of software project managers who are unfamiliar with the Project Management Knowledge Areas that are used without considering the company's situations and project contexts. This thesis aims to design a machine-learning model for the prediction of project management knowledge areas failure in software companies. Because software companies cannot consider any things as other related established or experienced business companies with different conditions. Like Experience, Money, Marketing and sales, Planning, Finding the right people, Team management, Scaling up, Holding competitors, Management, Mentorship, and others. This thesis followed experimental methodology using anaconda environment python programming Jupyter notebook. The machine-learning techniques that are suitable for our failed software project data and used the snowball sampling method by filling questionnaires to collect the data from software companies found in Bahir Dar and Addis Ababa cities. The study used about five machine learning algorithms among them SVM 92.13%, DT 89.88%, KNN 87.64%, LR 76.4%, and NB 65.16%. When we look at the results, we observed that SVM outperformed the other four machine learning methods. Because SVM works with categorical data, more effective high dimensional data and contains nonlinear transformations. We would like to recommend predicting the failure of predicting project management knowledge areas by collecting more failed project datasets from software companies using deep learning approaches to compare with our results.