Abstract
software effort estimation is the first step in every software development process. The accuracy of the software effort estimation can seriously affect the success of projects. Over estimating will result in allocating excess and unneeded resources that will cost the company unexpected expenses, while under estimation will lead to delays in software delivery. The dynamic and evolving nature of the software development process makes it challenging to build a reliable effort estimation system. One of the common models used for an efficient software cost estimation is the Constructive Cost Model (COCOMO) model. The COCOMO model is flexible, it is used in different environments. In this paper we aim to explore non-algorithmic approaches to effort estimation that attempts to improve upon the COCOMO model. We will focus on machine learning based methods trained on a public dataset. Furthermore, we will present some of the datasets available to use in the software development process and how they can be utilized to train machine learning models.