Software cost estimation predicts the amount of effort and development time required to build the system. Instead of just putting values into giving equation to calculate the cost and effort, we require more work on a scale and cost drivers to increase the accuracy of software cost estimation. The Software cost estimation process depends on the attributes such as peoples working in teams, programming language used and software tools used, salaries and overhead costs associated with the development team, database size used, training cost, accidental rework, a policy used in an organization, cost of shared facilities such as a library, restaurant, resources used such as light, network etc, which gives a clear idea about software cost estimation. There are so many models available categorized into algorithmic and non-algorithmic model each of their strengths and weakness. We propose a hybrid approach, which consists of Functional Link Artificial Neural Network (FLANN) and COCOMO-II with training algorithm. FLANN reduces the computational complexity in multilayer neural network. It does not have any hidden layer, and it has fast learning ability.