Software Cost Estimation Using a Hybrid Deep Convolutional LSTM and Optimization Approach

Abstract

The importance of such software cost estimation (SCE) is that it can be relied upon in the planning and management of large software projects, especially when one is working in an environment that is dynamic and uncertain. The current models can be difficult to scale and model. In these respects, this paper will recommend a hybrid optimization–based deep convolutional LSTM (HO-DCLSTM) framework for solving an SCE. The proposed system takes the form of a deep learning model of a deep convolution LSTM to enhance cost estimates. The suggested solution combines a profound convolutional LSTM network and parameter refinement based on gazelle and Al-Biruni Earth radius optimization algorithms to increase the prediction accuracy of costs and reduce the computation load and volume. The suggested method generates a better solution, has a low cost of computing, and has the ability to locate the best solution. Three datasets are used to analyze the proposed method's performance based on NASA, COCOMO 81, and KEMERER data. The analysis of the results with these datasets shows that the proposed method has high performance with lower error rates and greater prediction accuracy when compared to other current methods.