Software Development Effort Estimation Techniques Using Long Short Term Memory

• Published in: 2022 International Conference on Computer Science and Software Engineering (CSASE) pp. 182 - 187
• Main Authors: Ahmad, Farah B., Ibrahim, Laheeb M.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 15-03-2022
• Subjects: Analogy Based Estimation (ABE); Estimation; Forestry; General Linear Model (GLM); Long Short-Term Memory (LSTM); Machine Learning (ML); Machine learning algorithms; Measurement uncertainty; Prediction algorithms; Predictive models; Random Forest (RF); Software algorithms; Software Effort Estimation (SEE)
• DOI: 10.1109/CSASE51777.2022.9759751

The Software Effort Estimation (SEE) process is used to predict the effort involved in developing a software project inaccuracy. This process permeates the development stages of the software project. Although there are many models for estimating the effort, we will do our best to follow the best model for estimating the effort. The main objective of this research is to apply the Long Short-Term Memory (LSTM) algorithm and discover its accuracy in estimating the software effort. Then, compare the results with other models taken from previous work and find the best one among the models. The LSTM algorithm has been used with two data sets: China, which contains 499 projects, and Kitchenham, 145 projects. The metrics adopted to measure the accuracy of the models are Root Mean Squared Error (RMSE), respectively, Mean Absolute Error (MAE), and R_Squared. The results of RMSE, MAE, and R_Squared are 0.016, 0.019, and 0.972, respectively, in the china dataset. At the same time, the RMSE, MAE, and R_Squared results are also 0.017, 0.058, and 0.896, respectively, in the Kitchenham dataset. Therefore, the experiments' results showed the LSTM algorithm's superiority over the other algorithms in both data sets.