المستخلص: |
The aim of this study is to compare two learning models used for estimating software effort. The first model includes a layer, Gated Recurrent Unit (GRU), dropout layer, flatten layer and two dense layers. In contrast the other model is an Artificial Neural Network (ANN). These models will undergo testing, on three datasets. Desharnais, ISBSG10 and Finnish. To assess their performance. The evaluation will consider metrics such as Mean Absolute Error (MAE) Mean Squared Error (MSE) Root Mean Squared Error (RMSE) and R squared (R2). The result reveal that the first model exhibits performance across all datasets with R2 values and high MAE and RSMA values signifying its inefficacy. Similarly the second model demonstrates performance on the Finnish dataset with negative R2 values. From these findings it is apparent that hybrid deep learning models have limitations in software effort estimation, thus warranting research, in this domain.
|