GTAP Resources: Resource Display
| GTAP Resource #7530 |
|---|
|
"Expanding capabilities for policy evaluation through deep learning emulation of CGE models" by Nechifor, Victor, Mohammed Basheer and Emanuele Ferrari Abstract The study explores the potential of deep learning (DL) models, particularly Artificial Neural Networks (ANNs), to emulate complex structural models such as computable general equilibrium (CGE) models. Despite the promising capabilities of DL models to capture non-linear relationships in high-dimensional data, their application in economy-wide modeling remains limited. This research aims to address this gap by demonstrating the reliability of ANNs in emulating CGE models, using the DEMETRA model—a single-country CGE model for the Kenyan economy—as a case study. Key benefits of using DL emulators include enabling non-technical users to interact with complex models, promoting participatory modeling, reducing costs, enhancing computational efficiency, and facilitating large-scale sensitivity analyses. The study employs a fine-tuning framework using multi-objective evolutionary algorithms (MOEA) to optimize the hyperparameters of ANNs, balancing model performance (R2 on training and testing sets) and overall inference time to avoid overfitting and achieve efficient configurations. The research involves generating extensive training data through Latin Hypercube sampling, running the DEMETRA model on a cluster computing facility, and aims to integrate behavioral parameters (elasticities) as inputs for faster Monte Carlo analyses. Ultimately, the study introduces an interactive tool for policy analysis, allowing users to explore various policy scenarios within predefined ranges, thereby enhancing decision-making processes. |
| Resource Details (Export Citation) | GTAP Keywords | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
- Model validation and sensitivity analysis - Software and modeling tools - Africa (East) |
| Attachments |
|---|
|
If you have trouble accessing any of the attachments below due to disability, please contact the authors listed above.
Public Access  Paper (635.0 KB) Replicated: 0 time(s)Restricted Access No documents have been attached. Special Instructions No instructions have been specified. |
| Comments (0 posted) |
|---|
|
You must log in before entering comments.
No comments have been posted. |
Last Modified: 9/15/2023 2:05:45 PM
