Abstract:
Poverty alleviation is widely recognized as a fundamental pillar of human
development, closely linked to improvements in quality of life, social equity, and longterm economic resilience. Effective poverty reduction strategies require a
comprehensive understanding of the underlying factors contributing to poverty and
their dynamic interactions over time. This research proposes a hybrid poverty
prediction framework that integrates the Seasonal AutoRegressive Integrated Moving
Average with eXogenous (SARIMAX) model with Long Short-Term Memory (LSTM)
neural networks. The SARIMAX model captures linear patterns, while the LSTM
model learns complex nonlinear relationships. By combining these complementary
approaches, the Hybrid SARIMAX–LSTM model aims to enhance predictive
performance and robustness. The empirical analysis is conducted on the dataset from
the U.S. Census Bureau covering the period from 2013 to 2024 using different sets of
features, these are selected based on observations, Random Forest Regression, Linear
Regression and three specified sets. The performance of SARIMAX, LSTM, and
Hybrid SARIMAX–LSTM models are evaluated and compared using common error
metrics. The proposed hybrid model demonstrates improved predictive performance in
terms of across the different feature sets and domains. The proposed hybrid model
recorded consistently achieves the lowest mean absolute error (MAE) across all feature
configurations, outperforming both SARIMAX and LSTM models. Particularly,
Random Forest regression feature set the best overall performance with an MAE of
0.1301 improved predictive performance in terms of across the different feature sets
and domains. These results indicate that the hybrid SARIMAX–LSTM model produces
smaller average absolute prediction errors and the proposed poverty prediction system
can support governments, non-governmental organizations (NGOs), researchers, and
other relevant institutions to identify vulnerable populations, allocate resources
effectively, and develop targeted strategies to poverty alleviation.