سامانه همایش های دانشگاه آزاد اسلامی واحد نجف آباد

صفحه اصلی / سومین کنفرانس بین المللی مهندسی و علوم کامپیوتر و اولین کنفرانس ملی سامانه های هوشمند، کلان داده و رایانش پیشرفته

Forecasting Budget of City-console Using LSTM Neural Networks and Bayesian Optimization: A Case Study

نویسندگان :

Hasan Mohammadi najafabadi¹ Hamid Rastegari²

1- Najafabad Azad University 2- Najafabad Azad University

کلمات کلیدی :

LSTM،Bayesian optimization،city-console budget forecasting،deep learning،time-series prediction،smart cities

چکیده :

Abstract—Accurate forecasting of city-console budget headings is essential for maintaining fiscal stability and supporting data-driven urban governance. This paper proposes a deep learning framework based on a Bayesian-optimized Long Short-Term Memory (LSTM) network to model nonlinear interactions between financial and environmental indicators affecting municipal revenues. Using monthly data from the Najafabad city-console over the period 2011–2023, the model incorporates dynamically weighted features, including inflation rate, added value, tax-related components, population growth, and pollution effects. Bayesian optimization is employed to tune key hyperparameters and feature weights, leading to a more robust representation of temporal dependencies under economic volatility. Experimental results show that the proposed model achieves a Root Mean Squared Error (RMSE) of 0.0051 and a Mean Absolute Error (MAE) of 0.0038, outperforming ARIMA, RNN, GRU, and CNN-LSTM baselines. The findings highlight the potential of Bayesian-optimized LSTM architectures as practical decision-support tools for city-console budget planning in emerging economies.