Evaluation of Urban Low-Carbon Economy Development Using MachineLearning and Optimization Techniques

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

🔬研究者:This paper demonstrates the application of deep learning and optimization for carbon emission prediction, which can be extended to other contexts.

Introduction/Objective: The development of a low-carbon economy in urban areas has become a critical priority to address climate change and inefficient resource utilization. Since urban regions contribute significantly to carbon emissions, it is essential to develop an algorithm to assess the development of the low-carbon economy and support sustainable growth through advanced technologies such as artificial intelligence. Methods: The study adopts a data-driven predictive modeling approach using urban datasets from 2015 to 2023, including economic, energy, and emission indicators. The proposed approach employs an Intelligent Remora Optimized Deep Recurrent Neural Network (IRODRNN) for accurate prediction of carbon emissions. In addition, AHP and MCDA techniques are integrated to analyze the potential performance of cities implementing low-carbon strategies. results: The LC urban economies through AI address urgent global issues, such as environmental degradation and climate change. The development environment is an Intel core i7 CPU running Windows 11, Python 3.8, 32 GB of RAM, and a Jupyter Notebook for interactive scripting and data processing. It presents an innovative IRO-DRNN for carbon emissions prediction with better precision of such predictions. This will evaluate the urban LC development in terms of economic sustainability, energy efficiency, and carbon emissions. Based on data from 2015 to 2023, the proposed model makes a highly accurate prediction for carbon emissions with good reliability in the diverse performance metrics. Comparison of our proposed method with the existing approaches, such as secure Genetic Algorithm – Back propagation (GA-BP) [23], Autoregressive Integrated Moving Average - Back propagation Neural Network (ARIMA-BPNN) [24] and Cuckoo Search Optimization – Feed forward Learning Network (CSO-FLN) [25] in the metrics of Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-Square (R^2). 4.1. MAPE To assess the accuracy of a forecast model, one statistical metric is the MAPE. The MAPEamong the actual and anticipated values is calculated. It is generally used in time series forecasting and regression to evaluate model performance. MAPE can be used to assess the accuracy of models showing growth, energy use, and carbon emission reductions in response to sustainable planning techniques in cities developing low-carbon economies. The outcomes of MAPE are shown in Table 1 and Figure 3. Table 1: Computations of MAPE Methods MAPE GA-BP [23] 6.3 ARIMA-BPNN [24] 8.09 CSO-FLN [25] 1.656 IRO-DRNN [Proposed] 1.023 Figure 3: Comparisons of MAPE The IRO-DRNN method MAPE achieved 1.023, which is compared with the existing methods including GA-BP [23] achieved 6.3, ARIMA-BPNN [24] attained 8.09, and CSO-FLN [25] 1.656. The recommended methodology outperforms the existing approaches in terms of rates. 4.2. RMSE RMSE is a frequently used metric to assess the correctness of the model. Finding the square root of the average of the squared difference among the predict and real values is the calculation's task. A model's fit to the data cans be evaluated using the RMSE. A low RMSE indicates a high degree of forecasting accuracy. In the context of urban LC economic development, RMSE can be used to examine the accuracy of models that forecast the success of certain policies or programs meant to lower carbon emissions, while fostering economic growth. Figure 4 and Table 2 depicts the outcomes of RMSE. Table 2: Computations of RMSE Methods RMSE CSO-FLN [25] 11.915 IRO-DRNN [Proposed] 7.601 Figure 4: Comparisons of RMSE The IRO-DRNN method RMSE achieved 7.601, which is compared with the existing methods including CSO-FLN [25] 11.915. It illustrateshow the recommended methodology outperforms the existing approaches in terms of rates. 4.3. MAE An additional indicator of a model's prediction accuracy is MAE. Between actual and expected values, it calculates the average absolute difference. It is the difference between the expected and actual numbers, ignoring whether the mistakes are positive or negative. This metric is appealingsupportive, when analyzing models related to complex fields, such as urban LC economic development, in so thatexact predictions are fairly crucial for sustainable growth and environmental impact assessments. Figure 5 and Table 3 shows the results of MAE. Table 3: Computations of MAE Methods MAE CSO-FLN [25] 11.609 IRO-DRNN [Proposed] 7.152 Figure 5: Comparisons of MAE The IRO-DRNN method MAE achieved 7.152, which is compared with the existing methods including CSO-FLN [25] 11.609. The recommended methodology demonstrates how it outperforms the existing approaches in terms of rates. 4.4. R-Square The coefficient of determination, or R-squared, measures the extent to, which the independent variables in a regression model explain the variation in the dependent variable. It captures the fit of the model and is an important aspect of urban LC economic development research. In this regard, R2 can help evaluate how well factors, such as energy usage or transportation efficiency predict outcomes concerning sustainability and economic development in urban regions. Figure 6 and Table 4 displays the result for R2. Table 4: Computations of R2 Methods R2 GA-BP [23] 0.85 ARIMA-BPNN [24] 0.951 IRO-DRNN [Proposed] 0.978 Figure 6: Comparisons of R2 The IRO-DRNN method achieved 0.978R^2, which is compared with the existing methods including GA-BP [23] achieved 0.85, ARIMA-BPNN [24] attained 0.951. It illustrates how the recommended methodology outperforms the existing approaches in terms of rates. 4.5. Urban Low-Carbon Growth: 2015-2023 Trends The indicates that data shows that urban LC economy development levels have been trending upward from 2015 to 2023. Figure 7 shows the economic development results from 2015 to 2023. Results: The proposed model demonstrates superior predictive capability compared to existing methods. The performance improvement is statistically significant, with a Mean Absolute Percentage Error (MAPE) of 1.023 and an R-Squared (R²) value of 0.978. The model effectively captures complex relationships among economic sustainability, energy efficiency, and emission trends. discussion: The methods approximating GA-BP [23], ARIMA-BPNN [24], and CSO-FLN [25] have their respective drawbacks. GA-BP, which combines Genetic Algorithms (GA) with Back propagation (BP) Neural Networks, often faces challenges in terms of computational complexity and the risk of getting stuck in local optima due to the limitations of the GA in finding the global optimum. ARIMA-BPNN, which integrates Autoregressive Integrated Moving Average (ARIMA) with BP, struggles with over fitting, especially, when applied to highly non-linear data, and cannot efficiently handle large datasets. It is called CSO-FLN, combining Cuckoo Search Optimization with a Feed forward Learning Network. The disadvantage here might include slower convergence and sensitivity to parameter tuning thus less performance. To counter these problems, the developed method, IRO-DRNN has the following benefits. IRO-DRNN uses an improved optimization strategy that enhances the rate of convergence and accuracy as compared to GA and CSO-based methods for obtaining stability and better global search abilities. The proposed IRO-DRNN employs a deep recurrent architecture; it can very well capture the temporal dependency, therefore being able to express complicated patterns in time series more accurately than ARIMA-BPNN. IRO-DRNN can effectively manage larger datasets compared with its competitors, thus offering superior suitability for complex dynamic data analysis. Discussion: The integration of optimization techniques with deep learning enhances model accuracy and robustness, though results may vary depending on data quality and regional variability. Conclusion: The Intelligent Remora Optimized Deep Recurrent Neural Network (IRO-DRNN) provides a novel framework for evaluating urban sustainability, providing meaningful information to decision makers and urban planners and supporting future energy management strategies.

• openalex https://doi.org/10.2174/0126662558444991260427143655first seen 2026-05-31 05:20:44 · last seen 2026-06-03 04:58:42