关键词: 物流, 经济, 线性回归, 神经网络

Abstract: In the contemporary context where globalization and informatization are continuously deepening, the logistics industry, as a core component of the modern economic system, plays a pivotal role in the development of regional economies. The efficient operation of the logistics system is not only crucial to the efficiency of the circulation of goods and services but also a key factor in enterprises' cost reduction and enhancement of market competitiveness. Therefore, an in-depth study of the interplay between regional logistics and the economy holds profound significance, both on a theoretical and practical level. Theoretically, this research contributes to the enrichment and refinement of the theoretical framework of logistics and regional economic development, offering new explanatory variables and analytical frameworks for regional economic theory. In practical terms, this study holds significant value for guiding regional logistics planning and policy formulation. It can provide a scientific basis for governments and relevant departments, assisting them in making more informed decisions in areas such as logistics infrastructure construction, logistics policy support, and regional economic development strategies.
In order to delve into this issue, this paper has selected Sichuan Province in China as the subject of study. Through the Sichuan Statistical Yearbook, a systematic collection of 28 indicators related to logistics and the economy has been conducted, spanning a period of 17 years from 2003 to 2019. The logistics indicators encompass five aspects: the scale of logistics demand, the length of transportation routes, the number of vehicles, the number of employees, and asset investment. This comprehensive dataset provides robust support for analyzing the interplay between logistics and the economy and for establishing predictive models.
During the initial data collection phase, we identify as many as 21 indicators related to logistics and seven related to the economy. Faced with such a multitude of indicators, directly constructing a predictive model would encounter issues such as an excessive number of variables and high model complexity, which would not only increase the computational burden of the model but also affect its practicality and interpretability. To address this issue, this paper employs a correlation coefficient matrix for indicator selection. By calculating the correlation coefficients between various indicators, we identify those that are highly correlated and selected eight indicators closely related to logistics and one closely related to the economy, laying a solid foundation for subsequent model construction.
After completing the indicator selection, we perform min-max normalization on the selected data to eliminate the impact of different indicators' units of measurement, ensuring the fairness and accuracy of the model. On this basis, we establish two predictive models: a linear regression predictive model and a neural network predictive model. The linear regression model, with its simplicity and ease of interpretation, is widely used in economic data analysis; the neural network model, with its strong nonlinear fitting ability, can capture more complex data relationships.
To verify the effectiveness of the models, we also construct linear regression predictive models and neural network predictive models without variable selection, as well as mainstream exponential smoothing and grey prediction methods, to forecast the regional gross domestic product. By comparing the mean squared error, mean absolute error, and determination coefficient of these six models, we find that the linear regression predictive model and the neural network predictive model after indicator selection have a clear advantage in prediction accuracy, indicating that our selection method not only reduces the complexity of the model but also effectively improves the model's predictive performance.
However, we are also aware that with the rapid development of the economy and the continuous changes in the logistics industry, historical data may not fully reflect the current and future actual situations. This means that our predictive models may have certain biases due to the limitations of the data. In addition, logistics and economic activities are complex systems with multiple dimensions and factors, and our research may only touch the tip of the iceberg, with many potential important influencing factors not yet fully explored or considered. Therefore, future research should pay more attention to the integration and analysis of multi-source data, trying to capture the interaction between logistics and the economy from a broader perspective. For example, data from fields such as supply chain management, international trade, and technological innovation can be considered to build more comprehensive and accurate predictive models. At the same time, with the development of big data and artificial intelligence technology, we can also use advanced algorithms such as machine learning and deep learning to improve the predictive ability and adaptability of the model.

Key words: logistics, economy, linear regression, neural network