Urban texture analysis and prediction method based on road network

Abstract

The invention discloses a city texture analysis and prediction method based on a road network, which is used for quantifying the influence of government policies on city development and predicting the future space-time evolution of a city. According to the method, the concept of city texture is introduced, road network data and a space-time analysis technology are combined, and the dynamic characteristics of city development can be accurately captured. According to the method, the neural network and the linear prediction model are utilized to perform quantitative analysis on the development levels of different regions of the city, identify the specific influence of policy intervention on city development, perform scientific prediction on future city expansion and infrastructure construction, and provide powerful support for city planning and policy making.

Description

Technical Field

[0001] The present invention belongs to the technical field of urban planning and development, and specifically is a development prediction method based on road network analysis and policy impact quantification. Background Art

[0002] With the acceleration of global urbanization, cities have become an important carrier of social and economic development. However, urbanization not only brings about the concentration of population and resources, but also brings many challenges, such as urban spatial expansion, infrastructure construction, social inequality, and environmental sustainability. In this context, how to scientifically plan and predict urban development has become a key issue that urban planners and policymakers need to solve.

[0003] Traditional urban studies have explored the driving forces of urbanization from the perspectives of socio-economics, environment, and technological development, but these studies often overlook the important role of government policies in urbanization. Over the past few decades, China has experienced rapid urbanization, with dramatic changes in urban scale and spatial pattern. This process is not only driven by economic development, but also inseparable from the guidance of government policies. For example, the role of urban planning policies, infrastructure construction policies, and major events in promoting urban development cannot be ignored.

[0004] However, existing urban research methods have limitations when analyzing the impact of policies on urban development. On the one hand, traditional urban texture analysis relies on static data and is difficult to capture the dynamic changes in urban development; on the other hand, existing methods lack systematicity and precision in quantifying policy impacts. In addition, as cities expand in size and complexity, traditional linear prediction methods can no longer meet the needs of accurate prediction of urban development.

[0005] In recent years, with the development of big data, geographic information systems (GIS) and artificial intelligence technologies, urban research has ushered in new opportunities. Neural network and machine learning technologies make it possible to process complex spatiotemporal data and more accurately simulate and predict the dynamic process of urban development. For example, the emergence of the Transformer architecture provides a powerful tool for processing sequence data, which can capture the spatiotemporal characteristics of urban road networks and provide new ideas for urban development prediction.

[0006] In China, urbanization is not only an economic and demographic issue, but also a policy issue. The Chinese government has guided urbanization through a series of policies, such as the "village-to-village" policy, urban functional zoning planning, and green belt policy. These policies have different impacts in different regions, resulting in significant regional differences in urban development.

[0007] Despite this, there are still deficiencies in the current research on how policies affect urban development. Existing research focuses on the qualitative analysis of policies and lacks quantitative evaluation of policy impacts. In addition, predictions of urban development are mostly based on linear extrapolation of historical data, ignoring the changes in urban development paths caused by policy interventions. Therefore, it is of great theoretical and practical significance to develop a method that can quantify policy impacts and combine spatiotemporal data to predict urban development.

[0008] This paper aims to fill this gap and proposes an urban texture analysis method based on road networks, combining neural networks and linear prediction models to quantify the impact of policies on urban development and predict the future spatiotemporal evolution of cities. By introducing the concept of urban texture, this paper can capture the historical traces and spatial characteristics of urban development and provide a scientific basis for urban planning and policy formulation. Summary of the invention

[0009] The present invention provides a road network-based urban texture analysis and prediction method for quantifying the impact of government policies on urban development and predicting the future spatiotemporal evolution of cities. This method can accurately capture the dynamic characteristics of urban development by introducing the concept of "urban texture" and combining road network data with spatiotemporal analysis technology. The present invention uses neural networks and linear prediction models to quantitatively analyze the development levels of different areas of a city, identify the specific impact of policy interventions on urban development, and scientifically predict future urban expansion and infrastructure construction, providing strong support for urban planning and policy formulation.

[0010] The basic principle of the present invention is to regard the city as a dynamic organism, whose development process can be reflected by the evolution of the road network. The urban texture is similar to the annual rings of trees, which records the historical traces and spatial characteristics of urban development. By analyzing the spatiotemporal changes of the road network, the dynamic process of urbanization can be revealed, and the regulatory effect of policies on urban development can be quantified. The present invention uses the Transformer architecture in the neural network to process the sequence data of the road network. At the same time, combined with the linear prediction model, the difference between the actual development and the natural development path is compared to identify the impact area of ​​policy intervention. In addition, the present invention also introduces the Manhattan distance formula and the quadratic polynomial model to set the threshold of urban development and predict future trends, so as to ensure the accuracy and practicality of the prediction results.

[0011] A method and prediction model for quantifying the impact of government policies on urbanization process, comprising the following steps:

[0012] Step 1: Collect urban road data. The data set includes road length data.

[0013] Step 2: Preprocess the collected data. Deleting abnormal data includes deleting empty data, attributes irrelevant to this method, abnormal road length data, and various invalid road data.

[0014] The data was then imported into ArcGIS Pro for topological analysis and road network segmentation, dividing the city into 5kmx5km grids to generate urban texture data. These data reflect the spatiotemporal development characteristics of the city and serve as the basis for subsequent analysis.

[0015] Step 3: Spatiotemporal data analysis and model building:

[0016] The Transformer neural network architecture is used to process the spatiotemporal sequence data of the road network. Through the encoder-decoder structure and the self-attention mechanism, the changing laws of the road network are learned and future development trends are predicted.

[0017] The Transformer loss function uses the MSE loss function, which is calculated as follows:

[0018]

[0019] in:

[0020] MSE stands for mean square error, which is used to measure the difference between the predicted value and the actual value of road data.

[0021] n represents the number of road data samples.

[0022] y i Represents the actual value of the i-th road data sample.

[0023] Represents the predicted value of the i-th road data sample.

[0024] At the same time, a linear prediction model is constructed to compare the differences between the natural development path and the actual development path, so as to identify the impact areas of policy intervention. The calculation formula is:

[0025] y=mx+c (2)

[0026] in:

[0027] y represents the predicted value of road data.

[0028] x represents the road data input value.

[0029] m represents the slope, which reflects the rate of change of y with x.

[0030] b represents the intercept, the value of y when x=0.

[0031] Then, by comparing the linear prediction and Transformer prediction parts, we can identify the areas affected by policies in urban development. The formula is:

[0032] ΔR dif1 =R transformer -R true (3)

[0033] ΔR dif2 =R linear -R true (4)

[0034] in:

[0035] ΔR dif1 Indicates the difference between the road value predicted by the Transformer model and the actual value.

[0036] ΔR dif2 Represents the difference between the linear model road prediction value and the actual road value.

[0037] R transformer Represents the predicted road value of the Transformer model.

[0038] R linear Represents the predicted road value of the linear model.

[0039] R true Indicates the actual road value.

[0040] Step 4: Set the threshold of urban development through the Manhattan distance formula to limit the overdevelopment of the city's peripheral areas. Combine the quadratic polynomial model to predict the future development trend of the city, comprehensively consider historical data and policy orientation, and provide a scientific basis for urban planning.

[0041] The Manhattan formula is as follows:

[0042] d=|x1-x2|+|y1-y2| (5)

[0043] in:

[0044] d represents the Manhattan distance.

[0045] x i and i Respectively represent the coordinates of the two intersection data in the i-th dimension.

[0046] The quadratic polynomial model is as follows:

[0047]

[0048] in:

[0049] y represents the predicted value of road data.

[0050] x1 represents the length of input road data 1.

[0051] x2 represents the length of input road data 2.

[0052] t represents the prediction time minus the road data 1 time.

[0053] a, b and c are model parameters obtained by fitting the road history data.

[0054] Step 5: Verify the accuracy of the model by comparing the actual data with the predicted results. The present invention has been experimentally verified on the road network data of Beijing from 2005 to 2022, and the prediction accuracy has reached more than 85%. This method can be widely used in urban planning, policy evaluation and sustainable development research, providing decision support for urban managers.

[0055] The traditional urban development model mainly relies on linear prediction and simple statistical methods. These methods are often unable to cope with complex spatiotemporal data and are difficult to accurately capture the dynamic changes of urban development. The present invention, by introducing the Transformer neural network architecture, can effectively process the spatiotemporal series data of the urban road network and accurately predict the future development trend of the city. In addition, the present invention combines the linear prediction model and can accurately identify the impact areas of policy intervention by comparing the differences between the natural development path and the actual development path, providing a more scientific basis for policy evaluation. The present invention is not only superior to the existing technology in terms of prediction accuracy, but also provides a more effective solution in terms of policy impact identification and urban development control. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a flow chart of Beijing road data processing provided by the present invention.

[0057] Figure 2 This is a graph of the evolution of road lengths in Beijing provided by the present invention.

[0058] Figure 3 This is the Beijing road data development map provided by the present invention.

[0059] Figure 4 The 2022 prediction verification map and 2030 prediction map of Beijing roads provided by the present invention. (a) Comparison between the predicted 2022 urban road data and the actual data; (b) Forecast of Beijing development in 2030.

[0060] Figure 5 Flow chart for implementing the present invention. DETAILED DESCRIPTION

[0061] The present invention will be explained and described below in conjunction with the accompanying drawings:

[0062] The basic principle of the present invention is to regard the city as a dynamic organism, whose development process can be reflected by the evolution of the road network. The urban texture is similar to the annual rings of trees, which records the historical traces and spatial characteristics of urban development. By analyzing the spatiotemporal changes of the road network, the dynamic process of urbanization can be revealed, and the regulatory effect of policies on urban development can be quantified. The present invention uses the Transformer architecture in the neural network to process the sequence data of the road network. At the same time, combined with the linear prediction model, the difference between the actual development and the natural development path is compared to identify the impact area of ​​policy intervention. In addition, the present invention also introduces the Manhattan distance formula and the quadratic polynomial model to set the threshold of urban development and predict future trends, so as to ensure the accuracy and practicality of the prediction results.

[0063] The data set used in the present invention is the road data of Beijing in 2005, 2015 and 2022. The method and prediction model for quantifying the impact of government policies on the urbanization process include the following steps.

[0064] Step 1: First, collect the road network data of Beijing in 2005, 2015 and 2022, including the geometric information, length and type of roads. Divide the city into 5km×5km grids, and import the road network data into ArcGIS Pro for topological analysis, fix potential topological errors, and ensure the integrity and accuracy of the data. Subsequently, segment the road network, mark the start and end nodes of each line segment, and calculate the total length of the road in each grid to generate urban texture data.

[0065] Step 2: Use the Transformer architecture to build a model, including an encoder and a decoder. The encoder processes the input spatiotemporal sequence data (road length) through a self-attention mechanism, and the decoder predicts future development trends based on the encoder's output. Use the road data from 2005 and 2015 as training input, the data from 2022 as the target output, and optimize the model parameters through the mean square error (MSE) loss function. After training, apply the model to the data from 2022, compare the predicted results with the actual data, and verify the accuracy and reliability of the model.

[0066] Step 3: Construct a linear prediction model to predict the road length in 2022 based on the road data in 2005 and 2015. By calculating the difference between the linear prediction results and the actual data, the areas where policy interventions affect urban development are identified. For example, if the actual road length in a certain area is significantly higher than the linear prediction value, it indicates that the area may be actively promoted by the policy. This method can quantify the specific impact of policies on urban development.

[0067] Step 4: Based on the maturity of the road network in the central area (such as within the Fourth Ring Road of Beijing), the Manhattan distance formula is used to set the development threshold of the peripheral area to limit overdevelopment. Combined with the quadratic polynomial model, historical data and time factors are considered to predict the road length and development trend in the future (such as 2030). This method provides a scientific basis for urban planning, avoids resource waste and disorderly expansion, and ensures the sustainability of urban development.

[0068] Step 5: Verify the accuracy of the model by comparing actual data with the prediction results. Experiments show that the prediction accuracy of the present invention reaches 88% in the peripheral areas of Beijing and 85% in the central urban area. This method can not only quantify the specific impact of policies on urban development, but also provide scientific guidance for future urban planning. It has broad application prospects, such as urban infrastructure planning, land use planning, and policy evaluation.

[0069] Based on urban texture analysis and spatiotemporal data modeling, the present invention realizes dynamic monitoring and accurate prediction of urban development by introducing Transformer neural network and linear prediction model. The present invention can not only quantify the specific impact of government policies on the urbanization process, but also provide a scientific basis for urban planning by setting development thresholds and predicting future trends, thereby avoiding disorderly expansion and waste of resources. In addition, the present invention has demonstrated high-precision prediction capabilities in data verification in Beijing from 2005 to 2022, and has broad application prospects. It can provide strong support for urban infrastructure planning, land use optimization, and policy evaluation, and help achieve urban sustainable development goals.

Claims

1. A method for analyzing and predicting urban texture based on road network, characterized in that: The following steps are involved: Step 1: Collect urban road data. The data set includes road length data. Step 2: Preprocess the collected data. Import the road data into ArcGIS Pro for topological analysis and road network segmentation, divide the city into 5kmx5km grids, and generate urban texture data. These data reflect the spatiotemporal development characteristics of the city and serve as the basis for subsequent analysis. Step 3: Spatiotemporal data analysis and model building: The Transformer neural network architecture is used to process the spatiotemporal sequence data of the road network. Through the encoder-decoder structure and the self-attention mechanism, the changing laws of the road network are learned and future development trends are predicted. A linear prediction model is constructed to compare the differences between the natural development path and the actual development path, so as to identify the impact areas of policy intervention. Step 4: Set the threshold of urban development through the Manhattan distance formula to limit the overdevelopment of the city's peripheral areas. Combine the quadratic polynomial model to predict the future development trend of the city, comprehensively consider historical data and policy orientation, and provide a basis for urban planning. Step 5: Verify accuracy by comparing actual data with predicted results.

2. The urban texture analysis and prediction method based on road network according to claim 1 is characterized in that: The Transformer loss function uses the MSE loss function, which is calculated as follows: in: MSE stands for mean square error, which is used to measure the difference between the predicted value and the actual value of road data. n represents the number of road data samples. y i Represents the actual value of the i-th road data sample. Represents the predicted value of the i-th road data sample.

3. The urban texture analysis and prediction method based on road network according to claim 2 is characterized in that: The linear prediction model calculation formula is: y=mx+c (2) Among them: y represents the predicted value of road data. x represents the input value of road data. m represents the slope, which reflects the rate of change of y with x. b represents the intercept, which is the value of y when x=0. By comparing the linear prediction and Transformer prediction parts, the areas affected by policies in urban development can be identified. The formula is: ΔR dif1 =R transformer -R true (3) ΔR dif2 =R linear -R true (4) Where: ΔR dif1 Indicates the difference between the road value predicted by the Transformer model and the actual value. ΔR dif2 Represents the difference between the linear model road prediction value and the actual road value. R transformer Represents the predicted road value of the Transformer model. linear Represents the predicted road value of the linear model. R true Indicates the actual road value.

4. The urban texture analysis and prediction method based on road network according to claim 1 is characterized in that: The Manhattan formula is as follows: d=|x1-x2|+|y1-y2| (5) in: d represents the Manhattan distance. x i and i Respectively represent the coordinates of the two intersection data in the i-th dimension. The quadratic polynomial model is as follows: in: y represents the predicted value of road data. x1 represents the length of input road data 1. x2 represents the length of input road data 2. t represents the prediction time minus the road data 1 time. a, b and c are model parameters obtained by fitting the road history data.

5. The urban texture analysis and prediction method based on road network according to claim 1 is characterized in that: The data preprocessing in step 2 includes deleting abnormal data, and deleting abnormal data includes deleting empty data, attributes irrelevant to this method, abnormal road length data, and various invalid road data.

Images