A multi-source data analysis and evaluation method for jogging index construction

Abstract

The present application belongs to the technical field of data processing, and specifically relates to a multi-source data analysis and evaluation method for constructing a jogging index, and specifically comprises the following steps: obtaining complete road network information and performing data cleaning and optimization on the same; superimposing movement trajectory data on the cleaned road network data, capturing the distribution of the trajectory data on each road segment, and calculating the jogging frequency; extracting the subjective and objective environmental related features respectively; establishing a random forest regression model, optimizing the key parameters of the random forest regression model, exploring the nonlinear relationship between the features and the jogging frequency, and obtaining the threshold value of each feature; and constructing a jogging index for evaluating the jogging capacity of urban streets. The present application overcomes the shortcomings of the prior art, integrates multi-source data, uses a random forest model to deeply explore the key influencing factors and the internal mechanism of the jogging capacity of urban streets, and provides technical support for scientifically evaluating the jogging suitability of streets.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to a multi-source data analysis and evaluation method for constructing a jogging index. Background Art

[0002] With socioeconomic development and the widespread adoption of health concepts, people are increasingly interested in improving their physical condition and quality of life through exercise. As a form of exercise with relatively low requirements for age, gender, physical condition, and skill level, jogging has gradually become a popular fitness method. Simultaneously, the country is vigorously promoting infrastructure development, with numerous fitness venues being built in major cities. While these venues meet residents' exercise needs to a certain extent, they still fall short in accommodating the personalized needs of joggers. Consequently, streets, as natural open spaces for exercise, have become the preferred venue for jogging for many.

[0003] Existing research indicates that street built-up environment characteristics (such as green view ratio and sky openness) significantly influence jogging frequency and route selection. However, the selection of specific indicators influencing jogging behavior fails to consider subjective perceptions, leaving room for improvement. Furthermore, scientific standards for selecting indicators influencing jogging behavior have yet to be established, and research into the weighting and mechanisms of various factors remains inadequate. Furthermore, traditional route recommendation methods typically rely on single-dimensional analysis, ignoring the complex relationship between environmental characteristics and jogging needs, resulting in recommendations that lack personalization and applicability. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-source data analysis and evaluation method for constructing a jogging index. This method overcomes the shortcomings of the existing technology. By integrating multi-source data, including motion trajectory data, street view images, built environment and perception scores, and using a random forest model, the method deeply explores the key influencing factors and internal mechanisms of urban street jogging ability, reveals the nonlinear relationship between various characteristics and jogging ability, and provides technical support for the scientific evaluation of street jogging suitability.

[0005] In order to solve the above problems, the technical solutions adopted by the present invention are as follows:

[0006] A multi-source data analysis and evaluation method for constructing a jogging index includes the following steps:

[0007] S1. Data preprocessing: Obtain complete road network information and perform data cleaning and optimization;

[0008] S2. Calculate jogging frequency: superimpose the motion trajectory data onto the cleaned road network data, capture the distribution of trajectory data on each road segment, and calculate the jogging frequency;

[0009] S3, multi-source data extraction: extract relevant features of subjective and objective environments respectively;

[0010] S4. Model building: Build a random forest regression model and optimize the key parameters of the random forest regression model through grid search and cross-validation;

[0011] S5. Model result analysis: Using feature importance analysis and partial dependence plot analysis based on the random forest regression model, we explored the nonlinear relationship between features and jogging frequency, and obtained the threshold value of each feature;

[0012] S6. Construct a jogging index: According to the feature importance score, set weights for each indicator, combine the feature thresholds extracted from the nonlinear analysis, optimize the index composition, and construct a jogging index for evaluating the jogging ability of urban streets.

[0013] Furthermore, the expression of the jogging frequency in S2 is:

[0014]

[0015] Among them, F r is the movement frequency of the road segment, which represents the total trajectory count of the road on the study day; N represents the total number of trajectories recorded on the day in the study area;

[0016] δ i represents the indicator function, setting a 50-meter buffer zone within each road, δ i It is used to determine whether the trajectory i is within the buffer zone of the road segment; if the trajectory i is within the buffer zone, δ i =1; otherwise, δ i =0.

[0017] Furthermore, the subjective and objective environment-related characteristics in S3 include road accessibility, road length, land use mixture, population density, green view rate, sky view, building view, and subjective perception score.

[0018] Furthermore, the specific steps of establishing the random forest regression model described in S4 include:

[0019] (1) Sampling the training data set, constructing several sub-data sets, and training a decision tree on each sub-data set;

[0020] (2) Each decision tree randomly selects features when splitting a node, reducing the correlation between features and enhancing the stability of the model;

[0021] (3) The final output value of the random forest is the average of the prediction values ​​of all decision trees. The specific formula is as follows:

[0022]

[0023] Where: is the predicted result, i.e., the jogging frequency; T b (X) is the predicted value of the b-th tree for the subjective and objective features X; B represents the number of decision trees in the forest.

[0024] (4) Optimize the key parameters of the random forest regression model through grid search and cross validation.

[0025] Furthermore, the random forest regression model uses goodness of fit and root mean square error to test model performance, and the specific expression is:

[0026]

[0027] Where n is the sample size; y i The true value of the dependent variable; predicted values ​​of the dependent variable; The mean of the predicted values ​​of the dependent variable.

[0028] Furthermore, the specific formula of the jogging index in S6 is:

[0029]

[0030] Where I represents the jogging index, w j For each variable X j The weight of is the error term; f j (X j ) is for each variable X j The nonlinear function is set.

[0031] Furthermore, the f j (X j ) is:

[0032]

[0033] Among them, k and X0 are parameters obtained by fitting the data to adjust the shape and position of the Sigmoid function.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. Based on the traditional built environment analysis, this invention further introduces the quantitative analysis of subjective feelings, comprehensively considering the objective environmental characteristics and people's subjective perception, making the types of independent variables more comprehensive and improving the scientificity and applicability of the analysis.

[0036] 2. This paper uses a random forest model to effectively capture the nonlinear relationship between features and jogging frequency, and quantify the importance and threshold of each variable. This method significantly outperforms traditional multivariate linear regression models and performs better when dealing with complex data relationships and multidimensional features.

[0037] 3. Based on the results of feature importance analysis, this paper assigns weights to each indicator and, in combination with variable thresholds, constructs a more scientific and reasonable jogging index to quantify the jogging suitability of urban streets. Compared with traditional index construction methods, the jogging index of this invention is more accurate and intuitive, providing joggers with a reliable reference for choosing routes.

[0038] 4. By constructing a jogging index, this invention provides joggers with a quantitative, personalized reference for jogging route selection, helping them choose the optimal route based on scientific data, thereby enhancing their jogging experience. Furthermore, the research results of this invention can indirectly provide data support for urban street planners to optimize jogging environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of the structure of a multi-source data analysis and evaluation method for constructing a jogging index.

[0040] Figure 2 This is an enlarged schematic diagram of point A in a multi-source data analysis and evaluation method for constructing a jogging index.

[0041] Figure 3 This is a schematic diagram of the structure of a support seat in a multi-source data analysis and evaluation method for constructing a jogging index.

[0042] Figure 4 Schematic diagram of the connection blocks in a multi-source data analysis and evaluation method for constructing a jogging index. DETAILED DESCRIPTION

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0044] As shown in the figure, the multi-source data analysis and evaluation method for constructing a jogging index according to the present invention includes the following steps:

[0045] Step 1: Select the research area

[0046] This paper selects the main urban area of ​​Shanghai, China as the research area, covering the core urban districts such as Pudong New Area, Xuhui District, Jing'an District and Huangpu District, and regards the roads in the area as the basic analysis unit. Figure 2 shown.

[0047] Step 2: Jogging data preprocessing

[0048] The present invention selects jogging track data of Shanghai on a certain day in 2019, totaling 11256 records. The data comes from the sports application "Keep" and contains the user's motion path information during jogging. Figure 3 shown.

[0049] First, we obtained the complete road network data for the study area from Open Street Map (OSM). We then cleaned and optimized the data to remove redundant information and irrelevant road types to ensure accurate matching between the road network data and the jogging trajectories. Subsequently, we overlaid the motion trajectory data onto the cleaned road network data. To accurately capture the distribution of trajectory data across each road segment, we set a 50-meter buffer zone for each road. The formula for calculating jogging frequency is as follows:

[0050]

[0051] Where: F r is the movement frequency of the road segment, which represents the total trajectory count of the road on the study day; N represents the total number of trajectories recorded in the study area on that day; δ i represents the indicator function, which is used to determine whether the trajectory i is located in the buffer zone of the road segment. If the trajectory i is in the buffer zone, δ i =1; otherwise, δ i =0.

[0052] Step 3: Feature extraction

[0053] This step uses multi-source big data, including street view images, subjective perception scores, road networks, building distribution, and population information, to construct the subjective and objective environmental feature system of the present invention.

[0054] Road network and river data were obtained from OSM, while points of interest (POI), building area (AOI), and bus stop information were obtained from Amap. Based on the standard 10-minute walking distance, a 500-meter buffer zone was established at the center of each road segment as a service area. This was used to analyze road accessibility, road length, and land use mix. Population data was obtained from the WorldPop database, from which population density was calculated.

[0055] Street view image data is obtained through Baidu Map API, as shown in the attached Figure 4As shown in the figure, sampling points were placed at 50-meter intervals along the road network, collecting over 48,000 panoramic images. After preprocessing, the image data was semantically segmented using the DeepLab V3+ model trained on the Cityscape dataset. Visual features such as buildings, greenery, roads, and sky were extracted, and visual landscape variables such as green view fraction (GVI), sky view fraction (SVF), and building view fraction (BVF) were calculated.

[0056] In terms of subjective perception, the Place Pulse 2.0 dataset was introduced. This dataset uses global participants' ratings of street view images to quantify the subjective attributes of urban environments across four dimensions: safety, aesthetics, vitality, and affluence. The scores from this dataset were combined with a CNN model to calculate indicators.

[0057] The specific calculation method is as follows:

[0058] (1) Population density

[0059] Population data was extracted from the World Pop website. The extracted population raster information was imported into ArcGIS, and the "Summary Statistics" tool was used to calculate the sum of the population within each study buffer zone, thereby calculating the population density. The calculation formula is as follows:

[0060]

[0061] Where: P is the population in the buffer zone; A is the area of ​​the 500-meter buffer zone.

[0062] (2) Land use mix

[0063] The obtained Shanghai land use SHP file was imported into ArcGIS. The land use types within each buffer zone were extracted using the Merge and Intersect tools. The land use area of ​​each type within each buffer zone was obtained using computational geometry. In this paper, the land use mix degree of each buffer zone was calculated using the entropy index method. The calculation formula is as follows:

[0064]

[0065] Where: H is the entropy value, which represents the diversity of land use; n is the number of land use types (such as residential, commercial, agricultural, green space, etc.); p i is the proportion of the i-th land use type (i.e. the proportion of land area of ​​this type in the total area).

[0066] (3) Road density

[0067] Road accessibility can be expressed by road density, and the calculation formula is as follows:

[0068]

[0069] Where: D is the road density; L 道路 is the total length of roads within the 500-meter buffer zone; A 缓冲区 is the area of ​​the 500-meter buffer zone.

[0070] (4) Green View Index (GVI)

[0071] Green view ratio measures the proportion of green space visible from an observation point and is commonly used to assess the visibility of green spaces in urban environments.

[0072]

[0073] (5) Sky View Field (SVF)

[0074] Sky viewshed is used to describe the proportion of sky that can be seen from a ground observation point and is a quantitative measure of sky exposure.

[0075]

[0076] (6) Building View Field (BVF)

[0077] The building viewshed refers to the proportion of buildings seen from the observation point, reflecting the proportion of buildings within a certain field of view.

[0078]

[0079] Step 4: Build the model

[0080] This paper uses the random forest model to conduct a regression analysis of the factors affecting jogging frequency. The core advantage of the random forest model lies in its good adaptability to multidimensional features and nonlinear relationships, and its strong ability to resist overfitting.

[0081] As an ensemble learning method, random forest ensures the generalization ability of the model by constructing multiple decision trees and randomly selecting features for node splitting based on each tree.

[0082] In the random forest regression model, the training dataset is first sampled to construct several sub-datasets, and a decision tree (=1, 2, ...,) is trained on each sub-dataset. Each tree randomly selects features when splitting a node to reduce the correlation between features and enhance the stability of the model. In this invention, the final output value of the random forest is the average of the prediction values ​​of all decision trees, as shown in the formula:

[0083]

[0084] Where: is the predicted result, i.e., the jogging frequency; Tb (X) is the predicted value of the b-th tree for the subjective and objective features X; B represents the number of decision trees in the forest.

[0085] Through grid search and cross-validation, we optimize the key parameters of the model. The optimal parameter setting ensures the model's prediction accuracy while reducing the risk of overfitting and enhancing the model's generalization ability.

[0086] Step 5: Model results analysis

[0087] This step uses the goodness of fit (R 2 ) and root mean square error (RMSE) to test the model performance. 2 The higher the RMSE, the better the model fitting performance. The lower the RMSE, the more accurate the model prediction and the higher the feasibility.

[0088] The specific formula is as follows:

[0089]

[0090]

[0091] Where: n is the sample size; y i The true value of the dependent variable; predicted values ​​of the dependent variable; The mean of the predicted values ​​of the dependent variable.

[0092] Feature importance analysis and partial dependence plot analysis based on random forest regression model were used to explore the nonlinear relationship between objective and subjective features and jogging frequency, and to obtain the threshold of each feature.

[0093] Step 6: Jogging Index Setting

[0094] Based on the nonlinear analysis in step 5, the importance of each feature and the threshold are combined to set the weight of each indicator, and a jogging index is constructed to evaluate the jogging ability on urban streets. The specific formula is as follows:

[0095]

[0096] Where: I represents the jogging index, w j Represents the importance of features output by the machine learning model for each variable X j Set weights. The weights reflect the relative importance of each variable in predicting jogging frequency. These weights are derived based on the results of machine learning analysis. is the error term; f j (X j ) is for each variable X jThe nonlinear function is set based on the nonlinear threshold of the variable to more accurately simulate its effect on jogging frequency. In addition, for f j (X j ) function, if the relationship between a variable and the target variable has a sudden change or saturation phenomenon at a certain point, the logarithm, Sigmoid or piecewise function will be used to adapt to this nonlinear feature. For example, X j is the green visual index (GVI). From the nonlinear graph, it is known that the relationship between GVI and jogging frequency tends to be stable after reaching a certain high level. The Sigmoid function is selected to simulate this relationship:

[0097]

[0098] Where: k and X0 are parameters obtained from data fitting, which adjust the shape and position of the Sigmoid function.

[0099] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A multi-source data analysis and evaluation method for constructing a jogging index, characterized by: The specific steps include: S1. Data preprocessing: Obtain complete road network information and perform data cleaning and optimization; S2. Calculate jogging frequency: superimpose the motion trajectory data onto the cleaned road network data, capture the distribution of trajectory data on each road segment, and calculate the jogging frequency; S3, multi-source data extraction: extract relevant features of subjective and objective environments respectively; S4. Model building: Build a random forest regression model and optimize the key parameters of the random forest regression model through grid search and cross-validation; S5. Model result analysis: Using feature importance analysis and partial dependence plot analysis based on the random forest regression model, we explored the nonlinear relationship between features and jogging frequency, and obtained the threshold value of each feature; S6. Constructing a jogging index: setting weights for each indicator based on feature importance scores, combining feature thresholds extracted from nonlinear analysis, optimizing index composition, and constructing a jogging index for evaluating jogging ability on urban streets; The subjective and objective environment-related characteristics described in S3 include road accessibility, road length, land use mix, population density, green view rate, sky view, building view and subjective perception score; the subjective perception score quantifies the subjective attributes of the urban environment in four dimensions: sense of safety, beauty, vitality and affluence.

2. The multi-source data analysis and evaluation method for constructing a jogging index according to claim 1, characterized in that: The expression of jogging frequency in S2 is: in, is the movement frequency of the road segment, which represents the total trajectory count of the road on the study day; N represents the total number of trajectories recorded on the day in the study area; represents the indicator function, setting a 50-meter buffer zone within each road. It is used to determine whether the track i is in the buffer zone of the road segment; if the track i is in the buffer zone, ;otherwise, .

3. The multi-source data analysis and evaluation method for constructing a jogging index according to claim 1, characterized in that: The specific steps for establishing the random forest regression model described in S4 include: (1) Sampling the training data set, constructing several sub-data sets, and training a decision tree on each sub-data set; (2) Each decision tree randomly selects features when splitting a node, reducing the correlation between features and enhancing the stability of the model; (3) The final output value of the random forest is the average of the prediction values ​​of all decision trees. The specific formula is as follows: Where: The predicted result is the jogging frequency; is the predicted value of the b-th tree for the subjective and objective features x; B represents the number of decision trees in the forest; (4) Optimize the key parameters of the random forest regression model through grid search and cross validation.

4. The multi-source data analysis and evaluation method for constructing a jogging index according to claim 3, characterized in that: The random forest regression model uses goodness of fit and root mean square error to test model performance. The specific expression is: in, is the sample size; The true value of the dependent variable; The predicted value of the dependent variable; The mean of the predicted values ​​of the dependent variable.

5. The multi-source data analysis and evaluation method for constructing a jogging index according to claim 1, characterized in that: The specific formula of the jogging index described in S6 is: Where I represents the jogging index, For each variable The weight of is the error term; For each variable The nonlinear function is set.

6. The multi-source data analysis and evaluation method for constructing a jogging index according to claim 5, characterized in that: described The expression is: in, and It is based on the parameters obtained by fitting the data to adjust the shape and position of the Sigmoid function.

Images