Urban farmland circulation price automatic evaluation method, device and equipment

Abstract

The invention relates to the field of land evaluation, and provides an urban farmland circulation price automatic evaluation method, device and equipment, and the method comprises the steps: constructing a multi-dimensional and multi-source evaluation index system, and calculating the index score of an evaluation index in each evaluation unit; collecting and pre-processing urban farmland circulation price data of actual transaction; constructing a multi-machine learning model fused with a Stacking integration algorithm; constructing a plurality of groups of sample point data by taking the urban farmland circulation price actually transacted by the evaluation unit as a dependent variable and taking the index score of the evaluation index of the evaluation unit as an independent variable; training and verifying each machine learning model by using the sample point data to obtain an optimal model passing verification; and using the verified optimal model to automatically evaluate the urban farmland circulation price. According to the method, cost reduction and efficiency increase of the urban farmland evaluation method can be realized, a simpler and more efficient automatic evaluation whole-process workflow is constructed, and the limitations of high subjectivity, high cost, low efficiency and the like of a traditional evaluation method are overcome.

Description

Technical Field

[0001] The present invention relates to the field of land valuation, and in particular, to an automatic valuation method, device and equipment for the transfer price of urban agricultural land. Background Art

[0002] As the "frontier" of urban-rural integration, urban agricultural land has the closest urban-rural relationship and more active land transfer compared with other agricultural production areas. Its transfer price is of great significance for coordinating urban development and cultivated land protection. At present, there are still problems such as implicit and unreasonable transfer prices of agricultural land in China, which are difficult to effectively reflect the fine-scale differences of agricultural land and are not conducive to the rational allocation of market resources; the allocation of urban agricultural land lags significantly behind the "hollowing out of villages", and higher requirements are also put forward for the complex regulation of land transfer. Under the new situation of the modernization of urban agricultural land governance, the continuously expanding scope and scale of urban agricultural land transfer also put forward a stronger practical need to quickly master and understand the changing rules of urban agricultural land prices at a fine scale. Therefore, how to give full play to the advantages of the "frontier" of urban-rural integration of urban agricultural land, explore a new and efficient method for valuing the transfer price of agricultural land, solve the "common problems" of the transfer price of agricultural land, and deeply explain the underlying logic of the transfer price of urban spatial agricultural land to meet the needs of refined valuation in complex real-world scenarios is crucial for optimizing the transfer configuration of urban agricultural land and improving the agricultural land valuation method system. Under the land price system in China, the evaluation of the benchmark land price of agricultural land provides a price reference for the transfer of urban agricultural land, but in practice, limitations such as strong subjectivity, low efficiency and high cost have been exposed. The advent of the era of data explosion and the development of machine learning methods have pointed the way for the optimization of traditional valuation methods. On the one hand, the current rich, accessible and geographically tagged multi-source data environment in cities provides a huge amount of data basis for machine learning, greatly improving the data mining ability of the model and facilitating the exploration of new methods for urban agricultural land valuation; on the other hand, compared with other traditional methods such as mathematical models and spatial analysis, machine learning overcomes the inherent defects of interpretive modeling, is more capable of capturing the complex relationships between variables, and enhances the prediction ability of the model. With the rapid development of artificial intelligence technology in recent years, more diverse and complex models such as random forest (RF) and extreme gradient boosting (XGBoost) have emerged in machine learning. To reduce the uncertainty of the model, multi-model integration methods represented by the stacking algorithm have also been widely developed and applied. At the same time, with the development of the research paradigm of geospatial big data, the valuation conditions of machine learning have been significantly improved, giving full play to the advantages of quickly processing a large amount of multi-source data, and showing significant application value and advantages in urban fine-scale valuation. Therefore, the combination of machine learning and multi-source data has important application potential for new urban agricultural land valuation. Summary of the Invention

[0003] In view of the deficiencies of the prior art, the present invention provides an automatic valuation method, device and equipment for the transfer price of urban agricultural land, aiming to solve at least some of the problems existing in the prior art.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] In the first aspect, the present invention provides an automatic valuation method for the transfer price of urban agricultural land, including the following steps:

[0006] Construct a multi-dimensional and multi-source valuation index system, and calculate the index scores of the valuation indexes in each valuation unit;

[0007] Collect the transfer price data of urban agricultural land actually transacted and perform preprocessing;

[0008] Construct a multi-machine learning model integrating the Stacking integration algorithm;

[0009] Taking the transfer price of urban agricultural land actually transacted in the valuation unit as the dependent variable and the index scores of the valuation indexes of the valuation unit as the independent variables, construct multiple sets of sample data, and divide the multiple sets of sample data into a training set and a validation set according to a preset ratio;

[0010] Input the training set into each machine learning model for training, select the optimal model, and use the validation set for verification to obtain the optimal model that passes the verification;

[0011] Use the optimal model that passes the verification for automatic valuation of the transfer price of urban agricultural land.

[0012] Preferably, the construction of the multi-dimensional and multi-source valuation index system and the calculation of the index scores of the valuation indexes in each valuation unit specifically include:

[0013] Construct a "natural - economic - social - ecological" multi-dimensional and multi-source data index system, and according to the influence radius of the source data of the valuation index, adopt a spatial representation method based on the valuation unit to calculate the global distance attenuation score, and perform spatial superposition on the scores calculated according to the source data of each index according to the preset weight to obtain the index scores of the valuation indexes in each valuation unit.

[0014] Preferably, after calculating the index scores of the valuation indexes in each valuation unit, it further includes: performing standardization processing on the index scores of all valuation indexes, sorting the importance of multiple valuation indexes, and retaining the most important n effective valuation indexes.

[0015] Preferably, the preprocessing includes: removing outliers and extreme values from the transfer price data of urban agricultural land actually transacted;

[0016] And / or, for the urban agricultural land transfer price data of different trading years, calculate the land price index for each year and uniformly correct the prices of each year to the latest year.

[0017] Preferably, the construction of the multi-machine learning model integrating the Stacking integration algorithm specifically includes: constructing multiple preset single machine learning models and integrating them into multiple ensemble learning models using the Stacking algorithm.

[0018] Preferably, the step of inputting the training set into each machine learning model for training and screening out the optimal model specifically includes:

[0019] Using the grid search method to search for the optimal parameter range of the single machine learning model and adjust the parameters;

[0020] After parameter adjustment, input the training set into the machine learning model and the ensemble learning model for training, use the ten-fold cross-validation method to evaluate the performance indicators, and screen out the optimal model.

[0021] Preferably, it further includes: calculating the shap values of multiple valuation indicators in the optimal model and generating a shap value ranking graph.

[0022] Preferably, it further includes: performing artificial intelligence analysis on the valuation data of urban agricultural land transfer prices.

[0023] In a second aspect, the present invention also provides an automatic valuation device for urban agricultural land transfer prices, including:

[0024] An index score calculation module, configured to construct a multi-dimensional and multi-source valuation index system and calculate the index scores of the valuation indicators in each valuation unit;

[0025] A transaction price preprocessing module, configured to collect the actual transaction price data of urban agricultural land transfer and perform preprocessing;

[0026] A model construction module, configured to construct a multi-machine learning model integrating the Stacking integration algorithm;

[0027] A sample point construction module, configured to use the actual transaction price of urban agricultural land transfer in the valuation unit as the dependent variable and the index scores of the valuation indicators of the valuation unit as the independent variable to construct multiple groups of sample point data, and divide the multiple groups of sample point data into a training set and a validation set according to a preset ratio;

[0028] A model training module, configured to input the training set into each machine learning model for training, screen out the optimal model, and use the validation set for verification to obtain the optimal model that passes the verification;

[0029] An automatic valuation module, configured to use the optimal model that passes the verification to perform automatic valuation of urban agricultural land transfer prices.

[0030] In a third aspect, the present invention further provides an automatic valuation device for the transfer price of urban agricultural land, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned automatic valuation method for the transfer price of urban agricultural land are implemented.

[0031] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0032] The present invention constructs a multi-dimensional and multi-source valuation index system, calculates the index scores of the valuation indexes in each valuation unit; collects and preprocesses the transfer price data of actual urban agricultural land transactions; constructs a multi-machine learning model integrating the Stacking integration algorithm; uses the actual transfer price of urban agricultural land in the valuation unit as the dependent variable, and the index scores of the valuation indexes of the valuation unit as the independent variable to construct multiple groups of sample data, and divides the multiple groups of sample data into a training set and a verification set according to a preset ratio; inputs the training set into each machine learning model for training, screens out the optimal model, and uses the verification set for verification to obtain the optimal model that passes the verification; uses the verified optimal model for automatic valuation of the transfer price of urban agricultural land. The present invention can achieve "cost reduction and efficiency improvement" of the urban agricultural land valuation method, constructs a simpler and more efficient workflow for the whole valuation process, and overcomes the limitations of traditional valuation methods such as strong subjectivity, high cost, and low efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flowchart of an automatic valuation method for the transfer price of urban agricultural land provided in this embodiment;

[0034] Figure 2 It is a module diagram of an automatic valuation device for the transfer price of urban agricultural land provided in an embodiment of the present invention;

[0035] Figure 3 It is an analysis result diagram of the "Rank" component in an embodiment of the present invention;

[0036] Figure 4 It is a precision evaluation result diagram of each machine learning model and integrated learning model in an embodiment of the present invention;

[0037] Figure 5 It is a shap value ranking diagram of the optimal prediction models for dry land and paddy fields in an embodiment of the present invention;

[0038] Figure 6 It is a land price frequency diagram, a land price box plot, and a land price distribution map processed by natural break grading generated in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0040] As Figure 1 shown, the embodiments of the present invention provide an automatic valuation method for the transfer price of urban agricultural land, including the following steps:

[0041] S1. Construct a multi-dimensional and multi-source valuation index system, calculate the index scores of the valuation indexes in each valuation unit, and select the valuation indexes according to the importance;

[0042] The valuation indexes refer to some indexes that affect the transfer price of urban agricultural land, such as: agricultural land natural grade index, farmland production potential, central town influence degree, farmers' market influence degree, tourism and leisure value, road accessibility, external traffic convenience degree, agricultural land utilization grade index, agricultural land economic grade index, land expropriation compensation standard, per capita cultivated land area, landscape ecological diversity index, and aesthetic landscape value, etc. The valuation unit refers to the basic spatial unit to be valued. Specifically, it refers to the basic spatial unit for estimating the transfer price of urban agricultural land by using machine learning methods, and at the same time, it is also the basic unit for collecting data and performing calculations in actual work. The division basis of the valuation unit can be the administrative region scope, for example, using administrative villages as the valuation units.

[0043] Specifically, construct a "natural - economic - social - ecological" multi-dimensional and multi-source data valuation index system. The valuation indexes include m multi-source data valuation indexes such as agricultural land natural grade index, farmland production potential, central town influence degree, farmers' market influence degree, tourism and leisure value, road accessibility, external traffic convenience degree, agricultural land utilization grade index, agricultural land economic grade index, land expropriation compensation standard, per capita cultivated land area, landscape ecological diversity index, and aesthetic landscape value, etc., and calculate the index scores of these m valuation indexes for the valuation unit.

[0044] The specific index scores of the calculation valuation indicators in each valuation unit include: according to the influence radius of the valuation indicator source data, a spatial representation method based on the valuation unit is used to calculate the global distance attenuation score, and the scores calculated according to each index source data are spatially superimposed according to a preset weight to obtain the index scores of the valuation indicators in each valuation unit. According to the spatial form of the carrier on which the multi-source data valuation indicators rely and different forms of influence on the agricultural land transfer price, they are divided into point valuation indicators, line valuation indicators, and surface valuation indicators. Point valuation indicators belong to diffusive influence indicators, such as farmers' markets, etc. The quality of their valuation indicators not only affects the plots with this carrier, but also affects the surrounding agricultural land; line valuation indicators belong to diffusive influence indicators, such as roads of different grades such as national roads, main roads, and secondary roads. The quality of their valuation indicators not only affects the plots with this carrier, but also affects the agricultural land within a certain distance range; surface valuation indicators belong to non-diffusive influence indicators, such as irrigation facilities. The quality of their valuation indicators only affects the plots with this carrier.

[0045] The point valuation indicators decay to their surroundings in the form of concentric circles, and the decay methods are divided into linear model and exponential model decay. According to different situations of the influence of the point valuation indicator source data on the valuation unit, different methods can be used to calculate the influence radius of the point valuation indicator source data. For example, a certain level of some point valuation indicators can have many source data, and all source data jointly affect the entire valuation area, and the influence degree is basically the same. When calculating, the influence radius of these valuation indicator source data needs to be arithmetically averaged and divided. The influence radius of these point valuation indicators is calculated according to the following formula:

[0046]

[0047] In the formula: d is the influence radius of the source data of a certain level of a certain point valuation indicator;

[0048] s is the total area of all valuation units;

[0049] n is the number of source data of a certain level of a certain point valuation indicator.

[0050] The formula for calculating the score of the point valuation indicator according to the linear attenuation method is as follows:

[0051] f i =M i ×(1 - r), r = d i / d

[0052] The formula for calculating the score of the point valuation indicator according to the exponential attenuation method is as follows:

[0053] f i =M i (1-r) , r = di / d

[0054] In the formula: f i is the score of the evaluation index;

[0055] M i is the scale index, that is, the influence degree of the carrier on which the evaluation index depends at the most central position;

[0056] d i is the actual distance between the carrier on which the evaluation index depends and the evaluation unit i;

[0057] d is the influence radius of the evaluation index;

[0058] r is the relative distance between the carrier on which the evaluation index depends and the evaluation unit i.

[0059] The linear evaluation index is an expansion factor, and the index score decreases with the increase of the distance. The attenuation mode attenuates according to the above linear model and exponential model.

[0060] The influence radius of the linear evaluation index is calculated according to the following formula:

[0061] d = s / 2L

[0062] In the formula: d is the influence distance of various linear evaluation indexes;

[0063] s is the total area of all evaluation units;

[0064] L is the total length of various linear evaluation indexes.

[0065] The quantification method of the areal evaluation index is as follows: According to the characteristics of the correlation between the evaluation index and the agricultural land quality, calculate the evaluation index values of each region or evaluation unit, and process the evaluation index values outside the significant interval according to the highest or lowest value within the significant interval; then, use a mathematical model to calculate the scores of each agricultural land evaluation index.

[0066] f i = 100×(X i - X min ) / (X max - X min )

[0067] In the formula: f i is the score of a certain agricultural land evaluation index;

[0068] X min 、X max 、X i are respectively the minimum value, the maximum value and the value of a certain agricultural land evaluation index of the index.

[0069] Specifically, each valuation index may correspond to multiple sources of data. For example, the sources of data for the tourism and leisure value index include tourist attractions and farmhouses. Tourist attractions are further divided into different levels. According to the "Classification and Evaluation of Quality Grades of Tourist Attractions" (GB / T 17775-2003) standard, the quality grades of tourist attractions are divided into five levels, from high to low, namely AAAAA, AAAA, AAA, AA, and A-level tourist attractions. The influence radii of tourist attractions of the same level are the same, while those of different levels are different. When calculating the index score of the tourism and leisure value evaluation index for a certain valuation unit, first determine the influence radii of tourist attractions of different levels and the influence radii of sources of data such as farmhouses, which can be obtained through calculation or directly preset by referring to existing data. Then, use the spatial representation method based on the valuation unit to calculate the global distance attenuation score to obtain the scores of tourist attractions of different levels and sources of data such as farmhouses for this valuation unit. Then, spatially superimpose the scores of tourist attractions of different levels and sources of data such as farmhouses for this valuation unit according to the preset weights to obtain the index score of the tourism and leisure value evaluation index for this valuation unit.

[0070] Taking the calculation of the tourism and leisure value index score in Wuhan as an example, as of now, there are 50 scenic spots at or above 3A level in Wuhan, including 3 5A-level scenic spots, 22 4A-level scenic spots, and 25 3A-level scenic spots. According to the quality grade levels of tourist attractions in Wuhan, scale indices of 100, 50, and 20 are respectively assigned. Each level of tourist attraction has corresponding service functions. High-level tourist attractions include the functions of low-level tourist attractions, and the higher the level, the higher the functional level. That is, 5A-level tourist attractions include both the service functions of 4A-level tourist attractions and the service functions of 3A-level tourist attractions; 4A-level tourist attractions include the functions of 3A-level tourist attractions; 3A-level tourist attractions only include their own service functions.

[0071] Based on the above principle, the service function score of tourist attractions is segmented according to the following formula.

[0072] G i =f i -f j

[0073] G min =f min

[0074] In the formula: G i —The function score of a certain level of tourist attraction;

[0075] f i —The average value of the scale index of a certain level of tourist attraction;

[0076] f j— The average value of the scale index of the secondary-level tourist attractions;

[0077] G min — The functional score of the lowest-level tourist attractions;

[0078] f min — The average value of the scale index of the lowest-level tourist attractions.

[0079] The service radius of a tourist attraction refers to the maximum distance from the tourist attraction to its influence boundary, which is related to the scale and influence ability of the tourist attraction. Starting from the edge of the tourist attraction, the determination method is as follows: The service radius of a 5A-level tourist attraction is equal to the maximum distance from the edge of the 5A-level tourist attraction to the edge of agricultural land in Wuhan; The service radius of other levels of tourist attractions is equal to the maximum service distance of the same-level tourist attractions. According to the above principles, combined with field investigations and the analysis of the distribution of tourist attractions, the service radii of tourist attractions at all levels are determined.

[0080] The relative distance is calculated according to the following formula:

[0081] r = d i / d

[0082] In the formula: r — relative distance;

[0083] d i — The actual distance of a certain point from the center within the service radius of the i-level tourist attraction;

[0084] d — The service radius of the i-level tourist attraction.

[0085] According to the functional scores and service radii of tourist attractions at all levels, the following table of the functional scores and service radii of tourist attractions in Wuhan is obtained:

[0086] Level Number of tourist attractions Functional score Service radius (m) 5A level 3 50 72773 4A level 22 30 10502 3A level 25 20 9812

[0087] The data of the benefits of each farmhouse collected in the investigation (passenger flow, turnover, total profit and tax, area, etc.) are summarized and statistically analyzed, and weights of 0.35, 0.35, 0.1, and 0.2 are respectively assigned to obtain the summary of relevant data of each farmhouse.

[0088] The farmhouse data is standardized according to the following formula to obtain a dimensionless statistical table of each farmhouse.

[0089] M i = 100×a i / a max

[0090] In the formula: M i —— The dimensionless index value after standardization;

[0091] a i —— The statistical value of the i-index data;

[0092] a max —— The maximum value of the data statistical value in this indicator.

[0093] Calculate the scale index according to the dimensionless statistical table of rural tourism and the index weights according to the following formula.

[0094]

[0095] In the formula: F j —— The scale index of the j-th rural tourism.

[0096] W i —— The weights of each indicator.

[0097] M ij —— The dimensionless index value of the i-th indicator of the j-th rural tourism after standardization processing.

[0098] Rural tourisms at all levels have corresponding function scores. The function scores of rural tourisms at different levels are as follows:

[0099] Rural tourism scale index Rural tourism level Functional score ≥60 1 100 [40，60) 2 80 ＜40 3 60

[0100] Determine the influence radius of rural tourisms at all levels:

[0101]

[0102] In the formula: d —— The influence radius of the i-th level of rural tourism;

[0103] n —— The number of rural tourisms at the i-th level;

[0104] s —— The total land area.

[0105] Based on the influence of the tourism and leisure value of rural tourisms and tourist attractions on the agricultural land in Wuhan, assign weights of 0.6 to rural tourisms and 0.4 to tourist attractions.

[0106] The calculation formula for the tourism and leisure value score is as follows:

[0107]

[0108] In the formula:

[0109] F —— The tourism and leisure value of a certain valuation unit;

[0110] W j —— The weight of the j-th type of tourism and leisure value measurement index;

[0111] f j —— The action score of the j-th type of measurement index of a certain valuation unit;

[0112] j —— Tourist attractions, rural tourisms;

[0113] f i —— The function score of the i-th level of tourist scenic area / farmhouse in a certain valuation unit;

[0114] M i —— The function score of the i-th level of tourist scenic area / farmhouse;

[0115] d i —— The shortest traffic distance from the center of the i-th level of tourist scenic area / farmhouse to the center of each valuation unit;

[0116] d—— The maximum influence radius of the i-th level of tourist scenic area / farmhouse;

[0117] r i —— Relative distance, (when d i ≥d, r i = 1)

[0118] The above is an example for calculating the index scores of the valuation indicators in each valuation unit. Specifically, other existing methods can also be used for calculation, which will not be elaborated here.

[0119] In some preferred embodiments, after calculating the index scores of the valuation indicators in each valuation unit, it further includes: performing standardization processing on the index scores of all valuation indicators to eliminate the influence of dimensions. Specifically, the standardization formula is as follows:

[0120]

[0121] Among them, Z is the standardized valuation indicator score, X is the valuation indicator score to be standardized, μ is the mean value of the scores of this valuation indicator in all valuation units, and σ is the standard deviation of the scores of this valuation indicator in all valuation units.

[0122] Preferably, it further includes: ranking the importance of multiple valuation indicators, and retaining the most important n effective valuation indicators to further improve the scientificity and reliability of the urban agricultural land valuation indicators.

[0123] In this embodiment, the importance parameter is at least one of univariate regression and RReliefF value. The importance of each indicator is judged by comparing the magnitudes of the average values of the two parameters, and the m - n indicators with the lowest importance rankings are removed from the preset m multi-source data valuation indicators, thereby forming n effective valuation indicators.

[0124] S2. Collect the actual transaction price data of urban agricultural land transfers and perform preprocessing;

[0125] The preprocessing includes: removing outliers and extreme values from the actual transaction price data of urban agricultural land transfers;

[0126] And / or, for the urban agricultural land transfer price data of different transaction years, calculate the land price index for each year and uniformly revise the prices of each year to the latest year.

[0127] Specifically, perform outlier and anomaly value removal on the actually transacted urban agricultural land transfer price data, including:

[0128] Collect the urban agricultural land transfer transaction case data for the preset range of years through the local government's urban agricultural land transaction data platform in the target area, and remove the obviously high or low abnormal price data. Use one of the algorithms of one-class support vector machine, covariance estimation, local outlier factor, and isolation forest to remove the outlier values. Usually, the proportion of outlier values removed is set to 5%.

[0129] Specifically, for the urban agricultural land transfer price data of different transaction years, calculate the land price index for each year and uniformly revise the prices of each year to the latest year, including:

[0130] According to the range of years of the collected sample points' transactions, calculate the transaction date correction coefficient and uniformly revise the sample point land prices to the latest year. The calculation formula for the transaction date correction coefficient is:

[0131] K = P / Pj

[0132] In the formula: K is the coefficient for revising the land price in the jth period to the latest year, P is the average land transfer price in the latest year, and Pj is the average land transaction price in the jth period.

[0133] In some preferred embodiments, after revising all the transaction prices to after 2023, take the average value of the sample point prices of the same valuation unit and the same land use type. That is, finally, each valuation unit with transaction sample points has only one price data. To eliminate the incomparability of the agricultural land transfer prices in different years due to factors such as time change and inflation.

[0134] S3. Construct a multi-machine learning model integrating the Stacking integration algorithm;

[0135] The specific construction of multiple machine learning models includes: constructing multiple preset single machine learning models including 4 classic machine learning models of K-nearest neighbor (KNN), Adaptive Boosting (AdaBoost), Random Forest (RF), and Extreme Gradient Boosting (XGBoost), and integrating them into multiple Stacking integrated learning models. By constructing multiple single machine learning models and multiple combined integrated models, it provides diverse model selections, integrates the advantages of multiple models, and improves the performance of the models.

[0136] S4. Taking the actual transfer price of urban agricultural land of the valuation unit as the dependent variable and the index scores of multiple valuation indicators of the valuation unit as the independent variables, construct multiple groups of sample data, and divide the multiple groups of sample data into a training set and a validation set according to a preset ratio.

[0137] Then, taking the actual transfer price of urban agricultural land of the valuation unit as the dependent variable and the index scores of n effective valuation indicators of the valuation unit as the independent variables, construct multiple groups of sample data. According to the principle of uniform spatial distribution, divide the multiple groups of sample data into a training set c1 and a validation set c2 according to a preset ratio.

[0138] S5. Input the training set into each machine learning model for training, screen out the optimal model, and use the validation set for verification to obtain the optimal model that passes the verification.

[0139] The step of inputting the training set into each machine learning model for training and screening out the optimal model specifically includes:

[0140] Use the grid search method to search for the optimal parameter ranges of K-Nearest Neighbor (KNN), Adaptive Boosting (AdaBoost), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) and perform parameter tuning.

[0141] After parameter tuning, input the training set c1 into the machine learning model and the ensemble learning model for training, use the ten-fold cross-validation method to evaluate the performance indicators, and screen out the optimal model.

[0142] After screening out the optimal model, use the validation set c2 that has not undergone any training to verify the optimal model. If the optimal model still performs well, the model passes the verification; if the optimal model performs poorly, continue to adjust the model parameters and retrain. By using a validation set that has not undergone any training to test the trained model, the possibility and contingency of model overfitting are reduced, ensuring that the final model has good generalization ability.

[0143] S6. Use the optimal model that passes the verification to perform automatic valuation of the transfer price of urban agricultural land and conduct artificial intelligence analysis.

[0144] Specifically, input the index scores corresponding to the n effective valuation indicators of all valuation units into the optimal model that has been trained and verified to achieve automatic valuation of all valuation units and generate price curve graphs, price frequency graphs, and land price maps in multiple downloadable formats. In this embodiment, by connecting to the "Distributions" component, a land price frequency graph can be automatically generated, and by connecting to the "Box Plot", a price box plot of each administrative region can be automatically generated, and multiple file format download methods such as svg, jpg, and png are provided. By importing the predicted price into GIS and performing natural break point method grading processing, a land price distribution map is obtained.

[0145] Preferably, the method further includes: calculating the SHAP values of n effective valuation indicators in the optimal model and generating a SHAP value ranking map. After verification, calculate the SHAP values of the n effective valuation indicators of the optimal model and provide SHAP value ranking maps in various downloadable formats. Introducing SHAP value calculation improves the interpretability of the model and provides decision-making support for exploring the dominant influencing factors of urban rural land transfer prices.

[0146] More preferably, project the obtained result map onto a display device and use a 3D projection device to apply generative AI and digital human technology to construct an application scenario of "human-computer interaction" between the projected digital human and the user. In this scenario, the user can not only intuitively see the prediction results but also communicate with the AI digital human about the prediction results to obtain a better user experience. Introducing the application scenario of "generative AI + digital human" enables users to have an efficient conversation with the digital human, realize intelligent analysis of the prediction results, and optimize the user experience.

[0147] Specifically, the AI technology comes from at least one of representative generative AI models such as Baidu Wenxin Yiyan, Alibaba Cloud Tongyi Qianwen, Google Bard, MetaMake-A-Video, OpenAI GPT-4, DALL-E 2, Stable Diffusion, etc.

[0148] The following uses a specific embodiment to illustrate the automatic valuation method for urban rural land transfer prices of the present invention.

[0149] From the perspective of data acquisition, the basic data is mainly divided into sample point data and multi-source data. In this embodiment, taking Wuhan City as the research area, the sample point data mainly comes from the Wuhan Rural Comprehensive Property Rights Exchange (https: / / www.whnccq.com / ), and is supplemented by field research. The time range is from 2014 to 2023; the multi-source data comes from the Wuhan Natural Resources Protection and Utilization Center, authoritative datasets of the Chinese Academy of Sciences, and web crawling.

[0150] From the perspective of the processing flow of the basic data, in this embodiment, the urban rural land transfer in Wuhan City is mainly dry land and paddy field. A total of 1391 transaction sample points are collected, including 699 paddy field sample points and 692 dry land sample points. Taking administrative villages as the valuation units, using the spatial decay function and spatial overlay function of Euclidean distance in GIS, calculate the index scores of each multi-source data valuation indicator for each administrative village, and extract the scores of the prediction factors to all administrative villages in the outer suburbs. All villages in the outer suburbs of Wuhan City are the prediction sets, and sample points are evenly selected in each district of the outer suburbs of Wuhan City as the verification set. Finally, 31 dry land sample points and 32 paddy field sample points are selected, and the remaining sample points are the training sets.

[0151] Import the Excel data table that combines the processed training set, validation set, prediction set names, longitude and latitude information, and index scores into the Orange platform for data preprocessing and index screening. Specifically, use the "Preprocess" component to standardize the scores of 13 indicators, and then take the average of the transaction prices of each administrative village. Input the preprocessed data into the "Rank" component, calculate the univariate regression, RReliefF value of each indicator and sort them, and eliminate the road accessibility indicator with the smallest average value. Finally, 12 multi-source data indicators and land use type classification indicators are retained, such as Figure 3 as shown

[0152] In this embodiment, the Bootstrap algorithm of the "Data Sampler" component is used to automatically sample the training set, and four classic machine learning models, namely K-Nearest Neighbor (KNN), Adaptive Boosting (AdaBoost), Random Forest (RF), and Extreme Gradient Boosting (XGBoost), are trained. Evaluate the model performance through Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R 2 ). The performance ranking of a single machine learning model is AdaBoost > RF > XGBoost > KNN. Based on the training performance of a single machine learning model, three ensemble learning models with excellent performance are constructed, namely the AdaBoost-RF stacked ensemble model (Stack-AR), the AdaBoost-RF-XGBoost stacked ensemble model (Stack-ARX), and the AdaBoost-RF-XGBoost-KNN stacked ensemble model (Stack-ARXK).

[0153] As Figure 4 shown, from the perspective of model performance, both the dryland and paddy field models perform well. Compared with the single model, the RMSE of the ensemble model is reduced by 12% on average, the MAE is reduced by 5% on average, and the R 2 is increased by 7% on average. It can be seen that through Stacking integration, the advantages of multiple models can be effectively combined to improve the prediction accuracy and stability. Finally, the AdaBoost model is selected to predict the dryland transfer price, and the Stack-ARXK model is selected to predict the paddy field transfer price. Input the dryland validation set data into the dryland AdaBoost model, and input the paddy field validation set data into the paddy field Stack-ARXK model. The R 2 values of dryland and paddy field are 0.781 and 0.737 respectively, and the model shows good generalization ability and passes the verification.

[0154] Connect the dryland AdaBoost model to the "Explain Model" component for model contribution analysis and visual display. Since the optimal model for paddy fields is the Stack-ARXK integrated model and cannot calculate the SHAP value, the AdaBoost model, which ranks second in terms of paddy field model performance, is connected to the "Explain Model" component. The automatically generated index SHAP value ranking diagrams for dryland and paddy field models are as Figure 5 shown.

[0155] Input the prediction set data into the optimal model, and use the "Geo Map" component to visualize the geospatial distribution of the predicted prices. Save the data through the "Save Data" component and export it to GIS software to achieve cross-platform linkage of "Orange + GIS" spatial analysis as Figure 6 shown.

[0156] Project the generated SHAP value diagram, price curve diagram, price frequency diagram, and land price map onto a large-screen display, connect a 3D projection device to OpenAI GPT-4 to construct a digital human, and the digital human can conduct in-depth analysis and interaction for users based on the result diagrams.

[0157] In summary, in view of the deficiencies in the prior art, the embodiment of the present invention proposes an automatic valuation method for urban agricultural land transfer prices, which can achieve the following technical effects: First, provide an objective index screening method: The embodiment of the present invention objectively evaluates and ranks each index through quantitative methods such as univariate regression and RReliefF value, selects the indexes that are more important for model training, reduces the influence of subjective judgment, and improves the objectivity and accuracy of the valuation results; Second, integrate multiple machine learning models and ensemble learning models: The embodiment of the present invention adopts 4 classic machine learning models, namely K-nearest neighbor (KNN), Adaptive Boosting (AdaBoost), Random Forest (RF), and Extreme Gradient Boosting (XGBoost), and constructs 3 ensemble learning models with excellent performance, namely the AdaBoost-RF stacked ensemble model (Stack-AR), the AdaBoost-RF-XGBoost stacked ensemble model (Stack-ARX), and the AdaBoost-RF-XGBoost-KNN stacked ensemble model (Stack-ARXK). Through training and evaluation by the ten-fold cross-validation method, the optimal model is selected to automatically value the price, improving the accuracy and stability of the model; Through the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R 2)Comprehensively evaluate the model performance to ensure the reliability of the model; use a completely untrained validation set to test the optimal model, ensuring the generalization ability of the final prediction model. Thirdly, establish models for different land types separately. For urban agricultural lands of different land types such as dry land and paddy field, parallel and non-overlapping workflows are used respectively to obtain the optimal models, taking into account the heterogeneity of urban agricultural land transfers of different land types and improving the applicability of the model. Fourthly, enhance the interpretability of the model: by introducing an intuitive and fast SHAP value calculation method, visualize the contribution degree of different indicators to the model, which helps to explore the dominant driving valuation indicators of urban agricultural land transfer prices of different types and provide scientific decision-making support for urban agricultural land market management, overcoming the limitations of traditional machine learning models that are difficult to explain based on their own "black box" interactions. Fifthly, "reduce costs and increase efficiency" in urban agricultural land valuation: traditional valuation methods expose limitations such as strong subjectivity, high costs, and low efficiency. The present invention provides empirical experience for realizing simple, fast, and efficient automatic valuation of urban agricultural land transfer prices. As long as the corresponding workflow is built, refined automatic valuation of the entire region's valuation units can be achieved faster. Sixthly, innovatively introduce the application scenario of "generative AI + digital human", enabling users to have an efficient conversation with the digital human, realizing intelligent analysis of the prediction results, and optimizing the user experience.

[0158] Based on the same inventive concept, as Figure 2 shown, an embodiment of the present invention further provides an automatic valuation device for urban agricultural land transfer prices, including:

[0159] An index score calculation module 100, configured to construct a multi-dimensional and multi-source valuation index system and calculate the index scores of the valuation indexes in each valuation unit;

[0160] A transaction price preprocessing module 200, configured to collect and preprocess the data of the actually transacted urban agricultural land transfer prices;

[0161] A model construction module 300, configured to construct a multi-machine learning model integrating the Stacking integration algorithm;

[0162] A sample point construction module 400, configured to use the actually transacted urban agricultural land transfer price of the valuation unit as the dependent variable and the index scores of multiple valuation indexes of the valuation unit as the independent variables to construct multiple groups of sample point data, and divide the multiple groups of sample point data into a training set and a validation set according to a preset ratio;

[0163] A model training module 500, configured to input the training set into each machine learning model for training, screen out the optimal model, and use the validation set for verification to obtain the optimal model that passes the verification;

[0164] An automatic valuation module 600, configured to use the optimal model that passes the verification to perform automatic valuation of urban agricultural land transfer prices.

[0165] It should be noted that for the technical details not described in detail in the embodiments of this urban agricultural land transfer price automatic valuation device, reference may be made to the application of an urban agricultural land transfer price automatic valuation method as described above in any embodiment of the present invention, which will not be elaborated here.

[0166] Based on the same inventive concept, an embodiment of the present invention further provides an urban agricultural land transfer price automatic valuation device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the urban agricultural land transfer price automatic valuation method described above are implemented.

[0167] In some embodiments, the processor may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. The processor is generally used to control the overall operation of the electronic device. In this embodiment, the processor is used to run the program code stored in the memory or process data, such as the program code of the vehicle collision warning method.

[0168] The memory includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (such as SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory may be an internal storage unit of the electronic device, such as the hard disk or memory of the electronic device. In other embodiments, the memory may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device. Of course, the memory may also include both the internal storage unit and the external storage device of the electronic device. In this embodiment, the memory is generally used to store the operation methods and various application software installed in the electronic device, such as the program code of the vehicle collision warning method. In addition, the memory may also be used to temporarily store various data that have been output or will be output.

[0169] Based on the same inventive concept, the present invention also provides a readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the steps of the XXX method described above are implemented.

[0170] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, apparatus, or computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.

[0171] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (apparatuses), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0172] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device realizes the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0173] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, so that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable devices provide steps for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0174] It can be understood that the above embodiments are merely exemplary embodiments adopted to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also regarded as the protection scope of the present invention.

Claims

1. An automatic valuation method for the transfer price of urban agricultural land, characterized in that, It includes the following steps: Construct a multi-dimensional and multi-source valuation index system, and calculate the index scores of the valuation indexes in each valuation unit; Collect the actually transacted urban rural land transfer price data and preprocess it; Construct a multi-machine learning model integrating the Stacking integration algorithm; Taking the actually transacted urban rural land transfer price of the valuation unit as the dependent variable and the index scores of the valuation indexes of the valuation unit as the independent variables, construct multiple sets of sample data, and divide the multiple sets of sample data into a training set and a validation set according to a preset ratio; Input the training set into each machine learning model for training, screen out the optimal model, and use the validation set for verification to obtain the optimal model that passes the verification; Use the optimal model that passes the verification for automatic valuation of urban rural land transfer prices.

2. The automatic valuation method for the transfer price of urban agricultural land according to claim 1, wherein The construction of the multi-dimensional and multi-source valuation index system and the calculation of the index scores of the valuation indexes in each valuation unit specifically include: Construct a "natural - economic - social - ecological" multi-dimensional and multi-source data index system. According to the influence radius of the valuation index source data, use a spatial representation method based on the valuation unit to calculate the global distance decay scores, and perform spatial superposition on the scores calculated according to each index source data according to the preset weights to obtain the index scores of the valuation indexes in each valuation unit.

3. The automatic valuation method for the transfer price of urban agricultural land according to claim 1, wherein, After calculating the index scores of the valuation indexes in each valuation unit, it also includes: performing standardization processing on the index scores of all valuation indexes, ranking the importance of multiple valuation indexes, and retaining the most important n effective valuation indexes.

4. The automatic valuation method for the transfer price of urban agricultural land according to claim 1, characterized in that The preprocessing includes: removing outliers and extreme values from the actually transacted urban rural land transfer price data; And / or, for the urban rural land transfer price data of different trading years, calculate the land price index of each year and uniformly revise the prices of each year to the latest year.

5. The automatic valuation method for the transfer price of urban agricultural land according to claim 1, characterized in that, The construction of the multi-machine learning model integrating the Stacking integration algorithm specifically includes: constructing multiple preset single machine learning models and integrating them into multiple ensemble learning models using the Stacking algorithm.

6. The automatic valuation method for the transfer price of urban agricultural land according to claim 1, characterized in that: The input of the training set into each machine learning model for training and screening out the optimal model specifically includes: Using the grid search method to search for the optimal parameter range of the single machine learning model and perform parameter tuning; After parameter tuning, input the training set into the machine learning model and the ensemble learning model for training, use the ten-fold cross-validation method to evaluate the performance indicators, and screen out the optimal model.

7. The automatic valuation method for the transfer price of urban agricultural land according to claim 1, characterized in that, It also includes: Calculating the shap values of multiple valuation indexes in the optimal model and generating a shap value ranking graph.

8. The automatic valuation method for the transfer price of urban agricultural land according to claim 1, wherein It also includes: Performing artificial intelligence analysis on the valuation data of urban rural land transfer prices.

9. An automatic valuation device for the transfer price of urban agricultural land, characterized in that, It includes: An index score calculation module, which is used to construct a multi-dimensional and multi-source valuation index system and calculate the index scores of the valuation indexes in each valuation unit; A transaction price preprocessing module, which is used to collect the actually transacted urban rural land transfer price data and preprocess it; A model construction module, which is used to construct a multi-machine learning model integrating the Stacking integration algorithm; The sample point construction module is used to construct multiple sets of sample point data with the actual transaction price of urban rural land transfer in the valuation unit as the dependent variable and the index scores of the valuation indicators of the valuation unit as the independent variable, and divide the multiple sets of sample point data into a training set and a validation set according to a preset ratio; The model training module is used to input the training set into each machine learning model for training, select the optimal model, and use the validation set for verification to obtain the optimal model that passes the verification; The automatic valuation module is used to automatically value the urban rural land transfer price using the optimal model that passes the verification.

10. An automatic valuation device for urban rural land transfer price, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the automatic valuation method for urban rural land transfer price as described in claims 1-8.

Images