House intelligent property transfer application method and system

Abstract

The invention provides an intelligent house characterization and migration application method and system. The method comprises the following steps: extracting key features from multi-source heterogeneous data, and forming feature vectors of feature migration project risks of each house in a target feature migration area, wherein the association value of the key features and target feature migration project risks is greater than a preset association value; inputting the feature transfer project risk feature vector into a pre-trained random forest model, generating a risk probability value of each house in the target feature transfer area, and forming a visual risk thermodynamic diagram; and based on the visual risk thermodynamic diagram, corresponding differential compensation strategies are generated for houses with different risk probability values. According to the technical scheme provided by the invention, the risk thermodynamic diagram is generated by integrating and analyzing the multi-source characteristic migration data, the differential compensation strategy is intelligently formulated, accurate management and control of characteristic migration risks are realized, and the efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of house acquisition and relocation project management, and in particular to an application method and system for smart house acquisition and relocation. Background Art

[0002] In relocation project management, dynamic quantitative assessments of relocation risks for housing within a region are necessary to address the complex correlations between multi-source, heterogeneous data. The core challenge is to use data-driven automation to identify the causal relationship between high-dimensional features and project risks, accurately predict the risk probability of individual housing units, intuitively visualize the spatial distribution of risks, and develop a basis for differentiated compensation strategies.

[0003] The existing scheme adopts a resettlement risk assessment model based on logistic regression, manually presets limited features from the basic data of house attributes and right holders, calculates the binary risk probability through the logistic regression model, and outputs the result as a risk level table or static distribution map.

[0004] However, existing solutions have the problem of being unable to effectively learn the complex nonlinear interaction features between multi-source heterogeneous data using logistic regression models, resulting in the omission of key risk factors; moreover, static tables cannot reflect the continuous distribution pattern of risk probabilities in geographic space, making it difficult to locate regional risk concentration points; in addition, model performance is limited by preset feature dimensions and does not incorporate a data-driven automatic feature screening mechanism, resulting in low sensitivity to hidden risk patterns. Summary of the Invention

[0005] The present application provides a smart housing relocation application method and system to solve the problems in the existing technology such as the failure of nonlinear interactive feature recognition, the lack of expression of risk spatial distribution rules, and the strong dependence on manual feature screening mechanisms, which lead to the inability to accurately control housing relocation applications.

[0006] In the first aspect, this application provides a smart housing relocation application method, including: Collect multi-source heterogeneous data in the target relocation area, including housing attribute data, right holder characteristic data, compensation negotiation data, and environmental related data; Extracting key features whose correlation values ​​with the target resettlement project risk are greater than a preset correlation value from the multi-source heterogeneous data to form a resettlement project risk feature vector for each house in the target resettlement area; Input the risk feature vector of the resettlement project into a pre-trained random forest model to generate a risk probability value for each house in the target resettlement area and form a visual risk heat map; Based on the visualized risk heat map, corresponding differentiated compensation strategies are generated for houses with different risk probability values.

[0007] Optionally, the risk feature vector of the resettlement project is input into a pre-trained random forest model to generate a risk probability value for each house in the target resettlement area to form a visual risk heat map, including: Inputting the risk feature vector of the resettlement project for each house in the target resettlement area into multiple decision trees in a pre-trained random forest model to obtain the risk judgment results output by each decision tree; Aggregate all risk judgment results of a single house to generate a risk probability value for the single house to form a risk probability value set; Associating each risk probability value in the risk probability value set with the coordinate position of the corresponding house; Determine the color depth identifier corresponding to the risk probability value of the house based on a preset mapping relationship between the risk probability value and the color depth identifier; The coordinate positions of all houses and the corresponding color depth labels are superimposed on the regional map to form a visual risk heat map.

[0008] Optionally, the risk judgment result is at least one of a risk label, a risk score, and a risk level; Aggregating all risk assessment results for a single house to generate a risk probability value for the single house includes: Gather all risk assessment results of a single house to generate a risk assessment result set for the house; Counting the number of risk judgment results that meet a preset high-risk condition in the risk judgment result set of the house; the predetermined high-risk condition comprising at least one of the following: the risk label type is high risk, the risk score exceeds a preset score threshold, and the risk level is ranked in the top N levels; N is an integer greater than or equal to 1; The ratio of the number of risk judgment results to the number of decision trees in the random forest model is used as the risk probability value of a single house.

[0009] Optionally, different decision trees output risk judgment results corresponding to the same risk feature vector of the resettlement project; The risk feature vector of the resettlement project is input into multiple decision trees in a pre-trained random forest model to obtain the risk judgment results output by each decision tree, including: Inputting the risk feature vector of the resettlement project into all decision trees of the random forest model simultaneously; In each decision tree, the feature values ​​are recursively matched according to the splitting rules of the tree structure, and finally the leaf nodes are reached, and the risk judgment results are extracted from the leaf node attributes.

[0010] Optionally, different decision trees output risk judgment results corresponding to different key features; The risk feature vector of the resettlement project is input into multiple decision trees in a pre-trained random forest model to obtain the risk judgment results output by each decision tree, including: According to the preset decision tree feature mapping table, determine the exclusive feature set corresponding to each decision tree; Extracting each exclusive feature set from the risk feature vector of the resettlement project; Input each unique feature set into the corresponding decision tree; In each decision tree, node splitting decisions are performed based on the input feature subset until a leaf node is reached, and the risk judgment results are extracted from the leaf node attributes.

[0011] Optionally, determining the color depth identifier corresponding to the risk probability value of the house based on a preset mapping relationship between the risk probability value and the color depth identifier includes: The HSV value is determined by using the mapping relationship between the range of the risk probability value and the HSV color space established based on the nonlinear mapping function; Based on the color conversion relationship, the HSV value is converted into the RGB color space to obtain the RGB value; Gamma correction is performed on the RGB value, and the corrected RGB value is encoded into a hexadecimal color code as the color depth identifier.

[0012] Optionally, generating corresponding differentiated compensation strategies for houses with different risk probability values ​​based on the visualized risk heat map includes: Based on the distribution of housing risk probability values ​​in the visualized risk heat map, a preset fracture algorithm is applied to calculate a demarcation threshold, and the risk interval to which each housing belongs is determined according to the demarcation threshold; For each house, generating initial compensation strategy parameters according to the data corresponding to the house in the multi-source heterogeneous data; The differentiated compensation strategy parameters corresponding to the risk interval and the initial compensation strategy parameters are integrated to obtain a differentiated compensation strategy including the integrated compensation strategy parameters.

[0013] Secondly, this application provides a smart housing relocation application system, including: A collection module is used to collect multi-source heterogeneous data in the target relocation area, wherein the multi-source heterogeneous data includes housing attribute data, right holder characteristic data, compensation negotiation data, and environmental related data; An extraction module is used to extract key features whose correlation value with the risk of the target resettlement project is greater than a preset correlation value from the multi-source heterogeneous data, so as to form a resettlement project risk feature vector for each house in the target resettlement area; An input module is used to input the risk feature vector of the resettlement project into a pre-trained random forest model to generate a risk probability value for each house in the target resettlement area and form a visual risk heat map; A generation module is used to generate corresponding differentiated compensation strategies for houses with different risk probability values ​​based on the visual risk heat map.

[0014] In a third aspect, the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a smart house relocation application method as described in any one of the first aspects.

[0015] In a fourth aspect, the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements a smart house acquisition and relocation application method as described in any one of the first aspects.

[0016] In the present application, a smart housing expropriation and relocation application method is provided, which includes: collecting multi-source heterogeneous data of a target expropriation and relocation area, the multi-source heterogeneous data including housing attribute data, right holder characteristic data, compensation negotiation data and environmental correlation data; extracting key features whose correlation values ​​with the target expropriation and relocation project risks are greater than preset correlation values ​​from the multi-source heterogeneous data, so as to form an expropriation and relocation project risk feature vector for each house in the target expropriation and relocation area; inputting the expropriation and relocation project risk feature vector into a pre-trained random forest model to generate a risk probability value for each house in the target expropriation and relocation area, and forming a visual risk heat map; based on the visual risk heat map, generating corresponding differentiated compensation strategies for houses with different risk probability values.

[0017] This application achieves comprehensive coverage of risk factors by collecting multi-source heterogeneous data, automatically extracts highly correlated features to construct housing risk vectors; uses random forest models to accurately quantify individual risk probabilities and generate spatial heat maps, intuitively revealing regional risk distribution patterns; and finally dynamically generates differentiated compensation strategies based on visualization results to achieve scientific prediction of resettlement risks and precise resource allocation.

[0018] Furthermore, this application processes housing risk feature vectors in parallel through multiple decision trees of a random forest model to output diverse risk assessment results. The application generates a risk probability value by counting the percentage of houses that meet preset high-risk conditions in the results. After associating the geographic coordinates, the housing data is overlaid onto the regional map according to the probability value-color depth mapping rule to dynamically generate a risk heat map. The high-risk result statistics method based on dynamic thresholds improves the interpretability of probability value calculations. The spatial mapping of geographic coordinates and probability values, combined with gradient color depth visualization, reveals the regional clustering patterns of resettlement risks in real time, providing a spatial decision-making basis for the formulation of differentiated strategies.

[0019] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0021] Figure 1 A flowchart of a smart housing relocation application method provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of a smart housing relocation application system provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0023] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 11, 12, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0024] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0025] In order to solve the problems in the existing technology that cannot accurately control house acquisition and relocation applications due to the failure of nonlinear interactive feature recognition, the lack of expression of risk spatial distribution rules, and the strong dependence on artificial feature screening mechanisms, the embodiment of the present application provides a smart house acquisition and relocation application method, which adopts the following ideas: by integrating multi-source heterogeneous data to construct a panoramic risk factor library, and using correlation threshold screening to eliminate noise interference to form a highly representative feature vector; introducing a random forest model to solve the problem of identifying nonlinear feature coupling effects, and outputting probability values ​​based on multi-tree integrated decision-making; realizing the visualization of risk spatial distribution through geographic coordinate mapping and gradient color depth conversion, and finally generating a compensation strategy closed loop based on the risk gradient of the heat map, to achieve scientific management and control of the entire chain from data to decision-making.

[0026] Figure 1 A flowchart of a housing smart relocation application method provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes: S11. Collect multi-source heterogeneous data in the target resettlement area, including housing attribute data, right holder characteristic data, compensation negotiation data and environmental related data.

[0027] Among them, housing attribute data can refer to a set of physical parameters that reflect the status of the relocated housing itself, including the type of building structure, construction year, property area, current use nature and ownership certificate integrity indicators. Owner characteristic data refers to the data dimensions used to describe the social attributes of the housing owner, covering age, occupation type, family population structure, historical records and social relationship network strength. Compensation negotiation data refers to a dynamic information set that records the negotiation process of all relocations, including the fluctuation trajectory of the compensation claim amount, the number of negotiation breakdowns, legal litigation history and special resettlement needs statements. Environmental-related data can refer to spatial parameters that characterize the environmental conditions surrounding the house, involving transportation network density, public facilities coverage, geological stability indicators and regional real estate price fluctuation trends.

[0028] In an embodiment of the present application, first, the housing attribute data, right holder characteristic data, compensation negotiation data and environmental related data of the target resettlement area are collected through the data interface of the resettlement management system, and then the format of the multi-source heterogeneous data is standardized using a data cleaning tool. Then, the data from different sources are associated and matched according to the unique identifier of the house through spatial geocoding, and finally, a fusion database containing the physical status of the house, the social attributes of the right holder, historical negotiation records and surrounding environmental elements is constructed.

[0029] S12. Extract key features whose correlation values ​​with the target resettlement project risk are greater than preset correlation values ​​from multi-source heterogeneous data to form a resettlement project risk feature vector for each house in the target resettlement area.

[0030] The correlation value of the target resettlement project risk refers to the quantitative correlation strength between a single feature and the probability of resettlement failure calculated through correlation analysis, and is represented by the Pearson correlation coefficient or mutual information entropy value. The preset correlation value can refer to a correlation coefficient threshold set based on manual experience or model validation, which is used to filter out the set of strong risk influencing factors retained after filtering out low-correlation features. The resettlement project risk feature vector can refer to a numerical matrix composed of the screened key features arranged in a fixed dimension, with each vector element corresponding to a standardized feature observation value.

[0031] In an embodiment of the present application, the Pearson correlation coefficient between each feature in the house attribute data, right holder characteristic data, compensation negotiation data and environmental association data and the target resettlement project risk is first calculated, and then the key features with correlation coefficients greater than the preset correlation value are screened. The screened features are then normalized to eliminate dimensional differences, and finally the key features are aggregated according to the house unit to generate a fixed-length numerical resettlement project risk feature vector.

[0032] S13. Input the risk feature vector of the resettlement project into the pre-trained random forest model to generate the risk probability value of each house in the target resettlement area and form a visual risk heat map.

[0033] The random forest model refers to an ensemble learning algorithm composed of multiple decision trees. It trains subtrees through sampling and uses a majority voting mechanism to output a final prediction. The risk probability value refers to the statistical probability of a single house being classified as a high-risk case. It is calculated by dividing the number of decision trees marked as high-risk by the total number of decision trees in the random forest. A visual risk heat map is a spatial density distribution map generated on a digital map by mapping the geographic coordinates of houses to risk probability values ​​using a gradient color-depth mapping rule. The depth of red is positively correlated with the risk probability.

[0034] In an embodiment of the present application, the risk feature vector of the resettlement project of each house is first input in parallel into multiple decision trees of the random forest model. Secondly, each decision tree independently outputs a risk judgment result in the form of a risk label, a risk score, or a risk level. Subsequently, the ratio of the number of high-risk marks in all decision tree results of a single house to the total number of decision trees is counted to generate a risk probability value. Finally, the risk probability value is associated with the geographical coordinates of the house and rendered on the regional map through the color depth mapping rule to generate a visual risk heat map.

[0035] S14. Based on the visualized risk heat map, generate corresponding differentiated compensation strategies for houses with different risk probability values.

[0036] Among them, the differentiated compensation strategy refers to a set of compensation plans that are dynamically adjusted according to the risk probability level, including the floating coefficient of the basic compensation amount, the priority signing bonus amount and the customized conflict mediation plan.

[0037] In an embodiment of the present application, the risk probability value intervals corresponding to different color depth areas in the visual risk heat map are first analyzed, and then the house risk level is matched according to the preset low-risk interval, medium-risk interval, and high-risk interval division rules. Then, the compensation calculation coefficients and negotiation process templates bound to each risk level in the compensation strategy knowledge base are called, and finally a compensation strategy set is generated that includes differentiated compensation amounts, priority negotiation sequences, and special conflict resolution solutions.

[0038] To summarize, the relocation management platform first accesses the real estate registration database to obtain information on the structure and ownership of the property. Simultaneously, the system collects data on the family composition of the property owner and retrieves historical negotiation recordings to convert them into structured conflict records. Satellite remote sensing imagery is then integrated to analyze surrounding geological parameters. Next, the correlation coefficient between each feature and the probability of relocation delay is calculated, and features exceeding a threshold of 0.3 are retained to form the property feature vector. This feature vector is then fed into a trained random forest model. The proportion of high-risk items in 500 decision trees is calculated as the probability value. The longitude and latitude coordinates of the properties are then linked and mapped to red, yellow, blue, and dark colors according to the probability range. Finally, based on the properties in the red areas of the heat map, the base compensation amount is increased by 15%, and a legal advisor intervention mechanism is implemented. A fast-track signing incentive channel is implemented for properties in the blue areas.

[0039] The following is a specific example: This method first collects data from various sources and formats within the target relocation area, including information on the property itself (e.g., structural type, area, and age); information on the property owner (e.g., age, family structure, and historical records); past negotiation processes (e.g., changes in the requested amount and the number of failed negotiations); and surrounding environmental data (e.g., transportation, geology, and housing prices). The system then analyzes the correlation between each feature within this data (e.g., construction age, number of failed negotiations) and the risk of relocation project failure. By calculating the correlation coefficient and comparing it with a preset threshold (e.g., 0.3), it identifies key features that significantly influence risk, generating a numerical feature vector representing the risk profile of each property in the area. This feature vector is then fed into a pre-trained random forest model; this model consists of multiple (e.g., 500) decision trees, each of which independently determines a property's risk level or score based on its characteristics. Ultimately, by calculating the proportion of decision trees that consider a property to be high risk, a precise risk probability value is derived. For example, 0.84 indicates that 84% of the trees consider the property to be high risk. These calculated risk probability values ​​are associated with the geographic location of each house (i.e., longitude and latitude coordinates), and different probability value intervals are mapped to different color depths according to preset rules. For example, the darker the red, the higher the risk. Ultimately, a clear and intuitive risk heat map is overlaid on the map of the target area. Finally, the system identifies low-, medium-, and high-risk houses based on the risk distribution presented in the heat map. For example, it uses a breakpoint algorithm to automatically divide the intervals and generates a basic compensation plan based on the original multi-source data of the house. The system dynamically adds differentiated adjustment factors to houses of different risk levels, such as increasing the baseline compensation amount for high-risk houses by 15% and providing legal counsel intervention. Ultimately, a closed-loop compensation strategy is formed for each house, achieving scientific management of the entire chain from data collection to precise decision-making.

[0040] By executing S11 to S14, the embodiment of the present application constructs a panoramic risk factor library by integrating multi-dimensional resettlement data, eliminates noise interference based on correlation threshold screening to form a highly representative feature vector; uses the integrated decision-making mechanism of random forest to accurately quantify the risk probability of individual houses, and intuitively presents the regional risk distribution pattern through spatial heat maps; finally, dynamically generates compensation strategies based on risk gradients to achieve scientific optimization of resettlement resource allocation and improve the efficiency of conflict resolution.

[0041] In one possible embodiment, S13 inputs the risk feature vector of the resettlement project into a pre-trained random forest model to generate a risk probability value for each house in the target resettlement area, forming a visual risk heat map, including: Step 131: For each house resettlement project risk feature vector in the target resettlement area, the resettlement project risk feature vector is input into multiple decision trees in a pre-trained random forest model to obtain the risk judgment result output by each decision tree.

[0042] Multiple decision trees can refer to multiple weak classifiers trained through sampling and random feature selection in a random forest model. Each tree can make node splitting decisions based on the Gini coefficient or information gain. The risk assessment result refers to the prediction output by a single decision tree for a property's feature vector, including binary risk labels such as high or low risk, continuous risk scores, or discrete risk level codes.

[0043] In an embodiment of the present application, the risk feature vector of the resettlement project of a single house in the target resettlement area is first synchronously input into hundreds of independent decision trees pre-trained in the random forest model. Secondly, each decision tree recursively judges the input feature vector based on its internal division rules, and then each decision tree outputs the individual risk judgment result in the form of a risk label or risk score or risk level corresponding to the house.

[0044] Step 132: Aggregate all risk judgment results of a single house to generate a risk probability value of the single house to form a risk probability value set.

[0045] The risk probability value set may refer to a data structure storing the risk probability values ​​of all houses in the area, where each element consists of a unique house code and its corresponding risk probability value.

[0046] In an embodiment of the present application, the risk judgment results generated by all decision trees corresponding to a single house are first collected to form a temporary result set for the house. Secondly, the number of judgment results marked as high-risk in the temporary result set is counted, and then the number of high-risk results is divided by the total number of decision trees in the random forest model. Finally, the calculated value is used as the risk probability value of the house and added to the regional risk probability value set.

[0047] Step 133: Associate each risk probability value in the risk probability value set with the coordinate position of the corresponding house.

[0048] In an embodiment of the present application, the risk probability value corresponding to each house in the risk probability value set is first read, and then the house geographic information database is called to retrieve the latitude and longitude coordinates of the house, and then a key-value pair mapping relationship between the house risk probability value and the latitude and longitude coordinates is established, and finally a spatialized data table containing the house location identifier and risk probability value is generated.

[0049] Step 134 : Based on a preset mapping relationship between the risk probability value and the color depth identifier, determine the color depth identifier corresponding to the risk probability value of the house.

[0050] Among them, the color depth identification refers to the standardized color code assigned according to the risk probability value interval, which is used to represent the risk intensity in the heat map. The red identification code corresponds to the high-risk interval, and the blue identification code corresponds to the low-risk interval.

[0051] In an embodiment of the present application, a preset continuous color depth mapping rule table is first loaded, and then the interval to which the current house risk probability value belongs is determined according to the risk probability value segmentation threshold defined in the rule table, and then the standard color depth identification code corresponding to the interval is matched, and finally the identification code is associated with the corresponding record of the house spatialization data table.

[0052] Step 135: Overlay the coordinate positions of all houses and the corresponding color depth identifiers onto the regional map to form a visual risk heat map.

[0053] Among them, the regional map includes a digitized vector map of the geographical boundaries, road network and building outlines of the target resettlement area, using the latitude and longitude coordinate system as the spatial reference benchmark.

[0054] In an embodiment of the present application, a digital vector base map of the target resettlement area is first imported as a rendering base, and then the latitude and longitude coordinates of each house and its color depth identification code in the spatialized data table are traversed, and then circular heat points of specified color values ​​are drawn at the corresponding coordinate positions. Finally, a risk heat map with continuous gradient color depth is generated through a spatial interpolation algorithm.

[0055] Here's a specific example: During the risk assessment phase, the system feeds the relocation risk feature vectors prepared for each house in the area into each decision tree (e.g., 500 trees) in the random forest model. Each decision tree contains a series of splitting rules learned from training data. Based on the input house feature values, it makes decisions layer by layer along its tree structure, ultimately reaching a leaf node. This leaf node contains a pre-defined risk assessment result, which may be a simple "high risk / low risk" label or a specific risk score or level. After all decision trees have output their results for the current house, the system aggregates these results. For each house, the system counts the number of decision trees that were classified as "high risk" (or the number of results that met other pre-defined high-risk criteria). This number is then divided by the total number of decision trees in the model (e.g., 350 high-risk trees divided by 500 trees) to obtain the final risk probability value for the house (e.g., 0.7). The system then matches these risk probability values ​​with the geographic coordinates of the houses in the database, forming a dataset containing spatial location information. Next, based on predefined mapping relationships, such as a risk probability of 0.7 falling into the high-risk range, the system determines the color depth identifier for that house on the map, such as the RGB value corresponding to dark red. Finally, the system reads the coordinates and calculated color identifiers of all houses, overlays these point information onto a digital base map of the target relocation area, and uses interpolation technology to generate a color heat map that continuously displays the risk level of the entire area.

[0056] In another specific example, the feature vector of a house in an urban village renovation area is first input into a random forest model containing 300 decision trees, and each tree outputs a high-risk or low-risk label. Secondly, the number of decision trees that obtained a high-risk label for the house is counted as 210. 210 is divided by 300 to obtain a risk probability value of 0.7 and stored in a set. The longitude and latitude coordinates of the house in the geographic information system are then associated. Then, the dark red identification code corresponding to the high-risk interval to which 0.7 belongs in the rule table is matched. Finally, a dark red heat point is rendered at this coordinate position on the regional digital map, and a risk heat map covering the entire area is generated through kriging interpolation.

[0057] By executing steps 131 to 135, the embodiment of the present application enhances the robustness of risk assessment through parallel decision-making of multiple decision trees, and generates interpretable risk probability values ​​based on a statistical aggregation mechanism; spatial transformation of risk data is achieved through coordinate association, and abstract probability values ​​are converted into intuitive and visual heat maps in combination with gradient color depth mapping rules, providing spatial decision support for resettlement management.

[0058] In one possible embodiment, the risk judgment result is at least one of a risk label, a risk score, and a risk level. Step 132: Aggregate all risk judgment results for a single house to generate a risk probability value for the single house, including: Step a1: Gather all risk assessment results of a single house to generate a risk assessment result set for the house.

[0059] The risk judgment result set may refer to a container data structure that stores all decision tree output results corresponding to a single house, including the decision tree number, output result type, and specific value or label.

[0060] Step a2: Count the number of risk assessment results in the set of housing risk assessment results that meet a preset high-risk condition. Meeting the preset high-risk condition includes at least one of the following: the risk tag type is high risk, the risk score exceeds a preset score threshold, and the risk level is ranked in the top N levels. N is an integer greater than or equal to 1.

[0061] Among them, the preset high-risk conditions can refer to pre-set high-risk judgment logic rules, including any of the conditions that the risk label is a specific high-risk enumeration value, the risk score is greater than a floating threshold, and the risk level is in the top N discrete sequences. The number of risk judgment results can refer to the count of qualified results obtained after condition screening, which can reflect the strength of consensus on the high-risk status of the house in the decision tree group. The risk label refers to the discrete classification conclusion output by the decision tree, which usually represents the risk status of the house with a high-risk or low-risk binary enumeration value. The risk score refers to the continuous prediction value generated by the decision tree, and the score is positively correlated with the severity of the risk.

[0062] Step a3: The ratio of the number of risk judgment results to the number of decision trees in the random forest model is used as the risk probability value of a single house.

[0063] The number of decision trees refers to the total number of independent decision trees generated through training in the random forest model, which determines the stability and statistical cardinality of the integrated prediction.

[0064] Here is a specific example: In order to derive the final risk probability value of a house from the results of each decision tree, the system will first collect the output results of all decision trees corresponding to a certain house, such as labels, scores, and levels, to form a special result set. The system will then analyze each result in this set and perform statistics based on the pre-set "high-risk judgment conditions." These conditions can be flexible: for example, the judgment result is a "high risk" label; or a risk score that exceeds a certain score threshold (such as 80 points); or a risk level that ranks in the top few levels (such as the top 3 levels). Results that meet any of the preset high-risk conditions will be counted. Finally, the number of qualified judgment results, such as 400 trees that meet any of the above conditions, is divided by the total number of model decision trees, such as 500 trees. The resulting ratio of 0.8 is the risk probability value of this house.

[0065] By executing steps a1 to a4, the embodiment of the present application realizes unified management of multi-tree judgment results through collective storage, and dynamically screens high-risk consensus results based on configurable conditions; generates probability values ​​using the ratio of statistical results to the total amount, and provides robust risk assessment against single-tree bias; realizes the scientific transformation from discrete decision-making to continuous probability, and enhances the interpretability and reliability of model predictions.

[0066] In one possible embodiment, different decision trees output risk judgment results corresponding to the same resettlement project risk feature vector; step 131, inputting the resettlement project risk feature vector into multiple decision trees in a pre-trained random forest model, and obtaining the risk judgment results output by each decision tree, including: Step b1: Input the risk feature vector of the resettlement project into all decision trees of the random forest model at the same time.

[0067] Among them, all decision trees in the random forest model refer to a set of decision trees generated by resampling and random feature selection. Each tree is built based on differentiated training subsets and splitting rules, and supports synchronous prediction through a parallel architecture.

[0068] Step b2: In each decision tree, recursively match the feature values ​​according to the splitting rules of the tree structure, and finally reach the leaf node, and extract the risk judgment result from the leaf node attributes.

[0069] Among them, the tree structure refers to the hierarchical topology composed of nodes and edges in the decision tree, which includes three elements: root nodes, internal nodes and leaf nodes, and the nodes are connected by feature splitting paths. The splitting rule refers to the feature partitioning logic generated at each internal node during decision tree training, including the selected feature index, splitting threshold and the left and right subtree allocation direction, which is used to guide the feature vector traversal path during prediction. Feature refers to the numerical element in the input vector that represents the specific risk factor of the house, and the value at the index position specified by the splitting rule during the decision tree matching process. Leaf node refers to the terminal node with no child nodes at the end of the decision tree, which stores the category distribution or regression value obtained by statistics in the training phase as the prediction output benchmark. Leaf node attributes refer to the risk prediction result data encapsulated in the leaf node, including the mode category of the risk label, the mean of the risk score or the distribution probability of the risk level.

[0070] Here's a specific example: First, a copy of the house's feature vector is synchronously transmitted to the input ports of 500 decision trees. Next, in one decision tree, the feature value at index 3 is matched from the root node. Based on a threshold greater than 0.5, the flow proceeds to the right subtree. At the second-level node, the feature value at index 7 is matched. Based on a threshold less than 1.2, the flow proceeds to the left subtree. Finally, the tree reaches the leaf node and returns the stored high-risk label. After completing this recursive matching process, all decision trees output their respective risk assessment results.

[0071] By executing steps b1 to b2, the embodiment of the present application realizes multi-tree collaborative prediction through parallel distribution of feature vectors, greatly improving computing efficiency; the recursive matching mechanism based on the tree structure ensures that the single-tree decision process strictly follows the training logic, and the statistical attributes extracted from the leaf nodes ensure the traceability and model consistency of the prediction results.

[0072] In another possible embodiment, different decision trees output risk assessment results corresponding to different key features. Step 131: input the risk feature vector of the resettlement project into multiple decision trees in a pre-trained random forest model to obtain the risk assessment results output by each decision tree, including: Step c1: Determine the exclusive feature set corresponding to each decision tree according to the preset decision tree feature mapping table; The pre-set decision tree feature map refers to a relational table that records the indexes of the feature subsets randomly assigned to each decision tree during training, reflecting that a specific decision tree only responds to a subset of the dimensions of the full feature vector. A dedicated feature set refers to a feature subset customized for a single decision tree, containing the feature values ​​at the specific index positions that the tree's splitting rules depend on, and has a smaller dimension than the full feature vector.

[0073] In an embodiment of the present application, the decision tree feature mapping table generated during the random forest model training phase is first loaded, and then the correspondence between the decision tree number and the feature index recorded in the mapping table is determined, and finally the exclusive feature set to be used by each decision tree during prediction is determined.

[0074] Step c2: extracting each exclusive feature set from the risk feature vector of the resettlement project; Among them, in the embodiment of the present application, the risk feature vector of the relocation project of the target house is first read, and then the feature index position contained in the exclusive feature set is followed, and then the feature value of the corresponding index is extracted from the full feature vector, and finally the feature subset that matches the decision tree one by one is packaged and generated.

[0075] Step c3: input each unique feature set into the corresponding decision tree; Among them, in the embodiment of the present application, a transmission channel between the decision tree number and the exclusive feature set is first established, and then the generated feature subset is input into the decision tree with the corresponding number through the designated channel, and then the prediction calculation process of the decision tree is triggered.

[0076] Step c4: In each decision tree, node splitting decisions are performed based on the input feature subset until a leaf node is reached, and risk judgment results are extracted from the leaf node attributes.

[0077] Among them, node splitting decision refers to the binary judgment process of selecting the left subtree or right subtree path by comparing the input feature values ​​based on the feature index and splitting threshold stored in the current node during decision tree prediction.

[0078] In an embodiment of the present application, first, the splitting rule of the root node is matched according to the input feature subset within a single decision tree, and then the subtree branch is selected based on the comparison result between the feature value and the threshold, and then the node splitting decision is recursively executed until the leaf node is reached, and finally the risk judgment result pre-stored in the leaf node attribute is extracted.

[0079] The following is a specific example: When a single decision tree performs predictions, the model employs an optimization mechanism: During the model training phase, each decision tree is randomly assigned a capability to focus on a subset of the full set of resettlement risk features. During prediction, the system first consults the decision tree feature map to determine the unique feature set corresponding to the decision tree being predicted, for example, tree number 105. For example, this includes features with indices 2, 5, and 7. Then, from the complete feature vector of the current house, the system extracts the feature values ​​specified by this unique feature set, for example, 0.62, 0.90, and 0.45, to form a small feature subset containing only these features. This feature subset is then fed into its corresponding decision tree (tree number 105) for judgment. The decision tree uses its internal, trained splitting rules to perform node judgments on this input feature subset. For example, at the root node, it checks whether the value of feature index 2 is greater than 0.5, traversing downwards layer by layer until a leaf node is reached. Finally, the tree’s risk judgment result for the house, such as “high risk level”, is extracted from the attributes of the leaf node (which contains statistical information from the training phase).

[0080] By executing steps c1 to c4, the embodiment of the present application realizes dynamic binding of the decision tree and the exclusive feature subset through the feature mapping table, thereby reducing the computational complexity of a single tree; accurate extraction based on feature indexes ensures the effective execution of splitting rules, outputs reliable risk judgments from leaf node attributes, and improves the efficiency of large-scale prediction tasks and model interpretability.

[0081] In a possible embodiment, step 134, determining the color depth identifier corresponding to the risk probability value of the house based on a preset mapping relationship between the risk probability value and the color depth identifier, includes: Step d1: Determine the HSV value by using the mapping relationship between the range of the risk probability value and the HSV color space established based on the nonlinear mapping function.

[0082] The nonlinear mapping function refers to the conversion rules between risk probability values ​​and color parameters constructed through piecewise polynomials or logarithmic functions, which ensures that the color difference changes smoothly in low-probability intervals and dramatically in high-probability intervals, consistent with the human eye's perception characteristics. The HSV color space refers to a model system that describes color in three dimensions: hue, saturation, and lightness. Hue represents color type, saturation represents color purity, and lightness controls color intensity. HSV values ​​refer to tuple data consisting of the hue component value, saturation component value, and lightness component value in the HSV color space, and are used to uniquely identify a specific color state.

[0083] In an embodiment of the present application, the risk probability value of the house is first obtained, and then the hue component and saturation component of the probability value in the HSV color space are calculated according to a preset nonlinear mapping function. Then, the lightness component is determined according to the low-risk interval or high-risk interval to which the probability value belongs, and finally the complete HSV color space value is generated by combining.

[0084] Step d2: Based on the color conversion relationship, convert the HSV value into the RGB color space to obtain the RGB value.

[0085] The color conversion relationship refers to the HSV to RGB standard conversion algorithm based on three-dimensional geometric transformations, which achieves equivalent mapping of color models through color gamut coordinate system conversion. The RGB color space refers to a model system that describes color based on the superposition principle of the three primary colors red, green, and blue. Each color is composed of a mixture of red, green, and blue component values. An RGB value refers to a tuple of red, green, and blue component values ​​in the RGB color space, and its value range is an integer from 0 to 255.

[0086] In an embodiment of the present application, the three component values ​​of the HSV color space value are first read, and then the HSV components are converted into the red, green and blue primary color components of the RGB color space through the standard color space conversion formula, and then the converted component values ​​are normalized, and finally the standard RGB color space value is generated.

[0087] Step d3: Perform gamma correction on the RGB values, and encode the corrected RGB values ​​into hexadecimal color codes as color depth identifiers.

[0088] Gamma correction refers to a preprocessing operation that applies a power function transformation to the RGB components based on the human eye's nonlinear perception of brightness to compensate for color distortion on display devices. A hexadecimal color code is a six-digit string of characters formed by converting the red, green, and blue components of an RGB value into two hexadecimal digits, respectively. It is used to represent color in a standardized way.

[0089] In an embodiment of the present application, gamma correction calculation is first performed on the three components of the RGB color space value respectively, and then the corrected red, green and blue components are converted into eight-bit binary values ​​in integer format, and then the three groups of binary values ​​are spliced ​​in the order of red, green and blue, and finally the splicing result is converted into a six-digit hexadecimal color code as the final color depth identifier.

[0090] Here's a specific example: the system converts a house's risk probability value into the colors needed for visualization using a method more consistent with human visual perception. First, the system uses a pre-defined nonlinear conversion function. For example, low-risk areas have gentle color changes, while high-risk areas have sharp changes. Based on the house's risk probability value (e.g., 0.85), the system determines the corresponding color parameter values ​​in the HSV color space (composed of three dimensions: H hue, S saturation, and V value): for example, H = 240° deep red, S = 90% high saturation, and V = 70% medium value. Next, using standard color conversion formulas, this HSV color model value is accurately converted to the red, green, and blue primary color components of the RGB color model commonly used by displays: for example, R = 122 points, G = 30 points, and B = 150 points. To improve color accuracy and appearance on display devices, the system also performs gamma correction on the RGB components, a nonlinear brightness adjustment tailored to the human eye's visual characteristics. Finally, the corrected red, green, and blue component values ​​are each converted into a two-digit hexadecimal number, and then concatenated in sequence to form the final, standard hexadecimal color code, such as #6E198C, which is used as the color depth identifier for the house rendered on the heat map.

[0091] By executing steps d1 to d3, the embodiment of the present application establishes a perceptually optimized association between risk probability and color parameters through nonlinear mapping, and uses the HSV space to achieve independent control of hue and saturation; standard conversion is used to ensure color reproduction consistency, and gamma correction is used to eliminate device display deviation, and finally a standardized color code is generated, so that the color gradient of the heat map accurately reflects the risk distribution law.

[0092] In a possible embodiment, S14, based on the visualized risk heat map, generating corresponding differentiated compensation strategies for houses with different risk probability values, including: Step 141: Based on the distribution of housing risk probability values ​​in the visualized risk heat map, a preset fracture algorithm is applied to calculate a demarcation threshold, and the risk interval to which each housing belongs is determined according to the demarcation threshold.

[0093] The preset break algorithm refers to a spatial data analysis algorithm based on the natural breakpoint classification method. It automatically calculates the optimal demarcation threshold by maximizing the similarity and difference within intervals. The demarcation threshold is the set of critical point values ​​that divide the continuous risk probability value into discrete intervals. It is determined by the breakpoint algorithm to identify significant mutations in the data distribution. The risk interval refers to the discrete classification range of the housing risk probability value, which includes three mutually exclusive subsets: low risk interval, medium risk interval, and high risk interval.

[0094] Step 142: For each house, generate initial compensation strategy parameters based on the data corresponding to the house in the multi-source heterogeneous data.

[0095] Among them, the initial compensation strategy parameters refer to the basic compensation elements generated based on the objective attributes of the house and the characteristics of the owner, including core parameters such as the benchmark compensation amount, special resettlement conditions, and statutory subsidy items.

[0096] Step 143: Fuse the differentiated compensation strategy parameters corresponding to the risk interval and the initial compensation strategy parameters to obtain a differentiated compensation strategy including the fused compensation strategy parameters.

[0097] Among them, differentiated compensation strategy parameters refer to dynamic adjustment factors bound to risk ranges, including risk-responsive parameters such as amount floating coefficient, priority signing reward amount, and legal intervention trigger conditions.

[0098] Here's a specific example: When generating a compensation strategy, the system first uses mathematical algorithms, such as the natural breakpoint method, to analyze the risk probability values ​​of all houses in the heat map. It then automatically finds thresholds that rationally divide the risk data into different intervals (e.g., low, medium, and high risk). For example, 0.3 is used as the threshold between low and medium risk, and 0.69 is used between medium and high risk. This determines the specific risk interval for each house. Simultaneously, the system calculates initial compensation strategy parameters, such as a base amount of 300,000 yuan and a legal aid designation, based on the raw, multi-source, heterogeneous data collected about the house, such as its size, the owner's specific needs, and historical conflict records. Finally, the system combines the preset differentiated strategy parameters corresponding to the risk range to which the house belongs, such as the 15% increase in the amount for high-risk ranges and the on-site notarization process, with the initial compensation strategy parameters of the house to form a final compensation strategy that takes into account the objective attributes and historical situation of the house and dynamically optimizes its risk assessment results. For example, a base amount of 300,000 × 1.15 = 345,000, plus the original legal aid mark and the newly added notarization process.

[0099] By executing steps 141 to 143, the embodiment of the present application realizes the objective division of risk thresholds through the fracture algorithm, ensuring that the interval division conforms to the inherent distribution law of the data; generates initial compensation parameters based on multi-source heterogeneous data to ensure the basic rationality of the strategy; integrates risk interval factors to realize dynamic optimization of the compensation strategy, and improves the ability to resolve conflicts in high-risk houses and the accuracy of resource allocation.

[0100] Figure 2 A schematic diagram of the structure of a housing intelligent relocation application system provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system includes: The collection module 21 is used to collect multi-source heterogeneous data in the target resettlement area, where the multi-source heterogeneous data includes house attribute data, right holder characteristic data, compensation negotiation data and environment-related data.

[0101] The extraction module 22 is used to extract key features whose correlation values ​​with the target resettlement project risk are greater than a preset correlation value from multi-source heterogeneous data to form a resettlement project risk feature vector for each house in the target resettlement area.

[0102] The input module 23 is used to input the risk feature vector of the resettlement project into a pre-trained random forest model to generate a risk probability value for each house in the target resettlement area and form a visual risk heat map.

[0103] The generation module 24 is used to generate corresponding differentiated compensation strategies for houses with different risk probability values ​​based on the visual risk heat map.

[0104] Figure 2 The housing intelligent relocation application system can execute Figure 1 The implementation principle and technical effects of the smart housing relocation application method described in the embodiment are not described in detail here. The specific manner in which each module and unit performs operations in the smart housing relocation application system in the above embodiment has been described in detail in the embodiment of the method and will not be elaborated on here.

[0105] In one possible design, Figure 2 A housing smart relocation application system of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32 .

[0106] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0107] The processing component 32 is used to collect multi-source heterogeneous data in the target resettlement area, where the multi-source heterogeneous data includes house attribute data, property owner characteristic data, compensation negotiation data, and environmental association data. Key features whose correlation values ​​with the target resettlement project risk are greater than preset correlation values ​​are extracted from the multi-source heterogeneous data to form a resettlement project risk feature vector for each house in the target resettlement area. The resettlement project risk feature vector is input into a pre-trained random forest model to generate a risk probability value for each house in the target resettlement area, forming a visual risk heat map. Based on the visual risk heat map, corresponding differentiated compensation strategies are generated for houses with different risk probability values.

[0108] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0109] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as random access memory (RAM), static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0110] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0111] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0112] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0113] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0114] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 A smart housing relocation application method according to the embodiment shown.

[0115] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0116] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0117] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A smart housing relocation application method, characterized in that: include: Collect multi-source heterogeneous data in the target relocation area, including housing attribute data, right holder characteristic data, compensation negotiation data, and environmental related data; Extracting key features whose correlation values ​​with the target resettlement project risk are greater than a preset correlation value from the multi-source heterogeneous data to form a resettlement project risk feature vector for each house in the target resettlement area; Input the risk feature vector of the resettlement project into a pre-trained random forest model to generate a risk probability value for each house in the target resettlement area and form a visual risk heat map; Based on the visualized risk heat map, corresponding differentiated compensation strategies are generated for houses with different risk probability values.

2. The method according to claim 1, characterized in that The risk feature vector of the resettlement project is input into a pre-trained random forest model to generate a risk probability value for each house in the target resettlement area, forming a visual risk heat map, including: Inputting the risk feature vector of the resettlement project for each house in the target resettlement area into multiple decision trees in a pre-trained random forest model to obtain the risk judgment results output by each decision tree; Aggregate all risk judgment results of a single house to generate a risk probability value for the single house to form a risk probability value set; Associating each risk probability value in the risk probability value set with the coordinate position of the corresponding house; Determine the color depth identifier corresponding to the risk probability value of the house based on a preset mapping relationship between the risk probability value and the color depth identifier; The coordinate positions of all houses and the corresponding color depth labels are superimposed on the regional map to form a visual risk heat map.

3. The method according to claim 2, characterized in that The risk assessment result is at least one of a risk label, a risk score, and a risk level; Aggregating all risk assessment results for a single house to generate a risk probability value for the single house includes: Gather all risk assessment results of a single house to generate a risk assessment result set for the house; Counting the number of risk judgment results that meet a preset high-risk condition in the risk judgment result set of the house; the predetermined high-risk condition comprising at least one of the following: the risk label type is high risk, the risk score exceeds a preset score threshold, and the risk level is ranked in the top N levels; N is an integer greater than or equal to 1; The ratio of the number of risk judgment results to the number of decision trees in the random forest model is used as the risk probability value of a single house.

4. The method according to claim 2, characterized in that Different decision trees output the risk judgment results corresponding to the risk feature vector of the same resettlement project; The risk feature vector of the resettlement project is input into multiple decision trees in a pre-trained random forest model to obtain the risk judgment results output by each decision tree, including: Inputting the risk feature vector of the resettlement project into all decision trees of the random forest model simultaneously; In each decision tree, the feature values ​​are recursively matched according to the splitting rules of the tree structure, and finally the leaf nodes are reached, and the risk judgment results are extracted from the leaf node attributes.

5. The method according to claim 2, characterized in that Different decision trees output risk judgment results corresponding to different key features; The risk feature vector of the resettlement project is input into multiple decision trees in a pre-trained random forest model to obtain the risk judgment results output by each decision tree, including: According to the preset decision tree feature mapping table, determine the exclusive feature set corresponding to each decision tree; Extracting each exclusive feature set from the risk feature vector of the resettlement project; Input each unique feature set into the corresponding decision tree; In each decision tree, node splitting decisions are performed based on the input feature subset until a leaf node is reached, and the risk judgment results are extracted from the leaf node attributes.

6. The method according to claim 2, characterized in that The determining of the color depth identifier corresponding to the risk probability value of the house based on a preset mapping relationship between the risk probability value and the color depth identifier includes: The HSV value is determined by using the mapping relationship between the range of the risk probability value and the HSV color space established based on the nonlinear mapping function; Based on the color conversion relationship, the HSV value is converted into the RGB color space to obtain the RGB value; Gamma correction is performed on the RGB value, and the corrected RGB value is encoded into a hexadecimal color code as the color depth identifier.

7. The method according to claim 1, characterized in that The method generates corresponding differentiated compensation strategies for houses with different risk probability values ​​based on the visualized risk heat map, including: Based on the distribution of housing risk probability values ​​in the visualized risk heat map, a preset fracture algorithm is applied to calculate a demarcation threshold, and the risk interval to which each housing belongs is determined according to the demarcation threshold; For each house, generating initial compensation strategy parameters according to the data corresponding to the house in the multi-source heterogeneous data; The differentiated compensation strategy parameters corresponding to the risk interval and the initial compensation strategy parameters are integrated to obtain a differentiated compensation strategy including the integrated compensation strategy parameters.

8. A house intelligent relocation application system, characterized by: include: A collection module is used to collect multi-source heterogeneous data in the target relocation area, wherein the multi-source heterogeneous data includes housing attribute data, right holder characteristic data, compensation negotiation data, and environmental related data; An extraction module is used to extract key features whose correlation value with the risk of the target resettlement project is greater than a preset correlation value from the multi-source heterogeneous data, so as to form a resettlement project risk feature vector for each house in the target resettlement area; An input module is used to input the risk feature vector of the resettlement project into a pre-trained random forest model to generate a risk probability value for each house in the target resettlement area and form a visual risk heat map; A generation module is used to generate corresponding differentiated compensation strategies for houses with different risk probability values ​​based on the visual risk heat map.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a smart house relocation application method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, an intelligent housing relocation application method as described in any one of claims 1 to 7 is implemented.

Images