Real estate-based online value evaluation system

Abstract

The invention provides an online value evaluation system based on real estate, and belongs to the technical field of online evaluation, and the system comprises an online data obtaining module which is used for obtaining a basic feature vector and multi-modal evaluation data of a target real estate online; the feature extraction and mining module is used for carrying out space mapping on the explicit features and the defect features and carrying out potential association mining on the implicit features, and obtaining high-association implicit description in combination with the basic feature vector; the evaluation processing module is used for identifying real evaluation, false evaluation and fuzzy evaluation, and correcting the fuzzy evaluation after preliminary analysis according to a spatial mapping result and high-correlation implicit description in combination with social platform subject discussion data; and the value evaluation module is used for obtaining a comprehensive value according to the basic value reference line, the real evaluation and the correction evaluation, and outputting a value evaluation report. The problems that real estate evaluation is difficult to distinguish virtuality and reality and is fuzzy and difficult to solve are effectively solved, and the evaluation result can be comprehensive and accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of online evaluation, and in particular to an online value evaluation system based on real estate. Background Art

[0002] With the development of the real estate market, the demand for property valuation is growing. Currently, existing property valuation methods focus on basic property attributes, such as location, floor area, unit type, and construction year, calculating a property's base value based on these fundamental attributes. However, in actual transactions and market perception, property value is also influenced by additional factors such as user reviews, including evaluations of the property's surrounding environment, property services, and transportation accessibility. In the existing technology, although some online real estate appraisal platforms collect user reviews, they have the following defects: First, they are unable to effectively identify and filter out false reviews, which causes false reviews to affect the accuracy of the appraisal results; second, it is difficult to effectively handle fuzzy reviews (such as vague reviews such as "the surrounding environment is good" and "the transportation is convenient") and cannot convert them into quantifiable value factors; third, real reviews, false reviews, and fuzzy reviews are not systematically integrated and incorporated into the real estate value assessment system as added value, making the assessment results less comprehensive and accurate. Therefore, the present invention proposes an online value assessment system based on real estate. Summary of the Invention

[0003] The present invention provides an online value assessment system based on real estate to solve the above-mentioned technical problems.

[0004] The present invention proposes an online value assessment system based on real estate, comprising: An online data acquisition module, which is used to acquire basic feature vectors and multimodal evaluation data of target properties online through a distributed crawler cluster and an open API interface. The multimodal evaluation data includes text evaluations, image evaluations, and social platform topic discussion data. A feature extraction and mining module is used to extract explicit features and implicit features from the text evaluation, and at the same time, extract defect features from the image evaluation, spatially map the explicit features with the defect features, and mine potential associations of the implicit features, and obtain a highly correlated implicit description by combining the basic feature vectors; An evaluation processing module, configured to identify true, false, and fuzzy evaluations of the text and image evaluations based on a pre-built evaluation classification model, eliminate false evaluations, perform preliminary analysis on the fuzzy evaluations, and modify the fuzzy evaluations after preliminary analysis based on spatial mapping results, highly correlated implicit descriptions, and social platform topic discussion data; The value assessment module is used to obtain gridded transaction data in the same area as the target property to obtain a basic value baseline. At the same time, based on the deep belief network, it obtains weight evaluation factors of different value types, and obtains a comprehensive value based on the basic value baseline, the true evaluation and the revised evaluation, and outputs a value assessment report.

[0005] Preferably, the basic feature vector includes: location entropy value, spatial topological characteristics, physical attenuation coefficient and supporting efficiency index.

[0006] Preferably, the evaluation processing module includes: A cross-verification unit is used to cross-verify the identified fuzzy reviews using the user identity information stored in the blockchain. If the reviewer has not completed real-name authentication or the authentication information has no geographical association with the property, the fuzzy review is assigned a false and irrelevant label; otherwise, the fuzzy review is assigned a possibly false label; A possible assignment unit is used to quantify the uncertainty of the fuzzy evaluation based on the correlation between the historical fuzzy evaluation and the actual transaction price fluctuation, convert it into expectation, entropy and super entropy, and analyze the expectation, entropy and super entropy based on a preset analysis mechanism to assign a possible reference label or a non-reference label; a retaining unit, configured to retain the fuzzy evaluation only when the fuzzy evaluation is assigned both a possible false label and a possible reference label; Otherwise, the fuzzy evaluation is eliminated, wherein the fuzzy evaluation after the preliminary analysis is the retained fuzzy evaluation.

[0007] Preferably, the evaluation processing module further includes: a primary correction unit, configured to perform fuzzy element mapping on the fuzzy evaluation after the preliminary analysis to obtain a quantized ternary fuzzy array, and perform a primary correction on the ternary fuzzy array in combination with the fuzzy membership constraints of the basic eigenvectors, wherein the fuzzy elements include: object elements, feature elements, and quantization elements; A contribution determination unit is configured to extract the comparison relationship between the core value description of the real evaluation and the spatial mapping result and the highly correlated implicit description, and combine it with the social platform topic discussion data to obtain the value factor of the core value description, and determine the first contribution value in combination with the exponential decay weight of the evaluation release time; an element label assigning unit, configured to assign a label to the quantified fuzzy element based on a matching relationship between the core value description and the element evaluation description of the fuzzy element after the preliminary analysis, and based on the corresponding first contribution value, wherein the label is a strong correlation label, a weak correlation label, and an uncorrelated label; A secondary correction unit, configured to perform secondary correction on the ternary fuzzy array according to the label result; The correction evaluation unit is used to obtain a correction evaluation based on the first correction result and the second correction result.

[0008] Preferably, the feature extraction and mining module includes: A function construction unit, configured to construct an expected function of implicit features, determine a depth level for potential association mining based on the expected function, and mine from a preset knowledge graph based on the depth level; The associated feature extraction unit is used to extract the associated features of the mining results at each level based on the basic feature vector, and perform description conversion on the associated features at each level to obtain a highly associated implicit description.

[0009] Preferably, the feature extraction and mining module further includes: A horizontal mapping unit is used to perform horizontal mapping between the explicit features and the defect features to determine the horizontal matching degree. ,in, The horizontal offset between the three-dimensional coordinate point of the display feature and the three-dimensional coordinate point of the defect feature; The width of the wall corresponding to the explicit feature; A vertical mapping unit is used to map the explicit features and defect features vertically to determine the vertical matching degree. ,in, It is the vertical offset between the three-dimensional coordinate point of the explicit feature and the three-dimensional coordinate point of the defect feature; The height of the wall corresponding to the explicit feature; A diagonal mapping unit is used to diagonally map the explicit features to the defect features to determine the diagonal matching degree. , where d is the absolute distance difference between the three-dimensional coordinate point of the explicit feature and the three-dimensional coordinate point of the defect feature to the diagonal line; is the standard deviation of wall dimensions; is the diagonal length of the space where the explicit feature is located; A timing mapping unit is used to perform timing mapping between the explicit features and the defect features to determine the timing matching degree. ,in, is the maximum allowed time difference, which is 365 days; The time when the corresponding evaluation was published; For the present time; The time when the defect actually begins to appear; is the vertical height of the wall based on the current time; is the vertical height of the wall based on the time the review was published; The difference between the current time and the start time of recording the event corresponding to the explicit feature; is the error weight; is the attenuation coefficient; The horizontal direction matching degree Mx, the vertical direction matching degree My, the diagonal direction matching degree Md and the temporal direction matching degree Ms are taken as the spatial mapping results.

[0010] Preferably, the contribution determination unit includes: The positioning subunit is used to locate the attribute words and feature words in the real evaluation and construct a structured description ,in, A semantic vector representing the true evaluation; An association subunit for determining the structured description Attribute association with the basic feature vector , and retain the descriptions whose attribute correlation is greater than the preset correlation to form the core value description set ,in, represents the jth core value description; m represents the number of core value descriptions; The graph construction sub-unit is used to construct a three-layer heterogeneous graph G = (V, E) based on the correlation between the core value description, the spatial mapping results, and the highly correlated implicit description. The first layer is the core value description nodes, the second layer is the spatial mapping result nodes, and the third layer is the highly correlated implicit description nodes. V represents the nodes involved in the three layers, and E represents the edges corresponding to the nodes. A three-domain extraction subunit is used to extract three-domain features associated with the core value description from the social platform topic discussion data, wherein the three-domain features include: emotion domain, communication domain, and time and space domain; The factor determination subunit is used to obtain the value factor of the core value description based on the three-layer heterogeneous graph and the three-domain features.

[0011] Preferably, the value assessment module includes: A benchmark correction unit is used to correct the basic value baseline based on the spatiotemporal attenuation coefficient: A factor analysis unit, configured to assign initial weights to the weight evaluation factors output by the deep belief network according to feature categories, and at the same time, capture temporal correlations between the weight evaluation factors; The comprehensive acquisition unit is used to obtain the comprehensive value based on the revised baseline, the initial weight and time series association, the quantitative value of the value contribution of the real evaluation, and the quantitative value of the value contribution of the revised evaluation.

[0012] Compared with the prior art, the present invention has the following advantages: By integrating multimodal data (text, images, topic discussions) with basic features, combining spatial mapping to verify the authenticity of the evaluation, using deep learning models to extract key influencing factors, dynamically correcting fuzzy evaluations, and ultimately generating a comprehensive value assessment report, we can effectively solve the problem of real estate evaluations being difficult to distinguish between reality and fiction, and making the evaluation results comprehensive and accurate.

[0013] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0014] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 This is a structural diagram of an online value assessment system based on real estate in an embodiment of the present invention; Figure 2 4 is a structural diagram of a heterogeneous graph in an embodiment of the present invention. DETAILED DESCRIPTION

[0016] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0017] The present invention proposes an online value assessment system based on real estate, such as Figure 1 Shown, including: An online data acquisition module, which is used to acquire basic feature vectors and multimodal evaluation data of target properties online through a distributed crawler cluster and an open API interface. The multimodal evaluation data includes text evaluations, image evaluations, and social platform topic discussion data. A feature extraction and mining module is used to extract explicit features and implicit features from the text evaluation, and at the same time, extract defect features from the image evaluation, spatially map the explicit features with the defect features, and mine potential associations of the implicit features, and obtain a highly correlated implicit description by combining the basic feature vectors; An evaluation processing module, configured to identify true, false, and fuzzy evaluations of the text and image evaluations based on a pre-built evaluation classification model, eliminate false evaluations, perform preliminary analysis on the fuzzy evaluations, and modify the fuzzy evaluations after preliminary analysis based on spatial mapping results, highly correlated implicit descriptions, and social platform topic discussion data; The value assessment module is used to obtain gridded transaction data in the same area as the target property to obtain a basic value baseline. At the same time, based on the deep belief network, it obtains weight evaluation factors of different value types, and obtains a comprehensive value based on the basic value baseline, the true evaluation and the revised evaluation, and outputs a value assessment report.

[0018] Preferably, the basic feature vector includes: location entropy value, spatial topological characteristics, physical attenuation coefficient and supporting efficiency index.

[0019] In this embodiment, the distributed crawler cluster is a distributed data collection system composed of multiple servers. It improves efficiency through parallel crawling and avoids the restriction of a single IP. For example, 10 servers are deployed, each running 5 crawler processes, crawling the housing information of platforms such as Lianjia and Beike, and crawling the discussions about the target community on Weibo and Zhihu. Specifically: it is built based on the Scrapy-Redis framework, uses an IP proxy pool (such as Abu Cloud Proxy) to dynamically switch IPs, and sets a crawling frequency threshold (such as ≤30 times per minute for a single domain name) to avoid triggering anti-crawl.

[0020] In this embodiment, the open API interface is a standardized data interface provided by a third-party platform, which is used to legally obtain authorized data. For example, the AutoNavi Map API is called to obtain supporting facilities (schools, hospitals, subway entrances) within 3 kilometers of the target property; the open API of the housing and construction department is called to obtain the construction year and building area of ​​the property registration. Specifically, it is authenticated by the API key, requested according to the interface document specification (such as RESTful format), and a timeout retry mechanism is set (up to 3 times).

[0021] In this embodiment, the target property is a specific house to be evaluated, which has a unique identifier (such as property certificate number, community name + unit number + room number), for example, Room 501, Unit 2, Building 3, XX Community, Chaoyang District, Beijing.

[0022] In this embodiment, the basic feature vector is a set of quantitative data describing the physical and locational attributes of a property, and is formed into vectors by dimension, for example, [location entropy value 0.85, spatial topological characteristics 0.72, physical attenuation coefficient 0.3, supporting efficiency index 0.9] (each value is the result of normalization processing), specifically: through structured field extraction (such as building area obtained from registered information) and quantitative model calculation (such as location entropy value = number of business districts / sum of squared distances).

[0023] In this embodiment, the multimodal evaluation data is multiple types of user evaluation data, including text, images, topic discussions, and other forms.

[0024] Text evaluation: text description posted by users, such as "The community property management is very poor and no one has cleaned up the garbage for a week."

[0025] Image evaluation: User-uploaded photos / videos of houses, such as close-ups of cracks on the living room wall and videos of community greening.

[0026] Social platforms: Discussions on the target community on social platforms (Weibo, Tik Tok), such as the popularity and sentiment of the topic “XX community homeowners defending their rights.”

[0027] In this embodiment, explicit features are property attributes explicitly described in the text evaluation, which can be directly quantified or located, such as "building area 120 m2", "master bedroom facing south", and "built in 2010". Specifically, numbers, directional words, etc. are extracted based on regular expressions, and property attribute words are matched in combination with the knowledge graph (such as "facing south" is associated with the "lighting" feature).

[0028] In this embodiment, implicit features are potential information that is not directly expressed in the text evaluation but can be obtained through semantic reasoning. For example, "living comfortably" implies good lighting and good sound insulation; "convenient work" implies convenient transportation. Specifically, the BERT model is used to calculate the semantic similarity between the text and preset feature words (such as lighting and transportation). If the similarity is ≥0.7, the corresponding feature is associated.

[0029] In this embodiment, the defect features are physical defects of the house identified from image evaluation, such as structural damage and aging facilities. For example, the length of the wall crack in the image is 1.2m, the area of ​​the window glass damage is 0.05㎡, and there are traces of water seepage in the bathroom. Specifically, the defect area in the image is detected using the YOLOv8 model, and the defect contour is segmented in combination with Mask R-CNN. The defect size (length, area) is calculated and normalized (such as crack length / wall height).

[0030] In this embodiment, spatial mapping is to match the spatial position of the text description with the defect position in the image to verify the authenticity of the description. For example, the text describes a crack on the east wall of the master bedroom 1.5 meters above the ground, which is matched with the crack feature at that coordinate in the image (based on the three-dimensional model of the house). The matching degree is calculated to be 85%. Specifically, a three-dimensional grid coordinate system for the house is constructed, the text position is parsed as (room ID, wall orientation, height), and the image defect is generated into three-dimensional coordinates using SfM (structure from motion) technology. The Euclidean distance error is calculated to be ≤0.3m to determine the match.

[0031] In this embodiment, potential association mining is to mine the potential association between implicit features and basic features through knowledge graph reasoning.

[0032] In this embodiment, the highly correlated implicit description is an implicit feature description that is highly correlated with the basic feature vector and has practical reference value. For example, the implicit description "convenience in buying groceries" has a correlation of 0.82 with the basic feature "supporting efficiency index (supermarket density)" and is marked as highly correlated.

[0033] In this embodiment, the evaluation classification model is a machine learning model used to distinguish evaluation types (true, false, fuzzy), integrating multi-dimensional features. For example, based on the Bi-LSTM+GAT model, it inputs text semantics, user behavior, and image defect features, and outputs classification results (true evaluation probability 0.92, false 0.05, fuzzy 0.03). Specifically, it is trained with historical annotated data (100,000 true / false / fuzzy evaluations), and features include sentiment entropy, IP concentration, defect matching, etc., and F1-score ≥ 0.85 is used as the model compliance threshold.

[0034] In this embodiment, a true evaluation is posted by a real owner and the description is consistent with the actual situation of the house. For example, the owner (real-name authentication) posts that "the community property management is on duty 24 hours a day, and last year when there was a water leak at home, they arrived in 10 minutes", and attaches a video of the water leak treatment site, which has been verified to be consistent with the facts.

[0035] In this embodiment, false reviews are misleading reviews posted by non-real users, such as agents posting positive reviews or malicious negative reviews of competitors. For example, the same IP address posts five identical excellent reviews of a community within one hour, the user has not been authenticated with real name, and the distance between the IP address and the community is greater than 50 km.

[0036] In this embodiment, the fuzzy evaluation is an evaluation that is vague and difficult to quantify directly, such as OK, so-so, or okay. For example, the community environment is OK, the house is not too old, and the price is about the same, without clearly stating the specific characteristics.

[0037] In this embodiment, the preliminary analysis is a pre-processing of the authenticity verification and reference value screening of the fuzzy evaluation. For example, the geographical correlation of the fuzzy evaluation based on the blockchain verification that the user is the owner is 0.8, but the correlation with the historical transaction price fluctuation is only 0.3. It is determined to be retained but with a low weight. Specifically: combining blockchain identity authentication (whether real-name authentication is performed, the distance between the IP and the property) and gray correlation analysis (correlation with transaction data), the fuzzy evaluation that simultaneously meets the identity credibility + correlation ≥ 0.5 is screened.

[0038] In this embodiment, the revised evaluation is the result of quantitative correction of the fuzzy evaluation retained in the preliminary analysis, combined with the spatial mapping results, highly correlated implicit descriptions and topic discussion data. For example, the fuzzy evaluation that the community is a bit noisy is combined with the spatial mapping (near the main road), the implicit description of not sleeping well at night and the topic discussion data (the top 10% of surrounding noise complaints), and is corrected to a moderate noise impact with an average decibel level of 58dB at night. Specifically, the cloud model is used to convert the fuzzy description into a quantitative score (e.g., a bit noisy corresponds to 55-65dB), and the score range (58-68dB) is adjusted based on the emotional tendency of the topic discussion (negative accounts for 60%).

[0039] In this embodiment, the same-region gridded transaction data is obtained by dividing the area where the target property is located into grids (e.g., 500m×500m), and then counting the property transaction records (price, area, apartment type, etc.) within each grid over the past six months. For example, the target property is located in Grid A, Chaoyang District, Beijing. There were 20 transactions in this grid over the past six months, with an average price of 65,000 yuan / m2. Transactions in areas of 100-120 m2 accounted for 70%. Specifically, the grids are divided using GeoPandas, transaction data is extracted from Lianjia and the Housing and Urban-Rural Development Bureau databases, aggregated by grid ID, and outliers (e.g., prices deviating from the mean by ±3σ) are removed.

[0040] In this embodiment, the basic value baseline is a benchmark valuation of real estate based on the gridded transaction data of the same area and the basic feature vector. For example, the basic features of the target real estate (120 m2, built in 2010, supporting efficiency index 0.8) are used to refer to the transaction data of the same grid and calculate the basic value baseline of RMB 7.8 million. Specifically, the XGBoost model is used to train the relationship between transaction price and basic features, the target real estate features are input and the baseline is output, and the model .

[0041] In this embodiment, a deep belief network (DBN) is a deep neural network used to extract high-weight value influencing factors from evaluation data. For example, by analyzing 100,000 evaluation data through DBN, it is found that "property level", "transportation convenience" and "school district quality" are the top three evaluation factors with the highest weights. Specifically, a three-layer DBN is constructed (128-dimensional input layer, 64-dimensional hidden layer, and 10-dimensional output layer), which is trained with evaluation data (quantized true evaluation and revised evaluation) to output the weight of each factor (the sum of the weights is 1).

[0042] In this embodiment, the weighted evaluation factors are evaluation dimensions that have a significant impact on property value. The higher the weight, the greater the impact of the dimension on the valuation. For example, the weighted evaluation factors are "property level (0.3), transportation convenience (0.25), school district quality (0.2), environmental noise (0.15), and green coverage rate (0.1)".

[0043] In this embodiment, the comprehensive value is the final valuation that integrates the basic value baseline and the evaluation factors (real evaluation and revised evaluation), reflecting the comprehensive market value of the property.

[0044] In this embodiment, the value assessment report is a structured report that includes basic real estate information, valuation results, influencing factor analysis, and data sources. For example, the report includes the target real estate address, basic characteristics, a comprehensive value of 7.95 million yuan (±5%), weighted evaluation factors (property, transportation), and data sources (transaction data, 120 real reviews, and 30 revised reviews).

[0045] The location entropy value is used to quantify the location advantage of the area where the property is located. It reflects the economic vitality and living convenience of the area by calculating the spatial agglomeration and radiation intensity of surrounding core resources (such as business districts, transportation hubs, and educational and medical facilities). The higher the value, the better the location (after normalization, ∈ [0,1]). For example, the location entropy value of a school district house in Haidian District, Beijing is calculated as follows: within 3 kilometers, there are 5 large business districts (weight 0.3), 3 subway stations (weight 0.2), 2 key primary schools (weight 0.3), and 1 tertiary hospital (weight 0.2). The entropy value formula calculates 0.85 (high location advantage). On the other hand, there is only one small supermarket and no subway around a property in the suburbs, and the location entropy value is 0.23 (low location advantage). Specifically, a distributed crawler is used to collect resource POI data within a radius of 1 / 3 / 5 kilometers (obtained through the AutoNavi / Baidu Map API). Assign weights by resource type (education > transportation > commerce > medical care, with the sum of weights equal to 1); Calculated using the location entropy formula: Location entropy value , where resource density = number of resources / regional area (square kilometers), and normalized (divided by the maximum entropy value of the same city), where n1 represents the number of regions.

[0046] Spatial topology describes the rationality and functionality of a property's internal spatial structure. It is quantified through indicators such as the connectivity, squareness, and completeness of functional zoning of the apartment layout, reflecting the actual usage efficiency of the house. Higher values ​​indicate a better spatial structure (normalized to ∈ [0, 1]). For example, the spatial topology characteristics of a three-bedroom apartment were calculated: squareness (90%, wall turning angles close to 90°), functional area connectivity (living room and dining room directly connected, bedroom and bathroom traffic flow distance <5m), and space utilization (common area accounts for 15%). The overall score was 0.82 (high-quality apartment). On the other hand, the spatial topology characteristics of a "knife-handle-shaped" one-bedroom apartment (squareness 60%, chaotic traffic flow) was 0.35 (poor apartment). Specifically, the apartment structure data was extracted from the property's CAD drawings or panoramic images (using OpenCV edge detection to identify wall, door, and window locations). Calculation of core indicators: Squareness = area of ​​the smallest circumscribed rectangle of the apartment / actual interior area; Connectivity = number of direct connections between functional areas (living room, bedroom, kitchen and bathroom) / total number of possible connections; Utilization rate = usable area within the unit / building area; The analytic hierarchy process (AHP) was used to fuse the indicators, and the weighted sum was normalized (weights: squareness 0.4, connectivity 0.3, utilization 0.3).

[0047] The physical attenuation coefficient measures the degree of physical loss of a property due to its construction age, aging of building materials, and maintenance status, reflecting the durability and potential repair cost of the house. Higher values ​​indicate more severe loss (normalized to ∈ [0, 1], where 0 is brand new and 1 is severely aged). For example, a brick-concrete structure house built in 2000 has a construction age of 23 years (weight 0.5), 30% of the exterior paint peeling area (weight 0.2), and 3 records of water leaks due to rusted water pipes (weight 0.3). The physical attenuation coefficient is calculated to be 0.68 (high loss). On the other hand, a steel-concrete structure house built in 2020 (with no maintenance records and a building material aging rate of 5%) has a physical attenuation coefficient of 0.12 (slight loss). Specifically, basic data is collected: construction age (obtained from the housing and construction department API), building material type (brick-concrete / steel-concrete / frame), maintenance records (property system crawler), and defect features in image evaluation (such as cracks and leaks, identified by YOLOv8); Calculation of core indicators: Aging rate = years of construction / building design life (50 years for steel-concrete structures); Defect ratio = defect area in the image / total area of ​​the house; Maintenance frequency = number of maintenance records in the past five years / 5 (annual average); Weighted sum: physical attenuation coefficient = aging rate × 0.5 + defect ratio × 0.3 + repair frequency × 0.2, normalized (divided by the maximum attenuation value of the same type of house).

[0048] The supporting facilities efficiency index quantifies the comprehensive efficiency of the "accessibility + quality" of the supporting facilities around the property, covering core supporting facilities such as education, medical care, commerce, and transportation, reflecting the convenience of residents' daily lives. The higher the value, the better the supporting facilities (after normalization, ∈ [0,1]). For example, the supporting facilities efficiency index calculation of a community is: Education: 10 minutes’ walk to the city’s key primary school (quality score 0.9, accessibility 0.9, weight 0.3); Healthcare: 5 minutes by bike to a community hospital, 15 minutes by car to a tertiary hospital (quality score 0.7, accessibility 0.8, weight 0.2); Commercial: 5-minute walk to supermarket, 10-minute walk to shopping mall (quality score 0.8, accessibility 0.9, weight 0.2); Transportation: 3 minutes’ walk to the bus stop, 10 minutes to the subway station (quality score 0.85, accessibility 0.95, weight 0.3); Comprehensive calculation: (0.9×0.9×0.3)+(0.7×0.8×0.2)+(0.8×0.9×0.2)+(0.85×0.95×0.3)=0.83 (high-performance package).

[0049] Specifically: data is collected based on a three-level indicator system of type (education / healthcare, etc.) - quality (star rating) - accessibility (time / distance): Quality rating: education is based on enrollment rate, health care is based on tertiary qualifications, and commerce is based on the richness of business formats (quantified from 1 to 5 stars); Accessibility: walking / cycling / driving time (obtained through the route planning function of the map API), normalized to [0,1] (the shorter the time, the higher the value); for each type of supporting facilities, a "quality × accessibility" score is calculated, the score is summed by weight (education 0.3, transportation 0.3, commerce 0.2, health care 0.2), and then normalized (divided by the city's highest supporting efficiency value).

[0050] The beneficial effects of the above technical solution are: by fusing multimodal data (text, images, topic discussions) with basic features, combining spatial mapping to verify the authenticity of the evaluation, using deep learning models to extract key influencing factors, dynamically correcting fuzzy evaluations, and finally generating a comprehensive value assessment report, it effectively solves the problem of real estate evaluation being difficult to distinguish between reality and fiction and fuzzy and difficult to resolve, so that the evaluation results can be comprehensive and accurate.

[0051] The present invention proposes an online value assessment system based on real estate, wherein the assessment processing module includes: A cross-verification unit is used to cross-verify the identified fuzzy reviews using the user identity information stored in the blockchain. If the reviewer has not completed real-name authentication or the authentication information has no geographical association with the property, the fuzzy review is assigned a false and irrelevant label; otherwise, the fuzzy review is assigned a possibly false label; A possible assignment unit is used to quantify the uncertainty of the fuzzy evaluation based on the correlation between the historical fuzzy evaluation and the actual transaction price fluctuation, convert it into expectation, entropy and super entropy, and analyze the expectation, entropy and super entropy based on a preset analysis mechanism to assign a possible reference label or a non-reference label; a retaining unit, configured to retain the fuzzy evaluation only when the fuzzy evaluation is assigned both a possible false label and a possible reference label; Otherwise, the fuzzy evaluation is eliminated, wherein the fuzzy evaluation after the preliminary analysis is the retained fuzzy evaluation.

[0052] In this embodiment, a fuzzy evaluation is an evaluation posted by a user that is semantically ambiguous and lacks clear quantitative standards. It is impossible to directly determine the direction or degree of impact on the property value. For example, descriptions without specific indicators such as "the community environment is okay", "the house is not too old", and "the price is about the same" are possible.

[0053] The user identity information stored on the blockchain is the user's real-name authentication data stored on the blockchain (tamper-proof and traceable), including name, ID number, geographic location, etc., which is used to verify the authenticity of the review publisher. For example, after the user completes the real-name authentication on the review platform, the identity information is stored in the alliance chain (such as the real estate evaluation dedicated alliance chain) through hash encryption. Each piece of information is associated with a unique blockchain address and can be verified through smart contract query.

[0054] Cross-validation verifies the correlation between user identity and evaluation through multi-source data comparison to ensure that there is a reasonable correlation between the review publisher and the property. For example, if a user comments that "the sound insulation of XX community is average", the cross-validation shows that the permanent address of the user's identity on the chain is within 3 kilometers of the community, and there is a match with the express delivery address in the past 6 months, and the correlation is determined to be valid.

[0055] Real-name authentication is an identity verification completed by users through official channels (such as the public security system) to prove that the account belongs to a real natural person. For example, a user uploads a photo of his or her ID card on the evaluation platform and completes facial recognition. The information is verified by the public security interface and marked as "real-name authenticated."

[0056] Geographic association means that there is a reasonable spatial association (usually within a close range) between the user's identity information (such as permanent address and IP location) and the physical location of the target property. For example, if the target property is located in XX Community, Haidian District, Beijing, and the user's real-name authentication address is XX Street, Haidian District (within 3 kilometers), or the IP location of the review is Haidian District, it is determined to have a geographical association. If the user's address is Guangzhou City, Guangdong Province, there is no geographical association. Specifically, the straight-line distance between the user's address and the property is calculated through the AutoNavi Map API, a threshold is set (such as within 5 kilometers for a valid association), and the IP location (accurate to the city / district) is used for auxiliary verification.

[0057] The "false and irrelevant" label is used to mark ambiguous reviews that indicate "the identity is untrustworthy and has no connection with the property," indicating that the review is very likely false information and has nothing to do with the actual property. For example, a review of "XX community is very poor" posted by an unauthenticated user whose IP address is 100 kilometers away from the property is assigned the "false and irrelevant" label.

[0058] The "possibly false" label is used to mark a fuzzy evaluation that "the identity is credible but there is a potential risk of falsehood." This indicates that the identity of the publisher of the evaluation is real, but the authenticity of the content needs further verification. For example, a user who has been authenticated by real name posted "The community property is not good," but there is no specific description of the incident, and the user has no consumption record in the community in the past year, so the label "possibly false" is given.

[0059] Historical fuzzy evaluations are a collection of fuzzy evaluations of the same type and location as the target property over a period of time (e.g., one year). These evaluations are used to analyze the correlation between fuzzy evaluations and actual transactions. For example, 1,000 fuzzy evaluations from the past year in Haidian District, Beijing, such as "housing price is OK" and "environment is average," are associated with corresponding second-hand housing transaction data.

[0060] The actual transaction price fluctuation refers to the change in the actual transaction price of second-hand houses in the area where the target property is located over time, reflecting the market's true feedback on the value of the property. For example, the average transaction price of a certain community from January to June 2023 rose from 5 million yuan / unit to 5.3 million yuan / unit, with a fluctuation of 6%; from July to December, it fell to 5.1 million yuan / unit, with a fluctuation of -3.8%.

[0061] The correlation is the degree of association between the content of historical fuzzy evaluations and the fluctuations in transaction prices during the same period. The higher the correlation, the greater the reference value of the fuzzy evaluation for prices. For example, statistics show that when the number of fuzzy evaluations of the "good environment" type increases, housing prices in the same area increase by an average of 2%, with a correlation coefficient of 0.6 (medium correlation); when the number of evaluations of the "poor property" type increases, housing prices fall by an average of 1.5%, with a correlation coefficient of 0.5. Specifically, the Pearson correlation coefficient is used to calculate the correlation between the frequency of fuzzy evaluations (or emotional tendencies) and transaction price fluctuations. A coefficient absolute value ≥0.4 is considered to be a valid correlation.

[0062] Uncertainty quantification is to convert the semantic uncertainty of fuzzy evaluation into computable digital features, and use "expectation, entropy, and super entropy" to describe its degree of fuzziness. For example, the fuzzy evaluation "the lighting is okay" is quantified, with expectation = 0.6 (the average corresponds to "medium lighting"), entropy = 0.2 (degree of uncertainty), and super entropy = 0.05 (the fluctuation range of entropy), indicating that the fuzziness of the evaluation is relatively low. Among them, expectation reflects the central trend of the evaluation, entropy reflects the degree of fuzziness, and super entropy reflects the stability of entropy.

[0063] Expectation is the average quantitative value of the fuzzy evaluation, reflecting the central trend of the evaluation (such as the average score corresponding to "okay"). For example, 100 fuzzy evaluations of "the community facilities are okay" correspond to an average of 0.6 points (out of 1 point) after quantification, so expectation = 0.6.

[0064] Entropy describes the degree of uncertainty in fuzzy evaluations. Higher values ​​indicate more ambiguous evaluation semantics and greater divergence. For example, the entropy of "the community environment is OK" is 0.1 (small divergence in evaluations), and the entropy of "the community environment is average" is 0.3 (some people prefer "average", while others think it is bad, with large divergence).

[0065] Super entropy describes the degree of entropy fluctuation and reflects the stability of the uncertainty of fuzzy evaluation. The higher the value, the more unstable the fuzziness of the evaluation. For example, the super entropy of "the housing price is about the same" = 0.03 (the entropy fluctuation is small and the fuzziness is stable); the super entropy of "the apartment type is okay" = 0.08 (the entropy values ​​of evaluations released at different times vary greatly and the fuzziness is unstable).

[0066] The preset analysis mechanism is a rule set based on historical data, which is used to judge the reference value of fuzzy evaluation. It screens effective evaluations through the thresholds of expectation, entropy, and super entropy. For example, the rule is set as "expectation∈[0.3,0.7] (avoid extreme values), entropy<0.3 (low fuzziness), super entropy<0.1 (stable fuzziness)". If the rule is met, a possible reference label is assigned, otherwise a non-reference label is assigned.

[0067] The possible reference label is a fuzzy evaluation marked as "low uncertainty and related to transaction price fluctuations", indicating that the evaluation has potential reference significance for real estate valuation. For example, the fuzzy evaluation "transportation is relatively convenient" has an expectation of 0.6, entropy of 0.2, and super entropy of 0.04, which meets the preset rules and is assigned a possible reference label.

[0068] The no-reference label is a fuzzy evaluation marked as "high uncertainty or unrelated to transaction price fluctuations", indicating that the evaluation has no actual reference value for valuation. For example, the fuzzy evaluation "The house is like this" has an entropy of 0.4 (above the threshold) and a correlation of only 0.2 with house price fluctuations, so it is assigned a no-reference label.

[0069] The fuzzy evaluation after preliminary analysis is the fuzzy evaluation retained after identity verification and reference value screening, and must meet both the requirements of possible false labels (trusted identity) and possible reference labels (reference value).

[0070] The beneficial effects of this technical solution include filtering out fuzzy reviews with false associations through blockchain authentication, quantifying the reference value of reviews by combining historical data with the correlation of transaction prices, and ultimately retaining fuzzy reviews that are credible and have practical reference value. This not only eliminates invalid noise but also provides a high-quality data foundation for subsequent accurate correction of fuzzy reviews, effectively improving the reliability of evaluation data in online real estate appraisals.

[0071] The present invention proposes an online value assessment system based on real estate, wherein the assessment processing module further includes: a primary correction unit, configured to perform fuzzy element mapping on the fuzzy evaluation after the preliminary analysis to obtain a quantized ternary fuzzy array, and perform a primary correction on the ternary fuzzy array in combination with the fuzzy membership constraints of the basic eigenvectors, wherein the fuzzy elements include: object elements, feature elements, and quantization elements; a contribution determination unit, configured to extract the comparison relationship between the core value description of the real evaluation and the spatial mapping result and the highly correlated implicit description, and combine it with the social platform topic discussion data to obtain the value factor of the core value description, and determine the first contribution value in combination with the exponential decay weight of the evaluation release time; an element label assigning unit, configured to assign a label to the quantified fuzzy element based on a matching relationship between the core value description and the element evaluation description of the fuzzy element after the preliminary analysis and based on the corresponding first contribution value, wherein the label is a strong correlation label, a weak correlation label, and an uncorrelated label; A secondary correction unit, configured to perform secondary correction on the ternary fuzzy array according to the label result; The correction evaluation unit is used to obtain a correction evaluation based on the first correction result and the second correction result.

[0072] In this embodiment, fuzzy element mapping decomposes the fuzzy evaluation into three core elements: "things-features-quantification" and converts them into a structured fuzzy mathematical expression (mapped into a three-element fuzzy array). The thing element is the object of evaluation (real estate-related entities), such as "wall sound insulation," "property services," and "surrounding transportation." The feature element describes the attributes or dimensions of the thing, such as "sound insulation effect," "response speed," and "congestion level." The quantitative element is a fuzzy quantitative description of the feature (without a clear numerical value), such as "good," "average," "high," and "not too bad."

[0073] The quantized ternary fuzzy array is obtained through fuzzy element mapping, and describes the uncertainty of fuzzy evaluation with three quantitative indicators, namely expectation (Ex), entropy (En), and super entropy (He), which constitute the array [Ex, En, He]. For example, the ternary fuzzy array of the fuzzy evaluation "the wall sound insulation effect is good" is [0.6, 0.15, 0.04], where Ex = 0.6 (the average corresponds to "above average"), En = 0.15 (low fuzziness), and He = 0.04 (high stability).

[0074] In this embodiment, the fuzzy membership constraint of the basic feature vector is a fuzzy membership rule set based on the basic feature vector (location entropy value, spatial topological characteristics, etc.), which is used to constrain the quantitative range of the fuzzy evaluation so that it matches the actual attributes of the property. For example, if the "supporting efficiency index = 0.9" of the target property (dense surrounding supermarkets), then the fuzzy membership constraint related to "shopping convenience" (such as "convenient to buy things") is set: Ex ≥ 0.6 (to avoid underestimation); if the "physical attenuation coefficient = 0.7" (the house is relatively old), then the fuzzy evaluation of "wall quality" (such as "the wall surface is okay") is set with a constraint: Ex ≤ 0.5 (to avoid overestimation). Specifically, for each dimension of the basic feature vector (such as location entropy value, spatial topological characteristics), a fuzzy membership function (such as an S-type function) is trained through historical data to clearly define the reasonable range of Ex, En, and He corresponding to the fuzzy evaluation. If it exceeds the range, it is forcibly corrected (for example, the upper limit of Ex is set to the basic feature quantization value + 0.1).

[0075] In this embodiment, the first correction is the first adjustment of the ternary fuzzy array in combination with the fuzzy membership constraint of the basic eigenvector to ensure that the quantitative results of the fuzzy evaluation are consistent with the actual attributes of the property. For example, the original ternary array of the fuzzy evaluation "the community facilities are OK" is [0.5, 0.2, 0.05], but the "supporting facilities efficiency index = 0.9" of the basic eigenvector. According to the fuzzy membership constraint (Ex≥0.6), the array after the first correction is [0.6, 0.18, 0.05] (improving expectations and reducing ambiguity).

[0076] In this embodiment, the core value description of the real evaluation is a key description extracted from the real evaluation that has a significant impact on the property value and has clear attributes and characteristics. For example, the core value description of the real evaluation "There is Metro Line 3 300 meters away from the community gate, and there is no congestion during rush hours in the morning and evening" is "High transportation convenience (the subway is close and there is little congestion)."

[0077] The spatial mapping result is the spatial matching result between explicit features and defect features (related to the spatial mapping process mentioned above), including horizontal, vertical, diagonal, and temporal matching. For example, the spatial mapping result of the real-world evaluation of "water seepage on the east wall of the master bedroom" and the image defect is "horizontal matching 0.92, vertical matching 0.88" (high matching).

[0078] The contrast relationship is the matching or association relationship between the core value description and the spatial mapping results and highly correlated implicit descriptions, which is used to verify the authenticity of the core value description. For example, the core value description "convenient transportation" is contrasted with the spatial mapping result "200 meters away from the subway entrance (matching degree 0.9)" and the highly correlated implicit description "fast commute to work" to verify the authenticity of "convenient transportation".

[0079] The value factor is a quantitative indicator that measures the degree of influence of the core value description on the value of a property. The higher the value, the greater the impact of the description on the valuation. For example, the core value description of "convenient transportation" is combined with the topic discussion data (positive sentiment accounts for 80%) and the spatial mapping results (matching degree 0.9), and the calculated value factor is 0.35 (ranked first among all factors). Specifically: through the hierarchical analysis method (AHP), the "contrast relationship matching degree (0.4), topic discussion sentiment tendency (0.3), and associated implicit description weight (0.3)" are integrated, and the weighted summation is obtained to obtain the value factor, and the normalized sum is 1.

[0080] The exponential decay weight of the evaluation release time is assigned different weights based on the interval between the evaluation release time and the current evaluation time (the closer the time, the higher the weight), reflecting the timeliness of the evaluation. For example, setting the exponential decay formula (t is the number of days): The evaluation weight within 30 days of release is 0.97, 0.91 within 90 days, 0.83 within 180 days, and 0.6 after 365 days (the weight decays over time).

[0081] The first contribution value is the specific contribution of the core value description to the property value, which is determined by the value factor and the exponential decay weight. For example, the value factor of the core value description "convenient transportation" is 0.35, and the release time is 30 days (weight 0.97), then the first contribution value = 0.35×0.97≈0.34.

[0082] The element evaluation description is a structured description of the fuzzy evaluation based on the fuzzy elements after preliminary analysis (i.e., a combination of "thing elements + characteristic elements + quantitative elements"). For example, the element evaluation description of the fuzzy evaluation "the green coverage rate of the community is relatively high" after preliminary analysis is "things = community greening, characteristics = coverage rate, quantification = relatively high".

[0083] The matching relationship is the semantic similarity or correlation between the core value description and the element evaluation description. For example, the semantic similarity between the core value description "convenient transportation (near the subway)" and the element evaluation description "the surrounding transportation convenience is acceptable" is 0.82 (high match); and the similarity with "the community greening is okay" is 0.15 (low match).

[0084] Strong correlation labels, weak correlation labels, and no correlation labels are correlation strength labels assigned to the quantized fuzzy elements based on the matching relationship and the first contribution value, and are used for subsequent correction.

[0085] Strong correlation labels: Matching degree ≥ 0.7 and first contribution value ≥ 0.3 indicate that the fuzzy element is highly correlated with the core value description. For example, the element evaluation description "Transportation convenience is acceptable" matches the core value description "Convenient transportation" with a matching degree of 0.82 and a first contribution value of 0.34, thus giving it a strong correlation label. Weak correlation labels: Matching degree 0.3 ≤ < 0.7 or first contribution value 0.1 ≤ < 0.3 indicates that there is some correlation, but the strength is limited. For example, the element evaluation description "Average surrounding facilities" matches the core value description "Commercial convenience" with a matching degree of 0.55 and a first contribution value of 0.2, thus giving it a weak correlation label. Uncorrelated labels: Matching degree < 0.3 or first contribution value < 0.1 indicate that there is no substantial correlation. For example, the element evaluation description "Wall sound insulation is acceptable" matches the core value description "Convenient transportation" with a matching degree of 0.1, thus giving it a uncorrelated label.

[0086] The secondary correction is to adjust the ternary fuzzy array after the primary correction according to the label results (strong / weak / no correlation), strengthen the influence of elements with high correlation, and weaken the irrelevant elements. For example, the ternary array [0.6, 0.18, 0.05] after the primary correction is given a strong correlation label. During the secondary correction, the expectation is increased (Ex=0.65) and the entropy is reduced (En=0.15). The final array is [0.65, 0.15, 0.04] (enhanced certainty). If it is an irrelevant label, the expectation is reduced (Ex=0.5) and the entropy is increased (En=0.25).

[0087] The revised evaluation is the final result of the fusion of the primary correction (basic feature constraint) and the secondary correction (label association constraint), which converts the fuzzy evaluation into a quantitative description with higher certainty. For example, the fuzzy evaluation "the greening of the community is okay" is corrected once (Ex=0.6) and twice (the strong correlation label is increased to Ex=0.65), and the revised evaluation is "the green coverage rate of the community is above average (quantitative score 0.65±0.15)".

[0088] The beneficial effect of the above technical solution is: through fuzzy element decomposition-quantitative mapping-double correction, the semantically ambiguous evaluation is converted into a quantitative result with higher certainty: the first correction combines the basic characteristics of the real estate to ensure that the evaluation matches the actual attributes, and the second correction strengthens the relevance of the evaluation by associating core value descriptions and topic discussion data. The final revised evaluation not only retains the effective information of the user's subjective experience, but also reduces ambiguity through mathematical modeling, providing accurate evaluation dimension support for the comprehensive value assessment of the real estate.

[0089] The present invention proposes an online value assessment system based on real estate, wherein the feature extraction and mining module includes: A function construction unit, configured to construct an expected function of implicit features, determine a depth level for potential association mining based on the expected function, and mine from a preset knowledge graph based on the depth level; The associated feature extraction unit is used to extract the associated features of the mining results at each level based on the basic feature vector, and perform description conversion on the associated features at each level to obtain a highly associated implicit description.

[0090] In this embodiment, the expectation function is a mathematical model used to measure the potential association mining benefits. It integrates factors such as association degree, mining cost, and information gain to determine the optimal mining depth (level) so that the value and cost of the mining results are balanced. For example, the expectation function is constructed for the implicit feature "convenience of life". , where R(k) is the correlation degree of the kth level (the higher the value, the greater the benefit), C(k) is the mining cost (the higher the level, the higher the cost), and γ1 and λ1 are weight coefficients (e.g., γ1=0.6, λ1=0.4). The calculated values ​​are E(1)=0.5, E(2)=0.7, and E(3)=0.4, so the optimal depth level is 2. Specifically, the function parameters are trained using historical mining data, R(k) is calculated using the grey correlation degree (the degree of correlation with the basic features), and C(k) is quantified as the calculation time (seconds). The kth level at which the function value is maximized is the optimal depth level (usually 1-3 levels).

[0091] Potential association mining starts from implicit features and uses knowledge graphs to reason layer by layer to mine deep association information related to property attributes, reviews, and supporting facilities. For example, potential association mining is performed on the implicit feature "convenient schooling for children": Level 1 association: "near school"; 2nd level association: "near school → the school is a key primary school"; 3-level association: "Close to key primary schools → stable school district divisions → less pressure to enter higher education".

[0092] The depth level is the number of reasoning steps for potential association mining (the higher the level, the deeper the mining), which is determined by the calculation result of the expectation function to avoid over-mining (high cost, high noise) or under-mining (incomplete information). For example: 1st level: direct association (such as "comfortable" → "good property"); 2nd level: indirect association (such as "comfortable" → "good property" → "24-hour security"); 3rd level: cross-domain association (such as "comfortable" → "good property" → "24-hour security" → "good surrounding security"). For example, set the level threshold (1-3 levels), calculate the benefit value of each level through the expectation function, and select the level with the largest benefit as the mining depth (for example, the benefit of level 2 is the highest, then mine to level 2).

[0093] The preset knowledge graph is a pre-built semantic network that contains real estate entities (such as houses, supporting facilities, and evaluations) and relationships (such as "located in", "influence", and "include"), which is used to support hierarchical reasoning of potential associations. For example, the knowledge graph contains the entities "property", "security", and "public security", and the relationships "property contains security", "security affects public security", and "public security affects living experience", providing support for the reasoning of "comfortable → good property → strict security → good public security". Specifically: Neo4j is used to build the knowledge graph, and the entities cover real estate attributes (location, apartment type, etc.), supporting facilities (schools, hospitals, etc.), and evaluation dimensions (property, environment, etc.); relationships are extracted through manual annotation + machine learning (such as the TransE model) to ensure accurate reasoning logic.

[0094] In this embodiment, the mining result is a set of information associated with implicit features mined by a preset knowledge graph at a depth level. For example, the implicit feature "convenient commuting" is mined at a depth of 2 levels, and the result is "Level 1: close to the subway; Level 2: Metro Line 3 (no congestion during peak hours in the morning and evening)".

[0095] Correlation features are features extracted from the mining results that are significantly correlated with the basic feature vector (the correlation is higher than the preset threshold), ensuring that the mined information matches the actual properties of the property. For example, the correlation between the mining result "near Metro Line 3" and the "location entropy value of 0.85" of the basic feature vector (dense business districts usually have well-developed subways) is 0.75 (higher than the threshold of 0.6), so it is a correlation feature; if the correlation between the mining result "near mountainous area" and "location entropy value of 0.85" is 0.3 (lower than the threshold), then it is a non-correlation feature. The Pearson correlation coefficient is used to calculate the correlation between the mining results and each dimension of the basic feature vector, set a threshold (such as 0.6), and filter features with a coefficient ≥ the threshold as correlation features.

[0096] Description conversion converts associated features from structured data (such as "Metro Line 3, 500 meters away") into natural language descriptions, so that they more intuitively reflect the association with real estate value. For example, the associated feature "Metro Line 3 (500 meters away, congestion index 0.3 during peak hours in the morning and evening)" is converted into the description "Close to Metro Line 3, 5 minutes' walk away, smooth traffic during peak hours in the morning and evening." Specifically, natural language generation is performed based on a template library (such as "close to {transportation facilities}, {distance description}, {traffic status}"), and the template is filled in with the quantitative data of basic features (such as distance and congestion index) to ensure that the description is accurate and easy to understand.

[0097] The beneficial effect of the above technical solution is: through the logic of expectation function optimization of mining depth - knowledge graph hierarchical reasoning - basic feature screening association, fuzzy implicit features are converted into accurate high-correlation descriptions: it avoids the incomplete information of shallow mining and prevents the noise interference of deep mining, ensuring that the mining results are strongly correlated with the actual attributes of the property, and improving the depth and accuracy of online real estate evaluation.

[0098] The present invention proposes an online value assessment system based on real estate, wherein the feature extraction and mining module further includes: A horizontal mapping unit is used to perform horizontal mapping between the explicit features and the defect features to determine the horizontal matching degree. ,in, The horizontal offset between the three-dimensional coordinate point of the display feature and the three-dimensional coordinate point of the defect feature; The width of the wall corresponding to the explicit feature; A vertical mapping unit is used to map the explicit features and defect features vertically to determine the vertical matching degree. ,in, It is the vertical offset between the three-dimensional coordinate point of the explicit feature and the three-dimensional coordinate point of the defect feature; is the wall height corresponding to the explicit feature; A diagonal mapping unit is used to diagonally map the explicit features to the defect features to determine the diagonal matching degree. , where d is the absolute distance difference between the three-dimensional coordinate point of the explicit feature and the three-dimensional coordinate point of the defect feature to the diagonal line; is the standard deviation of wall dimensions; is the diagonal length of the space where the explicit feature is located; A timing mapping unit is used to perform timing mapping between the explicit features and the defect features to determine the timing matching degree. ,in, is the maximum allowed time difference, which is 365 days; The time when the corresponding evaluation was published; For the present time; The time when the defect actually begins to appear; is the vertical height of the wall based on the current time; is the vertical height of the wall based on the time the review was published; The difference between the current time and the start time of recording the event corresponding to the explicit feature; is the error weight; is the attenuation coefficient; The horizontal direction matching degree Mx, the vertical direction matching degree My, the diagonal direction matching degree Md and the temporal direction matching degree Ms are taken as the spatial mapping results.

[0099] In this embodiment, explicit features (text descriptions) and defect features (images) are mapped to a three-dimensional coordinate system of the property (X: horizontal direction, Y: vertical direction, Z: spatial dimension, to assist in diagonal calculation).

[0100] Horizontal offset : The difference between the horizontal coordinate of the explicit feature description (such as "1m left of the east wall") and the actual horizontal coordinate of the defect (take the absolute value, unit: meter).

[0101] Wall width : The horizontal length of the wall where the explicit feature is located (e.g., the width of the east wall is 5m, unit: meter), extracted from the basic feature vector.

[0102] Vertical offset : The difference between the vertical coordinate of the explicit feature description (such as "1.2m from the ground") and the actual vertical coordinate of the defect (take the absolute value, unit: meter).

[0103] Wall height : The vertical height of the wall where the explicit feature is located (such as floor height 3m, unit: meter), extracted from the basic feature vector.

[0104] Absolute distance difference d: The difference (absolute value, unit: meter) between the distances between the 3D points of the explicit feature and the defect feature and the wall diagonal (such as the upper left-lower right diagonal of the east wall).

[0105] Wall size standard deviation : The statistical standard deviation (unit: meter) of the wall dimensions (width / height) of the same apartment type, reflecting the fluctuation of wall dimensions (for example, σ = 0.1m, indicating that the wall width fluctuates by ±0.1m).

[0106] Diagonal length : The diagonal length of the space (such as the room) where the explicit feature is located (unit: meter, calculated by the Pythagorean theorem, for example, if the room is 4m long and 3m wide, then dmax=5m).

[0107] Maximum allowed time difference : The set time threshold (365 days) is exceeded, and the temporal matching degree is greatly reduced, reflecting the timeliness of the evaluation.

[0108] Review release time : The timestamp when the user posted the review (e.g., 2024-06-01 10:00:00).

[0109] Current time : The current time at the time of evaluation (e.g. 2025-07-11 15:00:00).

[0110] The actual time when the defect began to appear : The defect start time determined through historical maintenance records and inspection reports (such as 2024-01-01, obtained from property work orders or professional inspections).

[0111] Time Difference : The difference between the current time and the start time of the event corresponding to the explicit feature (for example, the explicit feature "water seepage" starts recording from 2024-06-01, =395 days).

[0112] Error weight : The influence coefficient of time recording error ( ,like , the larger the error, the stronger the correction).

[0113] Attenuation coefficient : The rate of time decay ( ,like =0.01, the longer the time, the faster the matching degree decays).

[0114] The beneficial effect of the above technical solution is: through four-dimensional modeling of horizontal, vertical, diagonal (three-dimensional space) + temporal (time dimension), it can achieve cross-modal precise association, scene adaptation and multi-dimensional fusion, and indirectly improve the comprehensiveness of real estate value assessment.

[0115] The present invention proposes an online value assessment system based on real estate, wherein the contribution determination unit includes: The positioning subunit is used to locate the attribute words and feature words in the real evaluation and construct a structured description ,in, A semantic vector representing the true evaluation; An association subunit for determining the structured description Attribute association with the basic feature vector , and retain the descriptions whose attribute correlation is greater than the preset correlation to form the core value description set ,in, represents the jth core value description; m represents the number of core value descriptions; The graph construction sub-unit is used to construct a three-layer heterogeneous graph G = (V, E) based on the correlation between the core value description, the spatial mapping results, and the highly correlated implicit description. The first layer is the core value description nodes, the second layer is the spatial mapping result nodes, and the third layer is the highly correlated implicit description nodes. V represents the nodes involved in the three layers, and E represents the edges corresponding to the nodes. A three-domain extraction subunit is used to extract three-domain features associated with the core value description from the social platform topic discussion data, wherein the three-domain features include: emotion domain, communication domain, and time and space domain; The factor determination subunit is used to obtain the value factor of the core value description based on the three-layer heterogeneous graph and the three-domain features.

[0116] In this example, attribute words represent the property being reviewed (e.g., "property," "greenery," and "transportation"). Feature words describe the specific characteristics of the property (e.g., "24-hour patrol," "80% coverage," and "300 meters from the subway"). Structured descriptions associate attribute words with feature words to form attribute-feature pairs (e.g., (property, 24-hour patrol)). Semantic vectors: Using a BERT model fine-tuned for the real estate sector (such as BERT-Property), structured descriptions are converted into 768-dimensional vectors (capturing semantic associations, such as the implicit relationship between "property patrol" and "security"). For example, to evaluate "a community's green coverage rate reaches 80%" → locate the attribute word "green" and the feature word "coverage rate reaches 80%" → structure the description (green, high coverage) → BERT encoding yields a semantic vector v = [0.12, -0.35, ..., 0.21] (768 dimensions). Specifically, part-of-speech tagging is performed using NLP tools (such as HanLP) to extract nouns (attribute words) and adjectives / quantifiers (feature words). A pre-trained BERT model (fine-tuned on real estate data) is loaded to generate a semantic vector.

[0117] In this embodiment, the attribute relevance is the cosine similarity between the semantic vector of the structured description and the basic feature vector dimension (which measures the degree of relevance between the description and the property attributes).

[0118] Preset correlation: A set threshold (such as 0.6) is used to filter descriptions with high correlation.

[0119] Core value description set: highly correlated structured descriptions (such as {(greenery, high coverage), (transportation, subway proximity)}) are retained. For example, the cosine similarity between the structured description (greenery, high coverage) and the basic feature "supporting efficiency index" (related to the greening dimension) is 0.75 (>0.6), so it is included in the set; the description (property, poor attitude) has a correlation of 0.4 (<0.6) with the "supporting efficiency index" and is removed.

[0120] Specifically: for each dimension of the basic feature vector (such as "supporting efficiency index" corresponds to "greening, commerce" and other attributes), an attribute-dimension mapping table is constructed; the cosine similarity between the structured description vector and the basic feature dimension vector is calculated, and descriptions ≥ the threshold are screened.

[0121] The core value description node is each “attribute-feature” pair in the core value description set (such as node N1=(greening, high coverage)).

[0122] The spatial mapping result node is the matching degree between the explicit feature and the defect feature (such as N2=(Mx=0.92, high horizontal matching degree), associated with the parameters such as Mx and My in the previous spatial mapping).

[0123] Highly correlated implicit description nodes are implicit descriptions that are highly correlated with basic features (e.g., N3 = "convenience in life", which is correlated with the previous implicit feature mining results).

[0124] Edge E: The association relationship between nodes, with the weight being the degree of association (e.g., the edge weight between N1 and N2 is 0.8, indicating the association between greening description and horizontal matching degree).

[0125] For example: heterogeneous graph structures, such as Figure 2 As shown: First layer (core value): N1 = (greenery, high coverage), N2 = (transportation, close to subway); Second layer (spatial mapping): N3=(Mx=0.92), N4=(My=0.85); The third level (implicit description): N5 = “convenient living”, N6 = “efficient commuting”; Edge: N1-N3 weight 0.8 (greening and horizontal matching association), N2-N5 weight 0.7 (transportation and living convenience association).

[0126] Specifically: nodes store description content / matching degree, and edge weights are calculated using cosine similarity or grey relational degree.

[0127] Emotional domain characteristics: The emotional tendency (positive / negative / neutral) and intensity of the topic discussion (e.g., the positive emotion of "greening is good" accounts for 85%).

[0128] Communication domain characteristics: the scale of the topic discussion (number of readings, forwarding numbers, such as 500,000 readings) and speed (popularity growth rate, such as becoming a hot search within 24 hours).

[0129] Spatiotemporal characteristics: The time of discussion (e.g., 2025-07-01) and geographical distribution (the IP addresses of the users discussing the topic are concentrated within 5 kilometers of the property).

[0130] For example, for (greening, high coverage), extract the topic discussion data: Emotional domain: 82% positive, emotional intensity 0.8 (after normalization); Dissemination domain: 300,000 views, dissemination speed 0.8 (daily popularity growth of 50%); Time and space domain: The discussion time is concentrated in the past month, and the average distance between the IP location and the property is 2 kilometers (0.9 after normalization).

[0131] Specifically: call the platform API (such as Weibo API) to obtain the topic flow; use sentiment analysis tools (such as SnowNLP) to calculate sentiment tendencies; use time series models (ARIMA) to analyze the speed of propagation; use geographic fences (AutoNavi API) to resolve IP locations.

[0132] The value factor is a quantitative indicator that measures the degree of influence of the core value description on the value of the property (integrating graph association and topic discussion characteristics, the higher the value, the greater the influence).

[0133] Computational logic: The Graph Attention Network (GAT) learns the association weights of nodes in a heterogeneous graph (e.g., the edge weight contribution between N1 and N3 is 0.4); Weighted fusion of three domain features (emotion × 0.3 + communication × 0.4 + time and space × 0.3, such as 0.38 after fusion); The final value factor = graph weight × 0.6 + topic discussion fusion value × 0.4 (e.g., 0.4 × 0.6 + 0.38 × 0.4 = 0.392). For example, the value factor of the core value description (greening, high coverage) is 0.35 (the highest proportion among all descriptions, reflecting the strong impact of greening on value). Specifically, a graph attention model (GAT implemented in PyTorch-Geometric) is constructed, heterogeneous graph and topic discussion features are input, the model is trained to learn the weight distribution rules of the value factors, and the normalized value factors (the sum is 1) are output.

[0134] The beneficial effect of the above technical solution is: through the whole process of structured analysis → association screening → graph modeling → topic discussion integration, the fragmented description of real evaluation is converted into a precise and quantified value factor. It relies on basic features to ensure that the description is strongly bound to the property attributes, and mines multi-dimensional correlation relationships through heterogeneous graphs. It also incorporates the emotions, communication and temporal and spatial characteristics of social topic discussions. The final output value factor accurately depicts the impact weight of real evaluation on property value, provides a key basis for subsequent comprehensive value calculation, and greatly improves the accuracy and depth of the application of evaluation data in real estate valuation.

[0135] The present invention proposes an online value assessment system based on real estate, wherein the value assessment module includes: A benchmark correction unit is used to correct the basic value baseline based on the spatiotemporal attenuation coefficient: A factor analysis unit, configured to assign initial weights to the weight evaluation factors output by the deep belief network according to feature categories, and at the same time, capture temporal correlations between the weight evaluation factors; The comprehensive acquisition unit is used to obtain the comprehensive value based on the revised baseline, the initial weight and time series association, the quantitative value of the value contribution of the real evaluation, and the quantitative value of the value contribution of the revised evaluation.

[0136] In this embodiment, the spatiotemporal attenuation coefficient is a coefficient (both normalized to [0,1]) that performs attenuation correction on the basic value baseline from the time dimension (the timeliness of transaction data, the closer the more reliable) and the space dimension (the distance between the transaction property and the target property, the closer the more comparable).

[0137] Time decay: In the area where the target property is located, the average transaction price six months ago was 5 million. If the monthly decay rate u1=0.01, then the time decay coefficient is (The reference rate dropped to 94.1% after 6 months).

[0138] Spatial attenuation: A transaction property is 5 kilometers away from the target property. The kilometer attenuation coefficient is set to 0.1. Then the spatial attenuation coefficient (The reference rate drops to 66.7% outside 5 km).

[0139] Comprehensive correction: Basic value baseline Vbase=5 million, after correction Vbase′=500×0.941×0.667≈3.14 million.

[0140] Feature categories are used to classify weight factors (such as physical features such as "house type and decoration" and service features such as "property and security"), which makes it easier to assign initial weights by category and reflect the differences in importance between categories.

[0141] For example: Service features: including property (0.3), security (if present, the weight is classified into this category); Transportation characteristics: including transportation (0.25), commuting convenience (if any); Educational characteristics: including school district (0.2); Environmental characteristics: including greenery (0.15); Physical characteristics: including apartment type (0.1) and decoration standards (if any).

[0142] In this embodiment, the initial weight is the basic weight ratio allocated by feature category (reflecting the overall impact priority of the category on the property value). For example, the initial weight of the service feature is 30% (corresponding to 0.3 for property), 25% for transportation feature (corresponding to 0.25 for transportation), 20% for education, 15% for environment, and 10% for physics, totaling 100%.

[0143] In this embodiment, temporal correlation refers to the dynamic change of weight factors over time (for example, after the subway opens, the weight of the "traffic" factor will increase). For example, the original weight of the traffic factor is 0.25. After the subway opens (time t=0), the LSTM model learns the evolution of the weight over time: (When t=1 month, the weight is ≈0.295, reflecting the sustained impact of traffic improvement). Specifically, the model uses the LSTM model to train the time series data of factor weights (with a sliding window of 3 months) to capture the dynamic trend of the weights; sets trigger conditions (such as policy releases and facility openings) and calls the model in real time to update the timing association (for example, when a subway line opens, LSTM retraining is triggered).

[0144] The modified baseline Vbase′ is the basic value baseline corrected by time and space attenuation (such as 3.14 million in the example), representing the benchmark reference for the regional market.

[0145] The quantitative value contribution of the real evaluation, Vreal, is the increase or decrease in the property value caused by the real evaluation after being quantified by the value factor (e.g., "24-hour property patrol" corresponds to a value factor of 0.3, contributing +200,000). For example, to convert the value factor to the amount: through the historical transaction regression model, it is fitted that "every 0.1 weight of the property factor corresponds to a value of +50,000", so a property factor of 0.3 contributes 0.3 / 0.1×5=+150,000; Multiple factors superposition: Traffic factor 0.25 contributes +100,000, total Vreal=15+10=+250,000.

[0146] The quantitative value of the value contribution of the revised evaluation, Vcorr, is the increase or decrease in the property value after the fuzzy evaluation is revised (e.g., greening generally corresponds to a value factor of -0.1 and a contribution of -50,000). For example, a revised green coverage rate of 60% (medium) corresponds to a value factor of -0.1. Through the regression model, the contribution is -0.1 / 0.1×5=-50,000 (negative impact).

[0147] The comprehensive value calculation is a final valuation that integrates the revised baseline (regional benchmark), evaluation contribution (user experience), and combines the initial weight (category priority) and time series correlation (dynamic change).

[0148] Basic proportion: Assuming the revised baseline proportion is 60%, then 314×0.6=1.884 million; Evaluation contribution ratio: set at 40%, then (25-5) × 0.4 = 80,000; Timing adjustment: Due to the opening of the subway, the weight of the traffic factor increases by an additional 5%, contributing 500×0.05×0.4=100,000 (simplified calculation, actual reference needs to be adjusted to the baseline); Comprehensive value: 1.884+8+10=2.064 million (the example is for simplified logic; a more complex dynamic weighted model is required in practice).

[0149] Construct a dynamic weighted model Vtotal=wbase Vbase′+weval (Vreal+Vcorr), where wbase+weval=1, and weval increases dynamically with the amount of evaluation data (for example, when there are ≥100 evaluations, weval increases from 40% to 50%); Introduce timing correction term: adjust the factor contribution (such as Vreal′=Vreal by changing the weight of LSTM output Δw(t) (1+Δw(t))).

[0150] The beneficial effects of the above technical solution are: through the full process of space-time attenuation calibration benchmark, layered weighted characterization factors, time series correlation to capture dynamics, and multi-source fusion output value, it breaks through the limitations of traditional assessment static benchmark + single evaluation and improves the accuracy of valuation response to market changes and real estate characteristics.

[0151] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. An online value assessment system based on real estate, characterized in that: include: An online data acquisition module, which is used to acquire basic feature vectors and multimodal evaluation data of target properties online through a distributed crawler cluster and an open API interface. The multimodal evaluation data includes text evaluations, image evaluations, and social platform topic discussion data. A feature extraction and mining module is used to extract explicit features and implicit features from the text evaluation, and at the same time, extract defect features from the image evaluation, spatially map the explicit features with the defect features, and mine potential associations of the implicit features, and obtain a highly correlated implicit description by combining the basic feature vectors; An evaluation processing module, configured to identify true, false, and fuzzy evaluations of the text and image evaluations based on a pre-built evaluation classification model, eliminate false evaluations, perform preliminary analysis on the fuzzy evaluations, and modify the fuzzy evaluations after preliminary analysis based on spatial mapping results, highly correlated implicit descriptions, and social platform topic discussion data; The value assessment module is used to obtain gridded transaction data in the same area as the target property to obtain a basic value baseline. At the same time, based on the deep belief network, it obtains weight evaluation factors of different value types, and obtains a comprehensive value based on the basic value baseline, the true evaluation and the revised evaluation, and outputs a value assessment report.

2. The real estate-based online value assessment system according to claim 1, characterized in that: The basic feature vectors include: location entropy value, spatial topological characteristics, physical attenuation coefficient and supporting efficiency index.

3. The real estate-based online value assessment system according to claim 1, characterized in that: The evaluation processing module includes: A cross-verification unit is used to cross-verify the identified fuzzy reviews using the user identity information stored in the blockchain. If the reviewer has not completed real-name authentication or the authentication information has no geographical association with the property, the fuzzy review is assigned a false and irrelevant label; otherwise, the fuzzy review is assigned a possibly false label; A possible assignment unit is used to quantify the uncertainty of the fuzzy evaluation based on the correlation between the historical fuzzy evaluation and the actual transaction price fluctuation, convert it into expectation, entropy and super entropy, and analyze the expectation, entropy and super entropy based on a preset analysis mechanism to assign a possible reference label or a non-reference label; a retaining unit, configured to retain the fuzzy evaluation only when the fuzzy evaluation is assigned both a possible false label and a possible reference label; Otherwise, the fuzzy evaluation is eliminated, wherein the fuzzy evaluation after the preliminary analysis is the retained fuzzy evaluation.

4. The real estate-based online value assessment system according to claim 3, characterized in that: The evaluation processing module further includes: a primary correction unit, configured to perform fuzzy element mapping on the fuzzy evaluation after the preliminary analysis to obtain a quantized ternary fuzzy array, and perform a primary correction on the ternary fuzzy array in combination with the fuzzy membership constraints of the basic eigenvectors, wherein the fuzzy elements include: object elements, feature elements, and quantization elements; A contribution determination unit is configured to extract the comparison relationship between the core value description of the real evaluation and the spatial mapping result and the highly correlated implicit description, and combine it with the social platform topic discussion data to obtain the value factor of the core value description, and determine the first contribution value in combination with the exponential decay weight of the evaluation release time; an element label assigning unit, configured to assign a label to the quantified fuzzy element based on a matching relationship between the core value description and the element evaluation description of the fuzzy element after the preliminary analysis, and based on the corresponding first contribution value, wherein the label is a strong correlation label, a weak correlation label, and an uncorrelated label; A secondary correction unit, configured to perform secondary correction on the ternary fuzzy array according to the label result; The correction evaluation unit is used to obtain a correction evaluation based on the first correction result and the second correction result.

5. The real estate-based online value assessment system according to claim 1, characterized in that: The feature extraction and mining module includes: A function construction unit, configured to construct an expected function of implicit features, determine a depth level for potential association mining based on the expected function, and mine from a preset knowledge graph based on the depth level; The associated feature extraction unit is used to extract the associated features of the mining results at each level based on the basic feature vector, and perform description conversion on the associated features at each level to obtain a highly associated implicit description.

6. The real estate-based online value assessment system according to claim 1, characterized in that: The feature extraction and mining module also includes: A horizontal mapping unit is used to perform horizontal mapping between the explicit features and the defect features to determine the horizontal matching degree. ,in, The horizontal offset between the three-dimensional coordinate point of the display feature and the three-dimensional coordinate point of the defect feature; The width of the wall corresponding to the explicit feature; A vertical mapping unit is used to map the explicit features and defect features vertically to determine the vertical matching degree. ,in, It is the vertical offset between the three-dimensional coordinate point of the explicit feature and the three-dimensional coordinate point of the defect feature; The height of the wall corresponding to the explicit feature; A diagonal mapping unit is used to diagonally map the explicit features to the defect features to determine the diagonal matching degree. , where d is the absolute distance difference between the three-dimensional coordinate point of the explicit feature and the three-dimensional coordinate point of the defect feature to the diagonal line; is the standard deviation of wall dimensions; is the diagonal length of the space where the explicit feature is located; A timing mapping unit is used to perform timing mapping between the explicit features and the defect features to determine the timing matching degree. ,in, is the maximum allowed time difference, which is 365 days; The time when the corresponding evaluation was published; For the present time; The time when the defect actually begins to appear; is the vertical height of the wall based on the current time; is the vertical height of the wall based on the time the review was published; The difference between the current time and the start time of recording the event corresponding to the explicit feature; is the error weight; is the attenuation coefficient; The horizontal direction matching degree Mx, the vertical direction matching degree My, the diagonal direction matching degree Md and the temporal direction matching degree Ms are taken as the spatial mapping results.

7. The real estate-based online value assessment system according to claim 4, characterized in that: The contribution determination unit includes: The positioning subunit is used to locate the attribute words and feature words in the real evaluation and construct a structured description ,in, A semantic vector representing the true evaluation; An association subunit for determining the structured description Attribute association with the basic feature vector , and retain the descriptions whose attribute correlation is greater than the preset correlation to form the core value description set ,in, represents the jth core value description; m represents the number of core value descriptions; The graph construction sub-unit is used to construct a three-layer heterogeneous graph G = (V, E) based on the correlation between the core value description, the spatial mapping results, and the highly correlated implicit description. The first layer is the core value description nodes, the second layer is the spatial mapping result nodes, and the third layer is the highly correlated implicit description nodes. V represents the nodes involved in the three layers, and E represents the edges corresponding to the nodes. A three-domain extraction subunit is used to extract three-domain features associated with the core value description from the social platform topic discussion data, wherein the three-domain features include: emotion domain, communication domain, and time and space domain; The factor determination subunit is used to obtain the value factor of the core value description based on the three-layer heterogeneous graph and the three-domain features.

8. The real estate-based online value assessment system according to claim 1, characterized in that: The value assessment module includes: A benchmark correction unit is used to correct the basic value baseline based on the spatiotemporal attenuation coefficient: A factor analysis unit, configured to assign initial weights to the weight evaluation factors output by the deep belief network according to feature categories, and at the same time, capture temporal correlations between the weight evaluation factors; The comprehensive acquisition unit is used to obtain the comprehensive value based on the revised baseline, the initial weight and time series association, the quantitative value of the value contribution of the real evaluation, and the quantitative value of the value contribution of the revised evaluation.

Images