Automatic residence estimation method based on big data and related equipment

Abstract

The invention relates to the technical field of big data processing and real estate estimation, in particular to a residence automatic estimation method based on big data and related equipment. The method comprises the following steps: acquiring and cleaning residence historical transaction data to obtain an effective case set; processing a to-be-evaluated residence address into a standardized address and coordinates, and extracting basic and dynamic characteristics; screening similar cases based on addresses, coordinates and features, correcting prices through correction coefficients, and performing weighted calculation on estimated values; and regularly updating the case set, and adjusting the correction coefficient and the weight optimization model according to the estimated value and the actual transaction deviation. According to the method, through multi-dimensional feature fusion, accurate matching and dynamic correction, the problems of low efficiency, insufficient accuracy and insufficient dynamic factor integration of traditional valuation are solved, efficient, objective and accurate valuation is realized, and the refined requirement of one room and one price is met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data processing and real estate valuation, and particularly relates to a residential automatic valuation method based on big data and related equipment. BACKGROUND

[0002] In real estate transactions, financial mortgage, asset accounting and other scenarios, residential value evaluation is a key link. The traditional residential valuation method mainly relies on two modes: one is to conduct manual inquiry by professional valuers, the valuers need to investigate the residential conditions on site, collect surrounding market data, analyze influencing factors and issue an evaluation report. Although this method can make comprehensive judgments combined with professional experience, it has the problems of low efficiency, long evaluation period, high labor cost, and is difficult to meet the large-scale and high-frequency valuation demand. At the same time, the evaluation result is easily affected by the personal experience and subjective preference of the valuer, and the objectivity and consistency are difficult to guarantee.

[0003] Another common mode is to obtain the average price of the building through a batch price checking tool, that is, to calculate the average price based on the historical transaction data of the same building, and to use it as a reference value for the residential property in the building. However, the value of residential property is affected by many individual factors, such as specific floor, house type structure, decoration level, building age, and surrounding supporting differences. The actual value of different residential properties in the same building often has significant differences. Batch price checking can only provide a general average price reference and cannot achieve precise valuation of "one house one price", which is difficult to meet the fine demand for individual residential property valuation.

[0004] With the development of big data technology, data-driven automatic valuation methods have gradually attracted attention, which aims to analyze massive historical transaction data and residential feature data through algorithms to achieve fast valuation. However, the existing automatic valuation methods still have obvious limitations, and the integration and analysis of dynamic factors affecting residential value (such as surrounding supporting updates, traffic network changes, etc.) are insufficient, and it is difficult to reflect the influence of market changes on residential value in real time. Therefore, how to efficiently process large-scale data and comprehensively consider multi-dimensional factors to achieve automatic and precise valuation of "one house one price" for residential property has become a technical problem to be solved. SUMMARY

[0005] The present application aims to provide a residential automatic valuation method based on big data and related equipment, which aims to solve the problems of low efficiency, high cost and strong subjectivity of manual evaluation in existing residential valuation methods, and the problem that batch price checking can only provide building average price and cannot achieve precise valuation of "one house one price".

[0006] The present application aims to provide a residential automatic valuation method based on big data, comprising: obtaining residential historical transaction case data; The historical transaction case data of the house is cleaned to eliminate abnormal data and cases with missing key information, and an effective historical case set is obtained. When the house value is evaluated, an input address of the house to be evaluated is obtained, the input address is analyzed and standardized, a standardized address of the house to be evaluated and corresponding geographic coordinates are determined; Feature information related to the house to be evaluated is extracted from the effective historical case set, including basic features and dynamic features; According to the standardized address, geographic coordinates and feature information, a similar case set similar to the house to be evaluated is filtered from the effective historical case set; The transaction price of each case in the similar case set is corrected by using a predetermined correction coefficient, and the transaction price is weighted calculated by combining the weight value of each case to determine the evaluation price of the house to be evaluated.

[0007] By using the above technical scheme, the historical transaction case data of the house is obtained and cleaned to ensure the effectiveness of the data, the input address of the house to be evaluated is analyzed and standardized to realize accurate spatial positioning, the basic features and dynamic features are extracted to fully describe the house attributes, and the similar case set is filtered and the multi-dimensional correction coefficient is used to correct and weight calculate the transaction price to determine the evaluation price. The evaluation process is automated and efficient, the labor cost is greatly reduced and the evaluation response speed is improved, the basic attributes and dynamic influencing factors of the house are considered, the similar cases are accurately matched and the price is corrected, the accuracy and objectivity of the house evaluation are significantly improved, and the evaluation result can truly reflect the market difference of "one house one price". The fine demand of the house value evaluation in the real estate transaction, financial mortgage and other scenes is met.

[0008] In a possible implementation of the present application, the input address is analyzed and standardized to determine the standardized address and corresponding geographic coordinates of the house to be evaluated, which includes: The input address is subjected to text error correction processing, input errors are identified and corrected based on a predetermined address dictionary library, and a corrected address is obtained; The corrected address is disassembled into a hierarchical structure of province, city, district, street, community name, building number, and house number to generate a standardized address; The standardized address is converted into corresponding latitude and longitude coordinates by a geographic information system as the geographic coordinates of the house to be evaluated.

[0009] By adopting the technical scheme, through text error correction processing based on a preset address dictionary on the input address, wrong characters, format disorder and the like in user input can be effectively identified and corrected, and subsequent valuation deviation caused by address input error is reduced; the address after correction is decomposed into a hierarchical structure of province, city, district, street, community name, building number and door number to generate a standardized address, and unified standardization of address information is realized, providing a consistent comparison basis for address data of different sources and different formats; and through a geographic information system, the standardized address is converted into longitude and latitude coordinates, accurate spatial positioning of the residential property to be evaluated is realized, and accurate spatial reference is provided for subsequent similar case screening based on geographic location, so that the accuracy and reliability of residential property valuation are further improved, and the address information plays an effective and accurate basic role in the entire valuation process.

[0010] In a possible implementation of the present application, the feature information related to the residential property to be evaluated extracted from the effective historical case set includes basic features and dynamic features. The basic features include building area, house type, floor, decoration level, building age and building structure of the residential property to be evaluated and the effective historical cases; The dynamic features include surrounding facility status, traffic network information and regional planning information of the residential property to be evaluated and the effective historical cases; The basic features and the dynamic features are quantitatively processed, and non-numerical features are converted into quantitative parameters.

[0011] By adopting the technical scheme, through extracting the basic features (such as building area and house type) and the dynamic features (such as surrounding facility status and traffic network information) of the residential property to be evaluated and the effective historical cases, and quantitatively processing the two types of features, non-numerical features are converted into quantitative parameters, the inherent properties and dynamic environmental influences of the residential property are comprehensively and meticulously described, the physical properties of the residential property itself are covered, and external influencing factors changing with time and environment are also included, and the one-sidedness of traditional valuation caused by relying on only a single or a small number of features is overcome; meanwhile, the feature parameters after quantitative processing provide a directly operable numerical basis for subsequent similar case screening, price correction and weight calculation, so that the system can accurately measure the influence of feature differences between different residential properties on value through an algorithm, the capture ability of the valuation model for "one house one price" differences is greatly improved, and the accuracy and scientificity of automatic residential property valuation are further ensured.

[0012] In a possible implementation of the present application, the similar case set similar to the residential property to be evaluated is screened from the effective historical case set according to the standardized address, geographic coordinates and feature information, and includes: screening a candidate case in the same region from the effective historical case set based on the administrative region and street information of the standardized address; calculating a distance between the candidate case and the geographic coordinates of the residential property to be evaluated, and selecting a case with a distance less than a preset distance threshold as a location similar case; screening a case with a consistent type from the location similar case according to the type of the house and the use in the feature information; calculating a transaction time difference between the case with a consistent type and the residential property to be evaluated, and selecting a case with a time difference less than a preset time threshold to form a similar case set.

[0013] By adopting the above technical solution, the candidate case in the same region is screened based on the administrative region and street information of the standardized address, the case with similar location attributes is first locked from the macro geographic range, and the interference of irrelevant cases across regions is reduced. Then, the case with similar location is selected by calculating the geographic coordinate distance, ensuring the close relevance of the case in space, which is consistent with the characteristics that the residential value is significantly affected by the location. Then, the case with a consistent type is selected according to the type of the house and the use, which ensures the comparability of the case and the residential property to be evaluated in the functional attribute. Finally, the recent transaction case is selected by the time difference, so that the selected case can reflect the current market situation, avoid the invalidity of the price reference due to the time being too long, and through this series of progressive screening steps, the case with high similarity in location, position, type and time dimension can be accurately positioned, which provides high-quality reference basis for subsequent price correction and valuation calculation, significantly improves the matching accuracy of the similar case, and further lays a solid foundation for the accurate evaluation of the residential value.

[0014] In a possible implementation of the present application, the use of the preset correction coefficient to correct the transaction price of each case in the similar case set includes: The correction coefficient includes a transaction condition correction coefficient, a transaction date correction coefficient, and a feature difference correction coefficient; The transaction condition correction coefficient is used to correct the special transaction price in the similar case to the normal market price; The transaction date correction coefficient is based on the regional residential price index to correct the price fluctuation difference between different transaction times and the evaluation reference date; The feature difference correction coefficient is based on the difference between the feature quantitative parameter of the residential property to be evaluated and the similar case, and corrects the influence of the feature difference of the house type, floor, decoration and surrounding facilities on the price; According to the transaction condition correction coefficient, the transaction date correction coefficient and the feature difference correction coefficient, the corrected price of each similar case is calculated.

[0015] By adopting the technical scheme, the transaction condition correction coefficient, the transaction date correction coefficient and the characteristic difference correction coefficient are used to correct the transaction prices of similar cases in multiple dimensions, so that the influence of various interference factors on the valuation can be effectively eliminated: the transaction condition correction coefficient can adjust the non-market price caused by special transactions (such as urgent sale, internal transaction, etc.) to the normal market level, so as to ensure the fairness of the price benchmark; the transaction date correction coefficient can accurately reflect the market fluctuation between different transaction times and the valuation benchmark date in combination with the regional residential price index, so as to avoid the price deviation caused by the time difference; and the characteristic difference correction coefficient can correct the influence of individual attribute differences such as house type, floor, decoration and surrounding facilities on the value based on the quantitative characteristic parameter difference, so that the price of the similar case is more suitable for the actual situation of the residential property to be evaluated. Through the synergistic effect of the three types of correction coefficients, the comprehensiveness and accuracy of the price correction are greatly improved, which provides a key guarantee for the reliability of the final valuation result, and ensures that the valuation can truly reflect the market value of the residential property to be evaluated.

[0016] In a possible implementation of the present application, the method further comprises: The weight value of each similar case is allocated according to the geographic coordinate distance, the transaction time difference and the characteristic similarity between the similar case and the residential property to be evaluated, and the weight value is greater when the distance is closer, the time difference is smaller and the characteristic similarity is higher; The valuation price of the residential property to be evaluated is calculated by weighted summation based on the corrected prices of the similar cases and the corresponding weight values.

[0017] By adopting the technical scheme, the weight value is allocated according to the geographic coordinate distance, the transaction time difference and the characteristic similarity between the similar case and the residential property to be evaluated, so that the case with closer distance, closer time and higher characteristic similarity plays a greater role in the valuation, which highlights the influence of space on the value of the residential property, emphasizes the market reference value of the recent transaction case, and takes into account the influence of individual characteristic difference of the residential property on the weight, thereby avoiding the valuation deviation caused by simple average calculation. On this basis, the valuation price of the residential property to be evaluated is calculated by weighted summation, which can integrate multiple dimensional similarity factors, reasonably integrate the corrected prices of the cases according to the close degree of the cases to the residential property to be evaluated, so that the final valuation result is more suitable for the market actuality, effectively improves the accuracy and persuasiveness of the “one house one price” valuation, and meets the demand for scientific valuation in the scenes of real estate transaction, financial mortgage and the like.

[0018] In a possible implementation of the present application, the method further comprises: New residential transaction data is collected based on a preset period, and the effective historical case set is updated; According to the deviation feedback of historical evaluation results and actual transaction prices, the correction coefficient and the weight distribution rule are adjusted, and the evaluation model is optimized.

[0019] By adopting the above technical scheme, by collecting new residential transaction data in a preset period and updating the effective historical case set, it can be ensured that the evaluation model is always based on the latest market transaction information, and timely reflects the dynamic changes of the real estate market, avoids the evaluation deviation caused by outdated data; at the same time, according to the deviation feedback of historical evaluation results and actual transaction prices, the correction coefficient and the weight distribution rule are dynamically adjusted, which can continuously optimize the parameter setting in the continuous learning of the model, automatically calibrate the evaluation logic, gradually improve the adaptability to complex market environment and residential feature difference, and then significantly enhance the accuracy and stability of the evaluation model, realize the upgrade from "static evaluation" to "dynamic optimization", guarantee the system to provide reliable "one house one price" precise evaluation service for a long time, and effectively cope with the challenges brought by market fluctuations and residential feature diversification.

[0020] The second object of the present application is to provide an automatic residential evaluation system based on big data, which comprises: Residential historical transaction case data acquisition module: acquiring residential historical transaction case data; Effective historical case set acquisition module: cleaning and processing the residential historical transaction case data, eliminating abnormal data and cases with missing key information, and obtaining an effective historical case set; To be evaluated residential input address processing module: when evaluating the value of a house, the input address of the house to be evaluated is obtained, the input address is analyzed and standardized, the standardized address and corresponding geographic coordinates of the house to be evaluated are determined; Residential feature information extraction module: extracting feature information related to the house to be evaluated from the effective historical case set, the feature information including basic features and dynamic features; Similar case set acquisition module: according to the standardized address, geographic coordinates and feature information, similar cases similar to the house to be evaluated are screened out from the effective historical case set; To be evaluated residential evaluation price determination module: the transaction prices of each case in the similar case set are corrected by using a preset correction coefficient, and the evaluation price of the house to be evaluated is determined by weighted calculation combined with the weight value of each case.

[0021] By adopting the technical scheme, the historical transaction case data of the residence is acquired and cleaned to ensure the validity of the data, accurate spatial positioning is realized by combining the input address resolution and standardization processing of the residence to be evaluated, the basic features and dynamic features are extracted to comprehensively depict the residence attributes, and the transaction price is corrected and weighted calculated to determine the evaluation price through the similar case screening and multi-dimensional correction coefficient, which can effectively solve the problems of low efficiency, high cost and strong subjectivity of traditional manual evaluation, and the problem that batch price checking cannot realize accurate evaluation of "one house one price", not only realizes automation and high efficiency of the evaluation process by integrating multi-source information through big data technology, greatly reduces the labor cost and improves the evaluation response speed, but also significantly improves the accuracy and objectivity of the residence evaluation by comprehensively considering the basic attributes and dynamic influence factors of the residence, combining the accurate matching and price correction of similar cases, and ensuring that the evaluation result can truly reflect the market difference of "one house one price", meeting the fine demand of residence value evaluation in real estate transaction, financial mortgage and other scenes.

[0022] The third object of the present application is to provide a residence automatic evaluation equipment based on big data, which comprises: a memory and a processor, the memory stores a computer program capable of being loaded and executed by the processor to execute the residence automatic evaluation method based on big data.

[0023] The fourth object of the present application is to provide a storage medium.

[0024] The fourth object of the present application is achieved by the following technical scheme: A storage medium, wherein the computer program capable of being loaded and executed by the processor to execute the residence automatic evaluation method based on big data is stored.

[0025] In summary, the present application includes at least one of the following beneficial technical effects: 1. By obtaining residential historical transaction case data and cleaning it to ensure data validity, combining the input address resolution and standardization processing of the residential property to be evaluated to achieve accurate spatial positioning, extracting basic features and dynamic features to fully depict residential attributes, and then modifying and weighting the transaction price through similar case screening and multi-dimensional correction coefficient to determine the estimated value, it can effectively solve the problems of low efficiency, high cost and strong subjectivity of traditional manual evaluation, and the problem that batch price checking cannot achieve accurate valuation of "one house one price". Not only does it integrate multi-source information through big data technology to realize the automation and efficiency of the valuation process, significantly reducing labor costs and improving valuation response speed, but also by considering the basic attributes and dynamic influencing factors of the residence, combined with the accurate matching of similar cases and price modification, it significantly improves the accuracy and objectivity of residential valuation, ensuring that the valuation result can truly reflect the market differences of "one house one price" and meet the fine needs of residential value evaluation in real estate transactions, financial mortgage and other scenarios.

[0026] 2. By extracting the basic features (such as building area, house type, etc.) and dynamic features (such as surrounding supporting facility status, traffic network information, etc.) of the residential property to be evaluated and effective historical cases, and quantifying both types of features, non-numerical features are converted into quantitative parameters, which can fully and meticulously depict the inherent attributes and dynamic environmental influences of the residence, covering both the stable physical properties of the residence itself and the external influencing factors that change over time and environment, overcoming the one-sidedness of traditional valuation that relies on only a single or a small number of features. At the same time, the quantified feature parameters provide a directly operable numerical basis for subsequent similar case screening, price modification and weight calculation, enabling the system to accurately measure the influence of feature differences between different residences on value through algorithms, thereby significantly improving the ability of the valuation model to capture "one house one price" differences, further ensuring the accuracy and scientificity of residential automatic valuation. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a flowchart of a residential automatic valuation method based on big data provided by an embodiment of the present application; Figure 2 is a virtual structure diagram of a residential automatic valuation system based on big data provided by an embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0029] In addition, the term "and / or" in the present application is only used to describe the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can mean that A exists alone, A and B exist together, and B exists alone. In addition, the character " / " in the present application generally represents an "or" relationship between the associated objects unless otherwise specified.

[0030] The embodiments of the present application will be described in further detail below with reference to the drawings of the specification.

[0031] The embodiments of the present application provide a residential automatic valuation method based on big data, referring to Figure 1 The main process of the method is described as follows. S1: Obtain residential historical transaction case data; The residential historical transaction case data includes but is not limited to transaction price, transaction time, specific address, building area, house type, floor, decoration level, building age, surrounding supporting facilities and traffic network and other information of different residences, which are collected from multiple channels such as real estate trading platform, real estate registration system and the like through network crawler, third-party data interface and the like, so as to ensure the extensive coverage and diversity of the data and provide sufficient data basis for subsequent valuation analysis.

[0032] S2: The residential historical transaction case data is cleaned and processed to eliminate abnormal data and cases with missing key information, and an effective historical case set is obtained; The abnormal data includes transaction cases obviously deviating from the reasonable price range of the market (such as transactions far below or far above the average price of the same type of residence in the same area), and the cases with missing key information include records without clear transaction time, address or core characteristic parameters. The data is automatically cleaned through data verification algorithm (such as standard deviation method to identify outliers and null value detection method to select complete records), so as to ensure the data quality of the effective historical case set and avoid the interference of low-quality data on the subsequent valuation accuracy.

[0033] S3: When the residential value is evaluated, the input address of the residence to be evaluated is obtained, the input address is analyzed and standardized, the standardized address of the residence to be evaluated and the corresponding geographic coordinates are determined; The input address can be text information manually input by a user (such as "XX City XX District XX Community X Building X Unit XXX Room"), the analysis and standardization processing includes format normalization of the address text (such as unified expression of building and unit), hierarchical information extraction (such as separation of province, city, district, community, and house number), and mapping of the standardized address to latitude and longitude coordinates through geographic coding technology, realizing spatial positioning of the residential property to be evaluated, and providing a basis for subsequent spatial dimension screening of similar cases.

[0034] S4: extracting feature information related to the residential property to be evaluated from the effective historical case set, the feature information including basic features and dynamic features; The basic features are inherent properties of the residential property, such as building area, house type (such as one bedroom and one living room, two bedrooms and two living rooms), floor (and total number of floors), decoration level (such as bare, simple, and elaborate), building age, building structure (such as brick and concrete, and frame), etc. The dynamic features are external environment and market factors that affect the value of the residential property, such as the distribution and level of supporting facilities within 3 kilometers, such as schools, hospitals, and commercial districts, the straight-line distance to the subway station and the bus station, the recent fluctuation trend of the regional residential price, etc. The feature quantization algorithm (such as converting the decoration level to a quantization value of 0-10 points) realizes the computability of non-numerical features, and provides multi-dimensional parameters for similarity analysis.

[0035] S5: screening similar case sets from the effective historical case set according to the standardized address, geographic coordinates, and feature information; The screening process combines geographical space and feature attributes in two dimensions: calculating the straight-line distance between the case and the residential property to be evaluated based on geographic coordinates, and selecting cases with a distance less than a preset threshold (such as 1 kilometer) as spatial candidates; then, through a feature matching algorithm (such as calculating the similarity score of basic features and dynamic features), cases with consistent house types and feature parameters within a preset range are selected from the spatial candidates, and transaction records within the past two years are selected in combination with transaction times, to finally form a similar case set, ensuring that the screening result has high comparability with the residential property to be evaluated.

[0036] S6: using a preset correction coefficient to correct the transaction prices of each case in the similar case set, and performing weighted calculation in combination with the weight values of each case to determine the estimated value of the residential property to be evaluated.

[0037] Among them, the correction coefficient includes the correction coefficient for special transaction conditions (such as forced sale, internal transfer, and other non-market normal transactions), the transaction date correction coefficient based on the regional price index (such as correcting the transaction price six months ago to the current point price according to the monthly price increase), and the correction coefficient for differences in characteristics such as house type, floor, and decoration (such as the premium coefficient of high-rise compared to mid-rise, and the value-added coefficient of fine decoration compared to simple decoration); the weight value is dynamically allocated according to the spatial distance, transaction time, and feature similarity between the case and the residential property to be evaluated, and the closer the distance, the closer the time, and the higher the similarity, the greater the weight. The final evaluation is calculated by a weighted sum formula (such as evaluation price = Σ (case corrected price x corresponding weight)) to achieve accurate output of "one house one price".

[0038] Specifically, in some possible embodiments, the resolving and standardizing the input address to determine the standardized address and corresponding geographic coordinates of the residential property to be evaluated comprises: performing text error correction processing on the input address, identifying and correcting input errors based on a preset address dictionary library to obtain a corrected address; disassembling the corrected address into a hierarchical structure of province, city, district, street, community name, building number, and house number to generate a standardized address; converting the standardized address into corresponding latitude and longitude coordinates through a geographic information system as the geographic coordinates of the residential property to be evaluated.

[0039] Among them, in actual application scenarios, the address input by the user may have various non-standard situations, such as "Beijing Haidian District Zhongguancun South Street No. 5" may be misinput as "Beijing Haidian District Zhongguancun South Street No. 5" or "Beijing Haidian District Zhongguancun Street No. 5". The system first calls a text correction model based on deep learning, which uses a BERT pre-training language model for address text correction based on a preset address dictionary library. The dictionary library not only contains basic terms such as standard administrative region names and community names, but also incorporates professional terms and common address abbreviations (such as "garden" "garden" "building" suffixes) in the field of real estate. In the correction process, the model not only corrects common errors, but also intelligently adjusts non-standard address expressions, such as automatically associating "Zhongguancun Street" to the standard expression "Zhongguancun South Street".

[0040] After completing the text correction, the system uses a hierarchical address resolution algorithm to structurally decompose the corrected address. This algorithm combines rule matching and machine learning classifiers. First, it identifies administrative region information such as provinces, cities, and districts in the address through regular expressions. Then, it uses a Conditional Random Field (CRF) model to perform sequence labeling on the remaining text, identifying hierarchical structures such as streets, communities, buildings, and house numbers. For complex address expressions, the system performs semantic analysis and context reasoning. For example, "Chaoyang District, Jianguo Road, 88, Modern City SOHO A" is accurately decomposed into: province (Beijing), city (Beijing), district (Chaoyang), street (Jianguo Road), community name (Modern City SOHO), building number (A), and house number (88). In particular, for addresses that lack some hierarchical information (such as "Haidian District, Zhongguancun"), the system intelligently completes the missing information by combining historical data and geographic spatial relationships to infer possible street and community information.

[0041] After obtaining the standardized address, the system converts the address to latitude and longitude by integrating multiple geographic information system (GIS) services. It uses a multi-engine fusion strategy to simultaneously call multiple GIS interfaces such as Baidu Map, Gaode Map, and Tencent Map for coordinate conversion. It then uses a consistency verification algorithm to compare and optimize multiple results. For old communities or newly developed areas with coordinate offsets, the system combines high-precision electronic maps and historical transaction case coordinate distributions for secondary calibration to eliminate positioning deviations caused by map precision differences. In addition, the system introduces a three-dimensional geographic information model to calculate more accurate spatial position coordinates for high-rise buildings, further improving the accuracy of geographic positioning.

[0042] Specifically, in some possible embodiments, the feature information related to the to-be-evaluated residence is extracted from the effective historical case set, and the feature information includes basic features and dynamic features, including: The basic features include the building area, house type, floor, decoration level, building age, and building structure of the to-be-evaluated residence and the effective historical cases; The dynamic features include the status of supporting facilities, traffic network information, and regional planning information around the to-be-evaluated residence and the effective historical cases; The basic features and dynamic features are quantitatively processed to convert non-numerical features into quantitative parameters.

[0043] In extracting the basic features, instead of simply listing the information, the core influencing dimensions of residential value assessment are combined for structured screening: the building area is directly quantified by the value accurate to the square meter; the layout type not only distinguishes the basic forms such as "one room and one living room" and "three rooms and two living rooms", but also enhances the distinction through derived parameters such as the number of light surfaces and the rationality of the traffic line (for example, the "north-south through" layout is quantified as a light coefficient of 1.2, and the "full south" as 1.1); the floor feature is associated with the total number of floors to calculate the relative value (for example, in a 30-story building, the floor coefficient of the 15th floor is 0.95, and the 25th floor is 1.05); the decoration level introduces a "decoration cost index", which is subdivided into 10 quantified levels according to building material standards and construction technology (for example, simple decoration corresponds to levels 3-5, and high-end decoration corresponds to levels 6-8); the building age calculates the depreciation coefficient through the difference from the average building age in the region (for example, a 10-year-old building has a depreciation coefficient of 0.92 compared to the average building age of 5 years in the region); and the building structure is given a quantified weight according to the seismic grade and material durability (for example, a frame structure has a weight of 15% higher than a brick-concrete structure). In extracting dynamic features, the limitations of static data are broken through: the supporting facility status not only covers basic elements such as schools and hospitals, but also obtains dynamic scores such as school admission rates and hospital treatment levels through real-time data interfaces (for example, the supporting coefficient of a residence near a provincial key school is 1.15); the traffic network information integrates real-time traffic data to refine the "1 km from the subway" into a time cost quantified value of "10 minutes of commuting time during the morning peak"; and the regional planning information introduces a "planning landing progress coefficient" to give dynamic weights of 0.3, 0.7, and 1.0 to different stages such as approved but not started, under construction, and completed. In the quantification process, the Analytic Hierarchy Process (AHP) is used to construct a feature weight matrix, and gradient mapping is performed on non-numeric features (for example, "near commercial district" is converted to a continuous quantified parameter of 0.8-1.2 according to the distance from the core area of the commercial district), so that the basic features and dynamic features form unified dimension data that can be directly operated, preserving the unique influence of each feature while achieving the organic integration of multi-dimensional data, providing a fine data foundation for subsequent similar case matching and price correction.

[0044] Specifically, in some possible embodiments, the filtering, from the effective historical case set, a similar case set similar to the to-be-evaluated residence according to the standardized address, geographic coordinates, and feature information comprises: filtering, from the effective historical case set, candidate cases in the same region based on administrative district and street information of the standardized address; calculating a geographic coordinate distance between the candidate cases and the to-be-evaluated residence, and selecting cases with a distance less than a preset distance threshold as position similar cases; filtering, from the position similar cases, cases with consistent types according to layout types and uses in the feature information; Calculate the time difference between the type consistent case and the transaction time of the residential property to be evaluated, select the cases with a time difference less than the preset time threshold, and form a similar case set.

[0045] In practical applications, the system first constructs a multi-level screening index based on the administrative district and street information of the standardized address. For example, when the residential property to be evaluated is located in "Zhongguancun Street, Haidian District, Beijing", the system will quickly locate all cases in the effective historical case set that are marked as "Haidian District" and have a street field containing "Zhongguancun", forming a preliminary candidate set. To improve screening efficiency, the system uses the Geohash technology to index the historical cases by space, encodes the geographic coordinates into a string prefix, and quickly narrows down the search range through prefix matching. This method can improve the query speed by more than 50% compared to traditional latitude and longitude distance calculation.

[0046] When calculating the geographic coordinate distance, the system not only considers the straight-line distance, but also introduces the concept of "traffic accessibility distance". By calling real-time traffic APIs, the system calculates dynamic parameters such as driving time and bus transfer times between candidate cases and the residential property to be evaluated, and converts them into equivalent distances. For example, two cases with a straight-line distance of 1 km but a detour of 3 km will be assigned a higher actual distance weight. The preset distance threshold is dynamically adjusted according to the size of the city, set to 800 meters for first-tier cities and 1200 meters for second-tier cities, while allowing users to customize the adjustment through an interface slider to enhance flexibility.

[0047] When filtering the type and purpose of the house, the system uses a semantic similarity matching algorithm. For standard house types such as "three-bedroom two-hall", direct accurate matching is used, and for house types such as "large two-bedroom" and "duplex" with ambiguous descriptions, the system uses a natural language processing model to analyze the house description text in historical cases, extracts key features such as room count and area ratio, constructs a house vector space, and calculates the cosine similarity. For example, the similarity between "2 room 2 hall 1 bathroom + variable space" and "2 room 2 hall" can be quantified as 0.85, and when the similarity threshold is set to 0.8, it can be considered as a match. Purpose screening combines real estate registration information and actual use to distinguish between "residential", "commercial and residential", "apartment", and other categories, ensuring that the purpose of the case is strictly consistent.

[0048] When calculating the transaction time difference, the system introduces a "market volatility sensitivity coefficient". In the period of rapid housing price rise (monthly increase > 1.5%), the preset time threshold is automatically shortened to 3 months, while in the market stable period it can be relaxed to 6 months. At the same time, the system will dynamically adjust the historical transaction price according to the regional housing price index, for example, if the transaction time of a case is 6 months ago and the regional housing price index has risen by 5%, the system will automatically correct the case price to the current level before participating in the similarity calculation. For cases that exceed the time threshold but have extremely similar features (such as the same building and same type), the system will separately mark and provide a manual confirmation option, ensuring both screening efficiency and not missing special value cases.

[0049] Specifically, in some possible embodiments, the using a preset correction coefficient to correct the transaction price of each case in the similar case set comprises: The correction coefficient includes a transaction condition correction coefficient, a transaction date correction coefficient, and a feature difference correction coefficient; The transaction condition correction coefficient is used to correct special transaction prices in similar cases to normal market prices; The transaction date correction coefficient is based on the regional residential price index to correct the price fluctuation difference between different transaction times and the evaluation benchmark date; The feature difference correction coefficient is based on the difference in feature quantitative parameters between the to-be-evaluated residence and the similar cases to correct the influence of the feature difference of the house type, floor, decoration, and surrounding facilities on the price; According to the transaction condition correction coefficient, the transaction date correction coefficient, and the feature difference correction coefficient, the corrected price of each similar case is calculated.

[0050] Among the transaction condition correction link, the system does not simply rely on artificial preset rules, but uses machine learning models to deeply mine historical transaction data and automatically identify special transaction patterns. For example, by analyzing multiple dimensions such as price deviation, transaction frequency, and buyer-seller relationship, a special transaction identification model is constructed to accurately label non-market transactions such as internal transfers, debt settlements, and urgent sales. The value of the transaction condition correction coefficient is based on the statistical distribution of normal market transaction prices of similar residences in the same area, such as the fact that urgent sale cases are usually 8-12% lower than normal prices. The system will automatically calculate and dynamically update this correction interval based on the price center of recent similar normal transactions, ensuring that special transaction prices are corrected to the objective market level.

[0051] The generation of the transaction date correction coefficient breaks through the limitations of traditional fixed cycle indexes and adopts a dynamic index model that is updated in real time. The system captures the latest transaction data in the region every day and constructs a residential price index that is subdivided to the street level through time series analysis algorithms. It not only includes overall market trends but also calculates price fluctuation coefficients for different types of residences, such as different types of residences. For example, during the adjustment period of the school district housing policy, the system can sensitively capture the differences in the price fluctuation range of 90-120㎡ three-bedroom houses in the corresponding area compared to other types of residences, and generate more accurate date correction coefficients for this type of residence, so that the transaction prices three months ago can be more closely related to the actual changes in the current market after correction.

[0052] The calculation of the characteristic difference correction coefficient reflects multi-dimensional and refined consideration. For the characteristics of the house type, not only the number of rooms and the layout are considered, but also the sunlight duration difference of different house types is calculated through lighting simulation algorithm and converted into a quantitative correction factor (such as 0.5% price correction for every 1 hour increase in sunlight duration); the floor correction introduces a "relative floor value curve", which dynamically adjusts factors such as total floor height, whether it is street-facing, whether it has an elevator, etc. For example, in a 30-story building, the 15-20th floor may have a 1.2 times correction coefficient due to the advantage of the view, while the ground floor may have a 0.8-1.0 difference coefficient if it has a garden; the decoration level correction is related to the price index of the main decoration materials in the past year and the construction cost data to form a dynamic updating decoration value evaluation model, avoiding the deviation caused by traditional fixed coefficients; the surrounding supporting correction integrates real-time operation data, such as the latest admission rate of schools, the daily passenger flow of commercial areas, and the diagnosis and treatment capacity of hospitals, to convert them into quantitative parameters, so that the influence of characteristic differences on prices can be more accurately reflected.

[0053] In the calculation of the corrected price, the system adopts a dynamic weighted fusion strategy rather than simply multiplying the coefficients. By analyzing the historical correction effect, the system automatically learns the influence weight of different correction coefficients in specific scenarios, for example, in a period of intense market fluctuations, the weight of the transaction date correction coefficient will automatically increase to 40%, while in a period of market stability, the weight of the characteristic difference correction coefficient can increase to 50%, so that the final corrected price can not only cover all kinds of influencing factors, but also highlight the role of key correction items according to the actual market situation, providing more reliable basic data for subsequent weighted calculation.

[0054] Specifically, in some possible embodiments, the combination of the weight values of each case for weighted calculation to determine the estimated value of the residence to be evaluated includes: According to the geographical coordinate distance, the transaction time difference and the characteristic similarity between the similar cases and the residence to be evaluated, the weight values of each similar case are assigned, the closer the distance, the smaller the time difference, the higher the characteristic similarity, the greater the weight value; Based on the revised prices of each similar case and the corresponding weight values, the estimated value of the residential property to be evaluated is calculated by weighted summation.

[0055] In practical applications, the system dynamically assigns weights to similar cases using a three-dimensional spatial weight distribution model. The geographical coordinate distance dimension introduces a Gaussian decay function, with the residential property to be evaluated as the center. Cases within a radius of 1 kilometer are weighted inversely proportional to the square of the distance, for example, cases 0.5 kilometers away have a weight of 0.8, and cases 0.8 kilometers away have a weight of 0.4. Cases that exceed 1 kilometer but are within a 3-kilometer range have an exponentially decaying weight below 0.1. In combination with traffic network data, cases along the subway line are additionally given a weight coefficient of 0.15, making the influence of spatial distance more consistent with actual market perception.

[0056] The transaction time difference dimension constructs a weight curve that changes dynamically over time. The system automatically adjusts the time decay factor based on the regional housing price fluctuation rate. In the rapid rising period when the monthly housing price increase exceeds 1.5%, the weight decays by 15% for every 1 month increase in time difference. In the market stable period (monthly increase <0.5%), the time decay factor is reduced to 5%. For historical cases exceeding 12 months, the system will make a second revision based on regional planning changes (such as new schools, subway openings, etc.) during that period. If there is a major favorable planning in place, the weight can be adjusted by 30%-50%, ensuring that historical cases can still provide effective reference under certain conditions.

[0057] The feature similarity dimension uses a multi-factor weighted cosine similarity algorithm. The basic features (building area, layout, etc.) and dynamic features (supporting facilities, transportation, etc.) are constructed into a feature vector space in a 7:3 ratio. The building area difference is controlled within ±15%, and each 1% exceeds 0.03 of the similarity score. The layout matching introduces a space utilization rate coefficient, for example, the evaluated residential property is a "three open south" layout, and the similar case is a "two open south" layout. The space utilization difference corresponds to a deduction of 0.12 in similarity score. For dynamic features, the system real-time grabs supporting data within 3 kilometers. Each additional key school or subway station increases the similarity score of related cases by 0.08.

[0058] In the weighted summation stage, the system introduces a confidence interval correction mechanism. First, calculate the standard deviation of the corrected prices of all similar cases, remove abnormal cases that exceed the mean value ± 2 standard deviations, and then use the iterative weighting method to calculate: the first calculation obtains the initial valuation, the second calculation gives higher weight to the cases closer to the initial valuation (up to 20% higher), and after 3-5 iterations, the confidence of the final valuation can be improved to more than 95%. At the same time, the system can automatically identify the price clustering phenomenon in similar cases. If there are two obvious price intervals (such as 100,000 yuan / square meter and 120,000 yuan / square meter), the weighted average of each cluster is calculated, and the final valuation is generated according to the case quantity ratio, effectively dealing with the market price stratification phenomenon.

[0059] Specifically, in some possible embodiments, the method further comprises: Based on the preset period, collect new residential transaction data, and update the effective historical case set; According to the deviation feedback of the historical valuation results and the actual transaction price, adjust the correction coefficient and the weight distribution rule, and optimize the valuation model.

[0060] Wherein, when collecting new residential transaction data based on the preset period, instead of using fixed period mechanical update, it is combined with the dynamic adjustment of regional market activity to shorten the update frequency to daily in the hot area with dense transactions or market fluctuation period, and to extend to weekly in the area with flat transactions. At the same time, through intelligent screening algorithm, the newly added cases with high matching degree to recent valuation demand (such as specific house type, specific price segment) are preferentially included, and the repeated or low value data are automatically marked and removed, so as to ensure that the effective historical case set reflects the latest market dynamics while keeping the size simple. In the deviation feedback link, the system does not simply adjust the parameters according to the overall deviation, but builds a multi-dimensional deviation analysis model: according to the dimensions of administrative region, house type, price interval, etc., the historical valuation deviation data is split to identify the subdivision rules such as "the valuation of 90-120 square meters three-bedroom in a certain area is generally low" and "the correction of residential property with a total price of 5 million yuan is insufficient", and then the correction coefficient (such as increasing the floor correction coefficient gradient of three-bedroom in this area from 0.02 to 0.03) and weight distribution rule (such as increasing the weight proportion of recent transaction cases of the same house type) of the corresponding dimension are adjusted. Through this "dynamic data update + accurate deviation tracing + targeted parameter optimization" closed loop mechanism, the valuation model can adapt to market changes, avoid precision decay after long-term use, and realize the upgrade from "static model" to "self-evolving system".

[0061] Another embodiment of the present application provides a residential automatic valuation system based on big data, wherein Figure 2 A residential automatic valuation system based on big data comprises: The residential history transaction case data acquisition module 100 acquires residential history transaction case data. The effective history case set acquisition module 200: The effective history case set is obtained by cleaning the residential history transaction case data and eliminating abnormal data and cases with missing key information. The input address processing module 300 of the residential property to be evaluated: When the residential property value is evaluated, the input address of the residential property to be evaluated is acquired, the input address is analyzed and standardized, the standardized address of the residential property to be evaluated and the corresponding geographic coordinates are determined. The feature information extraction module 400 of the residential property to be evaluated: The feature information related to the residential property to be evaluated is extracted from the effective history case set, and the feature information includes basic features and dynamic features. The similar case set acquisition module 500: According to the standardized address, the geographic coordinates and the feature information, the similar case set is screened from the effective history case set. The estimated value price determination module 600 of the residential property to be evaluated: The transaction prices of each case in the similar case set are corrected by using a preset correction coefficient, and the estimated value price of the residential property to be evaluated is determined by combining the weight values of each case.

[0062] The residential automatic valuation system based on big data provided in the embodiment can realize the steps of the foregoing embodiment, and thus can achieve the same technical effects as the foregoing embodiment. The principle analysis can be referred to the related description of the steps of the foregoing residential automatic valuation method based on big data, and will not be repeated here.

[0063] The embodiment of the present application also provides a residential automatic valuation device based on big data, which comprises a memory and a processor, and the memory stores a computer program capable of being loaded and executed by the processor to perform the foregoing residential automatic valuation method based on big data.

[0064] The embodiment of the present application also provides a storage medium, which stores a computer program capable of being loaded and executed by the processor to perform the foregoing residential automatic valuation method based on big data.

[0065] The storage medium provided in the embodiment can realize the steps of the foregoing embodiment, and thus can achieve the same technical effects as the foregoing embodiment. The principle analysis can be referred to the related description of the foregoing method steps, and will not be repeated here.

[0066] The storage medium includes, for example, a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0067] The steps of the methods or algorithms described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module can reside in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0068] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Also, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples, without contradiction.

[0069] In addition, the term defining the features of "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three, etc., unless otherwise specifically limited, only for descriptive purposes, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features.

[0070] Therefore, any process or method descriptions in the flow charts or otherwise described herein can be understood as representing code modules, segments, or portions of code that include one or more executable instructions for implementing the specified logical functions or processes, and the scope of the preferred embodiments of the present application encompasses additional implementation in which the functions are performed in a different order, in substantially simultaneous fashion, or in reverse order, as appropriate, depending upon the functions involved, as will be understood by those skilled in the art.

[0071] The embodiments of the specific implementation are the preferred embodiments of the present application, and do not limit the protection scope of the present application, so that: any equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A big data based residential automatic valuation method, characterized in that, The method comprises the following steps: obtaining historical transaction data of residential properties; cleaning the historical transaction data of residential properties to remove abnormal data and cases with missing key information, and obtaining an effective historical case set; when evaluating the value of a residential property, obtaining an input address of the residential property to be evaluated, analyzing and standardizing the input address, determining the standardized address of the residential property to be evaluated and the corresponding geographic coordinates; extracting feature information related to the residential property to be evaluated from the effective historical case set, the feature information including basic features and dynamic features; selecting a similar case set from the effective historical case set according to the standardized address, geographic coordinates and feature information, the similar case set being similar to the residential property to be evaluated; using a preset correction coefficient to correct the transaction price of each case in the similar case set, and combining the weight value of each case to perform weighted calculation, thereby determining the evaluation price of the residential property to be evaluated.

2. The method of claim 1, wherein, The method comprises the following steps: performing text error correction on the input address, identifying and correcting input errors based on a preset address dictionary, and obtaining a corrected address; decomposing the corrected address into a hierarchical structure of province, city, district, street, community name, building number and door number, and generating a standardized address; converting the standardized address into corresponding latitude and longitude coordinates through a geographic information system as the geographic coordinates of the residential property to be evaluated.

3. The method of claim 1, wherein, The method comprises the following steps: extracting basic features, including building area, house type, floor, decoration level, building age and building structure of the residential property to be evaluated and the effective historical cases; extracting dynamic features, including the status of supporting facilities, traffic network information and regional planning information around the residential property to be evaluated and the effective historical cases; quantifying the basic features and dynamic features, and converting non-numerical features into quantified parameters.

4. The method of claim 1, wherein, The method comprises the following steps: based on the administrative district and street information of the standardized address, selecting candidate cases within the same region from the effective historical case set; calculating the geographic coordinate distance between the candidate cases and the residential property to be evaluated, and selecting cases with a distance less than a preset distance threshold as location similar cases; selecting cases with consistent types from the location similar cases according to the house type and purpose in the feature information; calculating the transaction time difference between the cases with consistent types and the residential property to be evaluated, and selecting cases with a time difference less than a preset time threshold to form a similar case set.

5. The method of claim 1, wherein, The method comprises the following steps: the correction coefficient includes a transaction condition correction coefficient, a transaction date correction coefficient and a feature difference correction coefficient; the transaction condition correction coefficient is used to correct special transaction prices in similar cases to normal market prices; The transaction date correction coefficient is based on a regional residential price index, and corrects the price fluctuation difference between different transaction times and the evaluation reference date; The feature difference correction coefficient is based on the feature quantitative parameter difference between the residential property to be evaluated and similar cases, and corrects the influence of the feature difference such as house type, floor, decoration, and surrounding supporting facilities on the price; According to the transaction condition correction coefficient, the transaction date correction coefficient, and the feature difference correction coefficient, the corrected prices of each similar case are calculated.

6. The method of claim 1, wherein, The weighting calculation combined with the weight values of each case determines the evaluation price of the residential property to be evaluated, which includes: According to the geographical coordinate distance, the transaction time difference, and the feature similarity between similar cases and the residential property to be evaluated, weight values are assigned to each similar case. The closer the distance, the smaller the time difference, and the higher the feature similarity, the greater the weight value. Based on the corrected prices of each similar case and the corresponding weight values, the evaluation price of the residential property to be evaluated is calculated by weighted summation.

7. The method of claim 1, wherein, The method further includes: Based on the preset period, new residential transaction data is collected, and the effective historical case set is updated; According to the deviation feedback of the historical evaluation results and the actual transaction price, the correction coefficients and the weight distribution rules are adjusted to optimize the evaluation model.

8. A big data based residential auto valuation system, characterized in that, It includes: Residential historical transaction case data acquisition module: acquire residential historical transaction case data; Effective historical case set acquisition module: clean the residential historical transaction case data, eliminate abnormal data and cases with missing key information, and obtain an effective historical case set; Input address processing module of residential property to be evaluated: when performing residential value evaluation, acquire the input address of the residential property to be evaluated, analyze and standardize the input address, determine the standardized address and corresponding geographical coordinates of the residential property to be evaluated; Residential property to be evaluated feature information extraction module: extract feature information related to the residential property to be evaluated from the effective historical case set, including basic features and dynamic features; Similar case set acquisition module: according to the standardized address, geographical coordinates, and feature information, filter similar cases from the effective historical case set; Residential property to be evaluated evaluation price determination module: correct the transaction price of each case in the similar case set using the preset correction coefficient, and determine the evaluation price of the residential property to be evaluated by weighting calculation combined with the weight values of each case. 9.A big data-based residential automatic valuation device, characterized by, It includes: A memory and a processor, the memory stores a computer program capable of being loaded and executed by the processor to execute the automatic residential evaluation method based on big data according to any one of claims 1-7.

10. A storage medium, characterized by A computer program capable of being loaded and executed by the processor to execute the automatic residential evaluation method based on big data according to any one of claims 1-7.

Images