Abnormal housing resource identification method and device, storage medium and program product

Abstract

The embodiment of the invention relates to the technical field of artificial intelligence, in particular to an abnormal housing resource identification method and device, a storage medium and a program product. Target housing resource information of a to-be-identified housing resource is acquired; inputting the target housing resource information into a first recognition model and a second recognition model respectively, and outputting a first recognition result and a second recognition result of the to-be-recognized housing resource; carrying out fusion processing based on the first identification result and the second identification result, and generating a target identification result of the to-be-identified housing resource; according to the technical scheme, by introducing the two recognition models, analysis of the target house resource information in two dimensions of dealing and non-dealing is realized, so that the first recognition result and the second recognition result have higher accuracy in the corresponding dimensions; thereafter, through the fusion of the first recognition result and the second recognition result, the finally obtained target recognition result integrates the advantages of the two recognition models, so that the final recognition result is more accurate.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a method, device, storage medium, and program product for identifying abnormal housing listings. Background Art

[0002] The property display platform is a platform for users to learn about property information. The authenticity of the properties on the property display platform (such as whether the property is actually for sale or for rent) is very important to the user's house-hunting experience.

[0003] Currently, intelligent models are primarily used to identify abnormal listing statuses. However, due to the small sample size of abnormal listings, it is difficult for intelligent models to accurately learn the characteristics of these listings, resulting in poor prediction accuracy. Summary of the Invention

[0004] In order to solve the above technical problems, the present disclosure provides a method, device, storage medium and program product for identifying abnormal housing resources.

[0005] In a first aspect, an embodiment of the present disclosure provides a method for identifying abnormal housing listings, the method comprising:

[0006] Obtain target property information of a property to be identified; input the target property information into a first recognition model and a second recognition model respectively, and output a first recognition result and a second recognition result of the property to be identified; wherein the first recognition model is trained based on a sample of first property information that has been removed from the shelves due to a transaction; and the second recognition model is trained based on a sample of second property information that has been removed from the shelves due to a non-transaction; perform fusion processing based on the first recognition result and the second recognition result to generate a target recognition result of the property to be identified; wherein the target recognition result is used to characterize the possibility that the property to be identified is an abnormal property.

[0007] In some optional implementations, after obtaining target property information of a property to be identified, the method further includes:

[0008] Determining a comprehensive similarity between the target property information and each target property information sample; wherein the target property information sample includes the first property information sample and / or the second property information sample;

[0009] Based on the first recognition result and the second recognition result, a target recognition result of the property to be recognized is generated, including:

[0010] Based on the first recognition result, the second recognition result and the comprehensive similarity, a fusion process is performed to generate a target recognition result of the house to be identified.

[0011] In some optional implementations, a fusion process is performed based on the first recognition result, the second recognition result, and the comprehensive similarity to generate a target recognition result for the property to be recognized, including:

[0012] The first recognition result, the second recognition result, the comprehensive similarity and at least one additional property feature are input into a preset fusion model to generate a target recognition result for the property to be identified; wherein the additional property feature is property information that has an independent correlation with whether the property is abnormal.

[0013] In some optional embodiments, the additional listing features include the listing city identifier, the ratio of the number of listings removed by the listing manager within a first preset time period to the total number of listings, or the number of sessions corresponding to the listing within a second preset time period; and / or, the preset fusion model includes a tree model or a linear weighted model.

[0014] In some optional implementations, after generating a target identification result for a property to be identified, the method further includes:

[0015] The historical conversation data corresponding to the property to be identified is input into a preset language model to generate a model output result; the model output result includes a third recognition result of whether the property to be identified is an abnormal property and / or an information extraction dimension of the property to be identified; the target recognition result is adjusted based on the model output result to update the target recognition result.

[0016] In some optional embodiments, if the model output result includes the third recognition result and the information extraction dimension, adjusting the target recognition result based on the model output result to update the target recognition result includes:

[0017] If the third recognition result is that the property to be identified is an abnormal property and the number of information extraction dimensions does not exceed the preset number, the target recognition result is increased twice in a row at a first preset ratio to update the target recognition result; if the third recognition result is that the property to be identified is an abnormal property and the number of information extraction dimensions exceeds the preset number, the target recognition result is increased at the first preset ratio to update the target recognition result; if the third recognition result is that the property to be identified is not an abnormal property and the number of information extraction dimensions does not exceed the preset number, the target recognition result is increased at the first preset ratio to update the target recognition result; if the third recognition result is that the property to be identified is not an abnormal property and the number of information extraction dimensions exceeds the preset number, the target recognition result is reduced at the first preset ratio to update the target recognition result.

[0018] In some optional implementations, after generating a target identification result for a property to be identified, the method further includes:

[0019] If the current status of the property to be identified is on-shelf, and the property to be identified is off-shelf on the third platform, the target identification result is increased by a second preset ratio to update the target identification result.

[0020] In some optional implementations, the first recognition model and the second recognition model are both preset binary classification models; the preset binary classification model is a distributed gradient model based on a decision tree.

[0021] In a second aspect, an embodiment of the present disclosure provides a computer-readable storage medium, which stores a program or instruction, and the program or instruction enables a computer to execute the abnormal housing identification method of any embodiment of the present disclosure.

[0022] In a third aspect, an embodiment of the present disclosure provides a computer program product, which is used to implement the abnormal housing identification method of any embodiment of the present disclosure.

[0023] The abnormal property identification method provided by the embodiment of the present disclosure obtains target property information of the property to be identified; inputs the target property information into a first identification model and a second identification model respectively, and outputs a first identification result and a second identification result of the property to be identified; performs fusion processing based on the first identification result and the second identification result to generate a target identification result of the property to be identified; the above technical solution realizes the analysis of the target property information in two dimensions of transaction removal and non-transaction removal by introducing two identification models, so that the first identification result and the second identification result have higher accuracy in the corresponding dimensions; thereafter, by fusing the first identification result and the second identification result, the final target identification result combines the advantages of the two identification models, thereby making the final identification result more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0025] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0026] Figure 1 A flowchart of a method for identifying abnormal housing listings provided in an embodiment of the present disclosure;

[0027] Figure 2 A flowchart of another method for identifying abnormal housing listings provided in an embodiment of the present disclosure;

[0028] Figure 3 A flowchart of a method for updating target recognition results provided by an embodiment of the present disclosure;

[0029] Figure 4 A schematic diagram of the structure of an abnormal housing listing identification device provided by an embodiment of the present disclosure;

[0030] Figure 5 A schematic structural diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0031] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described in detail below. It should be noted that the embodiments of the present disclosure and the features therein can be combined with each other in the absence of conflict.

[0032] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0033] Listing platforms are platforms where buyers and renters can access information about properties. Abnormal listings on these platforms can severely impact users' house-hunting experience and their trust in the company.

[0034] Abnormal listings can be understood as listings that do not actually exist. There are two main types of non-existent listings. First, the house does not actually exist in a specific location and is a virtual house. Second, the listing status of the house on the listing platform does not match its actual status. For example, a house may appear to be sold or rented in real life, but still be listed as sold or rented on the listing platform.

[0035] To identify unusual listings on property listing platforms, related technologies primarily rely on intelligent models. The accuracy of these models' identification results is often related to the size of the training sample.

[0036] Analysis revealed that 80% of the anomalous listings on the property display platform fall into the second category, despite accounting for only 2% of all available properties. This phenomenon results in a small sample size of anomalous listings, making it difficult for the intelligent model to capture patterns in these listings, resulting in poor recognition accuracy.

[0037] Based on this, the abnormal housing source identification method provided by the embodiment of the present disclosure introduces two recognition models to realize the analysis of the target housing source information in two dimensions: transaction removal and non-transaction removal, so that the first recognition result and the second recognition result have higher accuracy in the corresponding dimensions; thereafter, by fusing the first recognition result and the second recognition result, the final target recognition result combines the advantages of the two recognition models, thereby making the final recognition result more accurate.

[0038] The abnormal property identification method provided in the embodiments of the present disclosure is used to identify properties that should be removed from a property display platform but are still on sale or for rent. The method can be performed by an abnormal property identification device, which can be implemented by software and / or hardware. The device can be integrated into an electronic device with data processing capabilities, such as a handheld computer, tablet computer, laptop computer, desktop computer, or backend server.

[0039] Figure 1 This is a flow chart of a method for identifying abnormal housing listings provided by an embodiment of the present disclosure. Figure 1 , the abnormal housing source identification method specifically includes:

[0040] S110: Obtain target property information of the property to be identified.

[0041] Among them, the target housing information is the housing information corresponding to the housing to be identified. The housing information at least includes the building area, unit price of the housing, average unit price in the same community, average unit price in the same business district, city unit price ranking, unit price of the housing / average unit price in the same community, unit price of the housing / average unit price in the same business district, listing time, listing time corresponding to the historical housing, broker activity (such as the broker has been active in the past XX days), search channels for exposure of the housing details page in the past XX days, recommended channels for exposure of the housing card in the past XX days, exposure of the housing details page in the past XX days, and other information.

[0042] Specifically, the abnormal property identification device obtains target property information of the property to be identified from the data storage space of the property display platform. The data storage space may be a database of the property display platform or a cloud server corresponding to the property display platform. This embodiment does not limit the data storage space.

[0043] S120: Input the target property information into the first recognition model and the second recognition model respectively, and output the first recognition result and the second recognition result of the property to be recognized.

[0044] Among them, the first recognition model is trained based on the first housing information sample that is removed from the shelves due to transaction; the second recognition model is trained based on the second housing information sample that is removed from the shelves due to non-transaction. The first recognition model and the second recognition model can be understood as different results obtained by training the same recognition model with different housing information samples. The first recognition model and the second recognition model can be preset probability prediction models, which are used to predict the probability that the housing to be identified is an abnormal housing based on the target housing information. The preset probability prediction model can be a deep learning model, a neural network model, etc. The first recognition model and the second recognition model can also be preset binary classification models, which are used to determine whether the housing to be identified is an abnormal housing based on the target housing information. The preset binary classification model can be an intelligent model that is easy to build and has strong interpretability, including but not limited to a distributed gradient model based on a decision tree, an adaptive base class boosting model, and a gradient boosting decision tree model.

[0045] The first recognition model and the second recognition model are trained at least before S120. Specifically, before S120, first, a housing information sample and the reason for the abnormal housing to which the housing information sample belongs are obtained from the housing information of all abnormal housing listings that have been removed from the shelves. The housing information sample is based on the time when the abnormal housing listing is removed from the shelves, and the housing information at the corresponding time is traced back for a preset period of time. Then, according to the reason for removal, the housing information sample is classified to obtain a first housing information sample corresponding to the transaction removal and a second housing information sample corresponding to the non-transaction removal. Finally, with the transaction removal as the label, the first housing information sample is used to perform supervised training on the pre-constructed recognition model to obtain the first recognition model; similarly, with the non-transaction removal as the label, the second housing information sample is used to perform supervised training on the pre-constructed recognition model to obtain the second recognition model. It can be understood that transaction removal means that the housing listing is removed from the shelves due to sale / rental, and non-transaction removal means that the housing listing is removed from the shelves due to reasons other than sale / rental. Non-transaction removal includes but is not limited to the platform administrator confirming that the property is a virtual / fictitious high-quality property used to attract traffic, and the property owner clearly stating that he has no intention to sell / rent the property for the time being.

[0046] In one embodiment, the preset time period is 7 days. The way to obtain the housing information sample is, for example, abnormal housing A was removed from the housing display platform on January 8 due to being sold. Therefore, the housing information corresponding to abnormal housing A on January 1 can be used as a sample, that is, the first housing information sample. Abnormal housing B was removed from the housing display platform on January 15 due to the owner's personal subjective reasons. Therefore, the housing information corresponding to abnormal housing B on January 8 can be used as a sample, that is, the second housing information sample. January 8 and January 15 are both historical moments. In addition,

[0047] Specifically, the target property information is input into the first recognition model, and the first recognition result is output. The target property information is input into the second recognition model, and the second recognition result is output. The first recognition result and the second recognition result are both results that characterize whether the property to be identified is an abnormal property, and the first recognition result and the second recognition result have the same form of expression. For example, the first recognition result and the second recognition result can be probability values ​​indicating that the property to be identified is an abnormal property, or they can be recognition results of the property to be identified, such as yes / no, where yes indicates that the property to be identified is an abnormal property, and otherwise indicates that the property to be identified is a normal property. The form of the first recognition result / second recognition result output by the model corresponds to the type of the first recognition model / second recognition model, and is not specifically limited here.

[0048] It should be emphasized that the reason why this embodiment uses two recognition models to identify the listings to be identified is that, on the one hand, the feature distribution of listings that have been removed from the shelves due to transactions is consistent with that of normal listings before they are removed from the shelves, while the feature distribution of listings that have not been removed from the shelves due to transactions differs significantly from that of listings that are still on sale. Therefore, using only one model to identify the listings to be identified will lead to inaccurate recognition results, especially for abnormal listings that have been sold but not removed from the shelves. Therefore, this embodiment introduces a first recognition model and a second recognition model, so that the first recognition model can identify listings that are abnormal due to transactions, and the second recognition model can identify listings that are abnormal due to other reasons, thereby improving the accuracy of the recognition results. On the other hand, using only one recognition model to identify the listings to be identified will increase the complexity of the recognition model, and the recognition model may learn noise and details in the input information, resulting in reduced accuracy of the recognition results. However, this embodiment uses two recognition models to identify the listings to be identified, which can conduct targeted learning of certain aspects of the input information, thereby improving the accuracy of the recognition results.

[0049] It should be noted that the first recognition model and the second recognition model in the embodiment of the present disclosure may be run in parallel or executed serially, and the embodiment of the present disclosure is not limited to this.

[0050] S130: Perform fusion processing based on the first recognition result and the second recognition result to generate a target recognition result of the house to be recognized.

[0051] Among them, the target recognition result is used to characterize the possibility that the house to be identified is an abnormal house.

[0052] Specifically, the first recognition result and the second recognition result are fused to combine the advantages of the two models in the fusion process to obtain a comprehensive target recognition result with reference significance, so that the output target recognition result of the house to be identified is more accurate. The fusion process can be achieved at least by a pre-trained tree model, linear weighted fusion, etc. For example, the first recognition result and the second recognition result can be input into a pre-trained tree model, and the pre-trained tree model is used to fuse the first recognition result and the second recognition result to obtain the target recognition result. For another example, weights are set for the first recognition result and the second recognition result according to their importance, and then a weighted sum is performed to obtain the target recognition result. In the linear weighted fusion process, the weight of the first recognition result and the weight of the second recognition result can be the same or different. In addition, it should be noted that the target recognition result is expressed in the same way as the first recognition result and the second recognition result, which will not be repeated here.

[0053] The above-mentioned technical solution of the embodiment of the present disclosure obtains the target property information of the property to be identified; inputs the target property information into the first recognition model and the second recognition model respectively, and outputs the first recognition result and the second recognition result of the property to be identified; performs fusion processing based on the first recognition result and the second recognition result to generate the target recognition result of the property to be identified; the above-mentioned technical solution realizes the analysis of the target property information in the two dimensions of transaction removal and non-transaction removal by introducing two recognition models, so that the first recognition result and the second recognition result have higher accuracy in the corresponding dimensions; thereafter, by fusing the first recognition result and the second recognition result, the final target recognition result combines the advantages of the two recognition models, thereby making the final recognition result more accurate.

[0054] Figure 2 This is a flowchart of another abnormal listing identification method provided by the embodiment of the present disclosure. It adds a step of calculating the comprehensive similarity. On this basis, the fusion process can be further optimized. The explanations of the terms that are the same or corresponding to the above embodiments are not repeated here. Figure 2 , the abnormal housing source identification method includes:

[0055] S210, obtain the target property information to be identified. Figure 1 Step S110 of the illustrated embodiment will not be described in detail here.

[0056] S220: Determine the comprehensive similarity between the target property information and each target property information sample.

[0057] The target property information sample includes a first property information sample and / or a second property information sample.

[0058] For example, in one embodiment, the target property information and target property information samples may be input into a pre-built similarity calculation model to obtain a comprehensive similarity between the target property information and each target property information sample.

[0059] For example, in another embodiment, the similarity between the target property information and each target property information sample may be directly calculated, and then the average of all similarities may be used as the comprehensive similarity.

[0060] When the target property information sample includes the first property information sample, the similarity between the target property information sample and each of the first property information samples (referred to as the first similarity) is calculated, and then a first average similarity is calculated based on all the first similarities, and the first average similarity is determined as the comprehensive similarity.

[0061] When the target property information sample includes a second property information sample, the similarity between the target property information sample and each second property information sample (referred to as a second similarity) is calculated, and then a second average similarity is determined based on all the second similarities, and the second average similarity is determined as the comprehensive similarity.

[0062] When the target property information sample includes the first property information sample and the second property information sample, the average of the first average similarity and the second average similarity obtained in the two aforementioned methods is determined as the comprehensive similarity.

[0063] For example, in another embodiment, each target property information sample can be treated as a cluster. Based on the property features contained in the target property information sample, the target property information sample is mapped into a feature vector, recorded as the first feature vector. Similarly, the target property information of the property to be identified is mapped into a feature vector, recorded as the second feature vector. Then, the distance between the second feature vector and each first feature vector is calculated. Finally, the average of the distances between the second feature vector and each first feature vector is used as the comprehensive similarity. In this embodiment, the distance between the second feature vector and the first feature vector includes, but is not limited to, Euclidean distance, Manhattan distance, cosine similarity, etc.

[0064] In this embodiment, the features in the mapped feature vector include at least the following dimensions, such as business opportunities in the past 90 days, search channels for property card exposure in the past 3 days, search channels for property card exposure in the past 7 days, house orientation - north, historical property verification times, number of users who clicked on property cards in the past 15 days, house orientation - east, number of users who clicked on property cards in the past 15 days, number of kitchens, search channels for property details page exposure in the past 3 days, business opportunities in the past 15 days, total price ranking of real estate projects, house orientation - west, number of living rooms, business opportunities in the past 60 days, recommended channels for property details page exposure in the past 15 days, house orientation - south, number of property card users in the past 3 days, property details page exposure in the past 3 days, recommended channels for property details page exposure in the past 7 days, and number of bathrooms, etc.

[0065] S230: Input the target property information into the first recognition model and the second recognition model respectively, and output the first recognition result and the second recognition result of the property to be recognized. Figure 1 Step S120 of the illustrated embodiment will not be described in detail here.

[0066] S240: Perform fusion processing based on the first recognition result, the second recognition result and the comprehensive similarity to generate a target recognition result of the house to be recognized.

[0067] Specifically, a fusion process is performed based on the first recognition result, the second recognition result, and the comprehensive similarity to generate a target recognition result for the unidentified property. The fusion process is similar to that in S130 and will not be repeated here. Since the comprehensive similarity represents the similarity in features between the unidentified property and the abnormal property sample, it can, to a certain extent, represent the category to which the unidentified property belongs. In this embodiment, the comprehensive similarity is integrated into the target recognition result to improve the accuracy of the target recognition result.

[0068] In another optional embodiment, S240 includes inputting the first recognition result, the second recognition result, the comprehensive similarity and at least one additional property feature into a preset fusion model to generate a target recognition result of the property to be identified.

[0069] Additional listing features are information that is independently relevant to determining whether a listing is unusual. Alternatively, additional listing features can be used to determine whether a listing is unusual without requiring cross-reference with other features.

[0070] Exemplarily, additional listing features include the listing city identifier, the ratio of the number of listings removed by the listing manager within the first preset time period to the total number of listings, or the number of sessions corresponding to the listing within the second preset time period. For example, if the ratio of the number of listings removed by the listing manager within the first preset time period to the total number of listings is too high, it means that the listings managed by the listing manager are removed frequently, which, to a certain extent, indicates that there is a high possibility that the listings managed by the listing manager are abnormal. For example, if the number of sessions corresponding to the listing within the second preset time period is large, it means that the maintenance frequency of the listing is high, which, to a certain extent, indicates that the listing is normal. The first preset time period and the second preset time period can be determined by those skilled in the art based on actual conditions.

[0071] Exemplarily, the preset fusion model includes a tree model or a linear weighted model.

[0072] Since the additional property features can be important features for determining whether the property to be identified is abnormal, this embodiment integrates the additional property features with the comprehensive similarity, the first recognition result, and the second recognition result. As a result, the obtained target recognition result not only combines the advantages of the first recognition model and the second recognition model, but also combines the feature similarity between the target property information and the abnormal property sample and the additional property features, thereby making the target recognition result more accurate.

[0073] In an optional embodiment, after S130 or S240, an update operation of the target recognition result is added, such as Figure 3 As shown, the following steps are included:

[0074] S310: Input the historical conversation data corresponding to the property to be identified into a preset language model to generate a model output result.

[0075] The model output includes a third identification result of whether the listing to be identified is an unusual listing and / or information extraction dimensions of the listing to be identified. The model output may include a third identification result of whether the listing to be identified is an unusual listing, information extraction dimensions of the listing to be identified, or both. It should be noted that the output of the preset language model corresponds to the model's settings. Information extraction dimensions of the listing to be identified include, but are not limited to, floor level, price, area, tax, and viewing intentions.

[0076] S320: Adjust the target recognition result based on the model output result to update the target recognition result.

[0077] Specifically, when the model output result includes the third recognition result and the information extraction dimension, S320 includes:

[0078] S3201. If the third recognition result is that the property to be identified is an abnormal property and the number of information extraction dimensions does not exceed the preset number, the target recognition result is increased twice in succession at a first preset ratio to update the target recognition result.

[0079] Specifically, when the number of information extraction dimensions does not exceed a preset number, it indicates that the conversation expanded on the page of the property to be identified involves relatively few information dimensions. This could mean that the customer is interested in the property to be identified, but upon learning from the property manager that the property is currently unavailable for sale / rental, ends the conversation. Therefore, when the number of information extraction dimensions does not exceed the preset number, it can be largely assumed that the property to be identified is an anomaly. Based on this, if the third identification result indicates that the property to be identified is an anomaly, then there is a high probability that the property to be identified is an anomaly. In this case, the target identification result is updated by increasing the target identification result twice by a first preset ratio. The first preset ratio is set by those skilled in the art, for example, 10%. If the target identification result obtained based on S130 or S240 is 0.5, then the target identification result is increased by 10% twice to obtain 0.7. If the target identification result obtained based on S130 or S240 is 0.9, since the upper limit of the target identification result is 1, the target identification result remains 1 after increasing by 10% twice.

[0080] S3202. If the third recognition result is that the property to be identified is an abnormal property and the number of information extraction dimensions exceeds a preset number, the target recognition result is increased by a first preset ratio to update the target recognition result.

[0081] Specifically, if the number of information extraction dimensions exceeds the preset number, it means that there are many information dimensions involved in the conversation launched in the interface of the house to be identified. It may be that the customer is interested in the house to be identified and has a conversation with the corresponding house manager based on the detailed information of the house to be identified. It may also be that after the house manager indicates that the house to be identified has been sold / rented, the customer provides a detailed description based on his own needs, or the house manager provides a description based on the detailed information of other houses. Therefore, when the number of information extraction dimensions exceeds the preset number, the house to be identified may be a normal house or an abnormal house. On this basis, if the third recognition result is that the house to be identified is an abnormal house, then it means that the possibility that the house to be identified is an abnormal house is relatively high. At this time, the target recognition result can be increased by the first preset ratio to achieve the update of the target recognition result.

[0082] S3203. If the third recognition result is that the property to be identified is not an abnormal property and the number of information extraction dimensions does not exceed a preset number, the target recognition result is increased by a first preset ratio to update the target recognition result.

[0083] Specifically, based on the analysis in S3201 that the number of information extraction dimensions does not exceed the preset number, it can be determined that when the number of information extraction dimensions does not exceed the preset number, the property to be identified is likely to be an anomaly. However, if the third identification result indicates that the property to be identified is not an anomaly, and comprehensive analysis indicates a high probability that the property to be identified is an anomaly, the target identification result is increased by the first preset ratio, thereby updating the target identification result.

[0084] S3204. If the third recognition result is that the property to be identified is not an abnormal property, and the number of information extraction dimensions exceeds a preset number, the target recognition result is reduced by a first preset ratio to update the target recognition result.

[0085] Specifically, based on the analysis in S3202 that the number of information extraction dimensions does not exceed the preset number, it can be seen that when the number of information extraction dimensions exceeds the preset number, it indicates that the property to be identified may be a normal property or an abnormal property. Based on this, if the third identification result indicates that the property to be identified is not an abnormal property, then it is likely that the property to be identified is a normal property. In this case, the target identification result is reduced by the first preset ratio to achieve an update of the target identification result.

[0086] In the above technical solution of the disclosed embodiments, when the output of the preset language model includes a third recognition result and an information extraction dimension, the likelihood of the identified property being an anomalous property is determined based on analysis of the third recognition result and the information extraction dimension, and the target recognition result is updated based on this likelihood. Because the information dimension in the conversation can reflect the authenticity of the identified property to a certain extent, updating the target recognition result using the output of the preset language can make the updated target recognition result more accurate.

[0087] In an optional embodiment, after S130 or S240, an update operation of the target recognition result is added, and the method includes: if the current status of the property to be identified is on the shelves, and the property to be identified is off the shelves in the third platform, then the target recognition result is increased by a second preset ratio to update the target recognition result.

[0088] Specifically, if the property to be identified is listed on the current listing platform and is removed from another listing platform (i.e., a third listing platform) besides the current listing platform, it is possible that the property to be identified has been removed from the third listing platform because it has been sold / rented or the owner has no intention of selling / renting it, but is still listed on the current listing platform. Therefore, it can be considered that the property to be identified is an abnormal property. In this case, the target identification result is increased by a second preset ratio to achieve an update of the target identification result.

[0089] Figure 4 This is a schematic diagram of the structure of an abnormal housing source identification device provided by an embodiment of the present disclosure. Figure 4 As shown, the abnormal housing source identification device includes:

[0090] The acquisition module 410 is used to obtain target property information of the property to be identified.

[0091] The output module 420 is used to input the target property information into the first recognition model and the second recognition model respectively, and output the first recognition result and the second recognition result of the property to be identified; wherein the first recognition model is trained based on the first property information sample that is removed from the shelf due to a transaction; the second recognition model is trained based on the second property information sample that is removed from the shelf due to a non-transaction.

[0092] The first generation module 430 is used to perform fusion processing based on the first recognition result and the second recognition result to generate a target recognition result of the house to be identified; wherein the target recognition result is used to represent the possibility that the house to be identified is an abnormal house.

[0093] In some optional embodiments, the device further comprises:

[0094] a determination module configured to, after obtaining target property information of a property to be identified, determine a comprehensive similarity between the target property information and each target property information sample; wherein the target property information sample includes the first property information sample and / or the second property information sample;

[0095] The first generation module 430 includes:

[0096] The first generation submodule is used to perform fusion processing based on the first recognition result, the second recognition result and the comprehensive similarity to generate a target recognition result of the house to be recognized.

[0097] In some optional implementations, the first generation module 430 includes:

[0098] The second generation submodule is used to input the first recognition result, the second recognition result, the comprehensive similarity and at least one additional property feature into a preset fusion model to generate a target recognition result for the property to be identified; wherein the additional property feature is property information that has an independent correlation with whether the property is abnormal.

[0099] In some optional embodiments, the additional listing features include the listing city identifier, the ratio of the number of listings removed by the listing manager within a first preset time period to the total number of listings, or the number of sessions corresponding to the listing within a second preset time period; and / or, the preset fusion model includes a tree model or a linear weighted model.

[0100] In some optional embodiments, the device further comprises:

[0101] The second generation module is used to input the historical conversation data corresponding to the property to be identified into a preset language model after generating the target recognition result of the property to be identified, and generate a model output result; the model output result includes a third recognition result of whether the property to be identified is an abnormal property and / or the information extraction dimension of the property to be identified; the first update module is used to adjust the target recognition result based on the model output result to update the target recognition result.

[0102] In some optional implementations, if the model output result includes the third recognition result and the information extraction dimension, the first updating module includes:

[0103] The first updating submodule is used to increase the target recognition result twice in a row by a first preset ratio if the third recognition result is that the property to be identified is an abnormal property and the number of information extraction dimensions does not exceed a preset number to update the target recognition result; the second updating submodule is used to increase the target recognition result by a first preset ratio if the third recognition result is that the property to be identified is an abnormal property and the number of information extraction dimensions exceeds a preset number to update the target recognition result; the third updating submodule is used to increase the target recognition result by a first preset ratio if the third recognition result is that the property to be identified is not an abnormal property and the number of information extraction dimensions does not exceed a preset number to update the target recognition result; the fourth updating submodule is used to reduce the target recognition result by a first preset ratio if the third recognition result is that the property to be identified is not an abnormal property and the number of information extraction dimensions exceeds a preset number to update the target recognition result.

[0104] In some optional embodiments, the device further comprises:

[0105] The second updating module is used to increase the target recognition result by a second preset ratio after generating the target recognition result of the house to be identified, if the current status of the house to be identified is on the shelf and the house to be identified is off the shelf on the third platform, so as to update the target recognition result.

[0106] In some optional implementations, the first recognition model and the second recognition model are both preset binary classification models; the preset binary classification model is a distributed gradient model based on a decision tree.

[0107] The abnormal housing source identification device provided by the embodiments of the present disclosure can execute the abnormal housing source identification method provided by any embodiment of the present disclosure, and has the corresponding functional modules and beneficial effects of the execution method.

[0108] It is worth noting that in the embodiment of the above-mentioned abnormal housing identification device, the various modules and sub-modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the various functional modules / sub-modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this disclosure.

[0109] Figure 5 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present disclosure. Figure 5 As shown, the electronic device 500 includes one or more processors 501 and a memory 502 .

[0110] The processor 501 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 500 to perform desired functions.

[0111] The memory 502 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on a computer-readable storage medium, and the processor 501 may run the program instructions to implement the abnormal housing source identification method of the embodiment of the present disclosure described above and / or other desired functions. Various contents such as target housing source information, first identification results, second identification results, target identification results, etc. may also be stored in the computer-readable storage medium.

[0112] In one example, the electronic device 500 may further include an input device 503 and an output device 504, which are interconnected via a bus system and / or other forms of connection mechanisms (not shown). The input device 503 may include, for example, a keyboard, a mouse, etc. The output device 504 may output various information to the outside, including target recognition results, information extraction dimensions, etc. The output device 504 may include, for example, a display, a speaker, a printer, a communication network and its connected remote output devices, etc.

[0113] Of course, to simplify, Figure 5 Only some of the components related to the present disclosure in the electronic device 500 are shown, and components such as buses, input / output interfaces, etc. are omitted. In addition, the electronic device 500 may further include any other appropriate components according to specific application scenarios.

[0114] In addition to the above-mentioned methods and devices, the abnormal housing source identification method in any embodiment of the present disclosure can also be implemented as a computer software program. For example, the embodiment of the present disclosure also includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through a communication device, or installed from a memory. When the computer program is executed by the processor, the processor causes the processor to execute the abnormal housing source identification method provided by any embodiment of the present disclosure.

[0115] The computer program product may be written in any combination of one or more programming languages ​​to implement the operations of the disclosed embodiments, including object-oriented programming languages ​​such as Java, C++, and conventional procedural programming languages ​​such as C or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's computing device, as a stand-alone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0116] In addition, an embodiment of the present disclosure also provides a computer-readable storage medium on which a computer program or instruction is stored. When the computer program or instruction is executed by a processor, the processor executes the abnormal housing identification method provided by the embodiment of the present disclosure.

[0117] Computer readable storage media can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can, for example, include but is not limited to a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0118] It should be noted that the terms used in this disclosure are only for describing specific embodiments and are not intended to limit the scope of this application. As shown in the present disclosure and claims, unless the context clearly indicates an exception, words such as "one", "a", "a kind of" and / or "the" do not specifically refer to the singular and may also include the plural. The term "and / or" includes any and all combinations of one or more related listed items. Relational terms such as "first" and "second" are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. The term "comprises", "includes" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method or device. In the absence of further restrictions, the elements defined by the sentence "comprising a..." do not exclude the presence of other identical elements in the process, method or device comprising the elements.

[0119] The foregoing are merely specific embodiments of the present disclosure, intended to enable those skilled in the art to understand and implement the present disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure is not to be limited to these embodiments, but is to be construed in the broadest manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying abnormal housing listings, characterized in that: include: Obtain target property information for the property to be identified; Inputting the target property information into a first recognition model and a second recognition model, respectively, and outputting a first recognition result and a second recognition result of the property to be identified; wherein the first recognition model is trained based on a sample of first property information that has been removed from the market due to a transaction; and the second recognition model is trained based on a sample of second property information that has not been removed from the market due to a transaction; A fusion process is performed based on the first recognition result and the second recognition result to generate a target recognition result of the house to be identified; wherein the target recognition result is used to characterize the possibility that the house to be identified is an abnormal house.

2. The method according to claim 1, characterized in that After obtaining target property information of a property to be identified, the method further includes: Determining a comprehensive similarity between the target property information and each target property information sample; wherein the target property information sample includes the first property information sample and / or the second property information sample; The generating a target recognition result of the to-be-recognized house by fusing the first recognition result and the second recognition result includes: A fusion process is performed based on the first recognition result, the second recognition result and the comprehensive similarity to generate a target recognition result of the house to be identified.

3. The method according to claim 2, characterized in that The generating a target recognition result of the to-be-recognized house by fusing the first recognition result, the second recognition result, and the comprehensive similarity includes: The first recognition result, the second recognition result, the comprehensive similarity and at least one additional property feature are input into a preset fusion model to generate a target recognition result of the property to be identified; wherein the additional property feature is property information that has an independent correlation with whether the property is abnormal.

4. The method according to claim 3, characterized in that The additional listing features include the listing city identifier, the ratio of the number of listings removed by the listing manager within the first preset time period to the total number of listings, or the number of sessions corresponding to the listing within the second preset time period; And / or, the preset fusion model includes a tree model or a linear weighted model.

5. The method according to any one of claims 1 to 4, characterized in that After generating the target recognition result of the to-be-recognized house, the method further includes: Inputting historical conversation data corresponding to the listing to be identified into a preset language model to generate a model output result; the model output result includes a third identification result of whether the listing to be identified is an abnormal listing and / or information extraction dimensions of the listing to be identified; The target recognition result is adjusted based on the model output result to update the target recognition result.

6. The method according to claim 5, characterized in that If the model output result includes the third recognition result and the information extraction dimension, adjusting the target recognition result based on the model output result to update the target recognition result includes: If the third recognition result is that the to-be-recognized property is an abnormal property, and the number of the information extraction dimensions does not exceed a preset number, increasing the target recognition result twice in succession by a first preset ratio to update the target recognition result; If the third recognition result is that the to-be-recognized property is an abnormal property, and the number of the information extraction dimensions exceeds the preset number, increasing the target recognition result by the first preset ratio to update the target recognition result; If the third recognition result is that the to-be-recognized property is not an abnormal property, and the number of the information extraction dimensions does not exceed the preset number, increasing the target recognition result by the first preset ratio to update the target recognition result; If the third recognition result is that the property to be identified is not an abnormal property, and the number of information extraction dimensions exceeds the preset number, the target recognition result is reduced by the first preset ratio to update the target recognition result.

7. The method according to any one of claims 1 to 4, characterized in that After generating the target recognition result of the to-be-recognized house, the method further includes: If the current status of the property to be identified is a listed status, and the property to be identified is in a delisted status on the third platform, the target identification result is increased by a second preset ratio to update the target identification result.

8. The method according to claim 1, characterized in that The first recognition model and the second recognition model are both preset binary classification models; the preset binary classification model is a distributed gradient model based on a decision tree.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program or instruction, and the program or instruction enables a computer to execute the abnormal housing identification method according to any one of claims 1 to 8.

10. A computer program product, characterized in that The computer program product is used to implement the abnormal housing listing identification method according to any one of claims 1 to 8.

Images