Model training method and device for identifying abnormal housing use
Through the federated learning model training method, the data is summarized and irreversible encoding is performed using multiple data sources, which solves the problem of difficult identification of abnormal use of residential buildings, and realizes efficient identification of abnormal use of houses and improves public safety.
Patent Information
- Application Number
- CN202011584090.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-12-28
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2040-12-28
AI Technical Summary
During the process of urbanization, abnormal use of residential buildings (such as illegal renovation and stealing of water and electricity) is difficult to be discovered by public security fire inspections or random inspections, resulting in public safety hazards, and regular inspections are high and incomplete.
Using the federated learning model training method, data is summarized through multiple data sources, irreversible encoding of house address information, generate house address identification, send it to the cloud server to determine the training sample, and conduct model training locally to ensure data security and model accuracy sex.
It realizes accurate and efficient identification of abnormal use of houses, reduces the risk of data leakage, and improves model training effect and public safety.
Smart Images

Figure CN113724117B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to a model training method and device for identifying abnormal use of a house. Background Art
[0002] With the progress of urbanization, there are more and more migrant workers in cities, and migrant workers will have more demand for renting houses, which leads to abnormal use of residential buildings, mainly including illegal renovation of group houses, theft of water and electricity, etc. Abnormal use is not easy to be discovered when the public security and fire departments cannot conduct regular inspections or random checks, which will bring great hidden dangers to public safety. Regular inspections or random checks require a large cost and there is also the problem of incomplete inspections. Therefore, how to identify abnormal use of houses has become a technical problem that needs to be solved urgently. Summary of the invention
[0003] The present application aims to solve one of the technical problems in the related art at least to some extent.
[0004] To this end, the first purpose of this application is to propose a model training method for identifying abnormal use of a house, by aggregating data from multiple data sources to obtain training samples for model training of a federated learning model, so that the trained model can accurately and efficiently identify abnormal use of a house.
[0005] The second objective of this application is to propose another model training method for identifying abnormal usage of a house.
[0006] The third objective of this application is to provide a method for identifying abnormal use of a house.
[0007] The fourth objective of the present application is to propose a model training device for identifying abnormal use of a house.
[0008] The fifth objective of the present application is to propose another model training device for identifying abnormal use of a house.
[0009] The sixth objective of the present application is to provide a device for identifying abnormal use of a house.
[0010] A seventh objective of the present application is to provide a computer device.
[0011] An eighth object of the present application is to provide a non-transitory computer-readable storage medium.
[0012] A ninth objective of the present application is to provide a computer program product.
[0013] To achieve the above-mentioned purpose, the first embodiment of the present application proposes a model training method for identifying abnormal use of a house, which is applied to at least two devices in a federated learning system, wherein the federated learning system also includes a cloud server, and the method executed by each of the devices includes the following steps:
[0014] Collecting a plurality of candidate records; wherein each of the candidate records includes house address information and associated information associated with the house address information;
[0015] Irreversibly encode the house address information of each candidate record to obtain a house address identifier corresponding to each candidate record;
[0016] Sending the house address identifier corresponding to each candidate record to the cloud server, so that the cloud server determines the training sample to which each candidate record belongs according to the house address identifier;
[0017] In response to the instruction information sent by the cloud server, determining the training sample to which each of the candidate records collected locally belongs;
[0018] The model nodes running locally in the federated learning model are trained according to the training samples to which each of the candidate records belongs that are collected locally.
[0019] To achieve the above-mentioned purpose, the second embodiment of the present application proposes a model training method for identifying abnormal use of a house, which is applied to a cloud server in a federated learning system, wherein the federated learning system also includes at least two devices; the method includes:
[0020] Acquire the house address identifier corresponding to the collected candidate records from at least two of the devices;
[0021] Determine, according to the house address identifier, the training sample to which each candidate record belongs;
[0022] Send corresponding indication information to each of the devices, where the indication information is used to indicate the training samples to which the candidate records collected by the corresponding device belong, so that the corresponding device uses the collected candidate records as training samples to train the model nodes running locally in the federated learning model.
[0023] To achieve the above-mentioned purpose, the third aspect of the present application proposes a method for identifying abnormal use of a house, comprising:
[0024] Acquire a target record to be identified; wherein the target record includes house address information and associated information associated with the house address information;
[0025] The associated information in the target record is input into a trained model node of a federated learning model to determine the abnormal usage of the house corresponding to the house address information; wherein the model node is trained using the model training method described in the first aspect, or is trained using the model training method described in the second aspect.
[0026] To achieve the above-mentioned purpose, the fourth aspect of the present application proposes a model training device for identifying abnormal use of a house, the device comprising:
[0027] A first acquisition module is used to collect multiple candidate records; wherein each candidate record includes house address information and associated information associated with the house address information;
[0028] An encoding module, used for irreversibly encoding the house address information of each candidate record to obtain a house address identifier corresponding to each candidate record;
[0029] A sending module, used for sending the house address identifier corresponding to each candidate record to the cloud server, so that the cloud server determines the training sample to which each candidate record belongs according to the house address identifier;
[0030] A determination module, configured to determine, in response to the instruction information sent by the cloud server, the training sample to which each of the candidate records collected locally belongs;
[0031] The training module is used to train the model nodes running locally in the federated learning model according to the training samples to which each candidate record belongs collected locally.
[0032] To achieve the above-mentioned purpose, the fifth aspect of the present application proposes another model training device for identifying abnormal use of a house, the device comprising:
[0033] A second acquisition module, used to acquire the house address identifier corresponding to the collected candidate records from at least two of the devices;
[0034] An identification module, used to determine the training sample to which each candidate record belongs according to the house address identifier;
[0035] An indication module is used to send corresponding indication information to each of the devices, wherein the indication information is used to indicate the training samples to which the candidate records collected by the corresponding device belong, so that the corresponding device uses the collected candidate records as training samples to train the model nodes running locally in the federated learning model.
[0036] To achieve the above-mentioned purpose, the sixth aspect of the present application provides a device for identifying abnormal use of a house, comprising:
[0037] A third acquisition module is used to acquire a target record to be identified; wherein the target record includes house address information and associated information associated with the house address information;
[0038] A processing module is used to input the associated information in the target record into a trained model node of a federated learning model to determine the abnormal usage of the house corresponding to the house address information; wherein the model node is trained using the model training method described in the first aspect or the model training method described in the second aspect.
[0039] To achieve the above-mentioned purpose, the seventh aspect embodiment of the present application proposes a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the model training method as described in the first aspect, or implements the model training method as described in the second aspect, or implements the abnormal use of a house identification method as described in the third aspect.
[0040] In order to achieve the above-mentioned objectives, the eighth aspect embodiment of the present application proposes a non-temporary computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the model training method as described in the first aspect, or implements the model training method as described in the second aspect, or implements the abnormal use of a house identification method as described in the third aspect.
[0041] In order to achieve the above-mentioned objectives, the ninth aspect embodiment of the present application proposes a computer program product. When the instructions in the computer program product are executed by the processor, the model training method as described in the first aspect is implemented, or the model training method as described in the second aspect is implemented, or the abnormal use of a house identification method as described in the third aspect is implemented.
[0042] In this embodiment, multiple candidate records are collected, and the house address information of each candidate record is irreversibly encoded to obtain the house address identifier corresponding to each candidate record, and the house address identifier corresponding to each candidate record is sent to the cloud server, so that the cloud server determines the training sample to which each candidate record belongs according to the house address identifier, and in response to the instruction information sent by the cloud server, determines the training sample to which each candidate record collected locally belongs, and trains the model nodes running locally in the federated learning model according to the training samples to which each candidate record collected locally belongs. Based on the house address identifier corresponding to each candidate record, data from multiple data sources are aligned to obtain the corresponding training samples, and the model nodes running locally in the federated learning model are trained based on the corresponding dimensional data of the training samples, thereby ensuring the data security of each device end and improving the training effect of each model node.
[0043] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:
[0045] Figure 1 A flow chart of a model training method for identifying abnormal use of a house provided in an embodiment of the present application;
[0046] Figure 2 A flow chart of another model training method for identifying abnormal use of a house provided in an embodiment of the present application;
[0047] Figure 3 A flow chart of another model training method for identifying abnormal use of a house provided in an embodiment of the present application;
[0048] Figure 4 An example of another model training method for identifying abnormal use of a house provided in an embodiment of the present application;
[0049] Figure 5 A flow chart of another model training method for identifying abnormal use of a house provided in an embodiment of the present application;
[0050] Figure 6 A flow chart of another model training method for identifying abnormal use of a house provided in an embodiment of the present application;
[0051] Figure 7 A flowchart of a method for identifying abnormal use of a house provided in an embodiment of the present application;
[0052] Figure 8 A schematic diagram of the structure of a model training device for identifying abnormal use of a house provided in an embodiment of the present application;
[0053] Fig. 9 A schematic diagram of the structure of another model training device for identifying abnormal use of a house provided in an embodiment of the present application;
[0054] Fig.10 A schematic diagram of the structure of a device for identifying abnormal use of a house provided in an embodiment of the present application;
[0055] Fig.11 A block diagram of an exemplary computer device provided for application embodiments. DETAILED DESCRIPTION
[0056] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.
[0057] The following describes the model training method and device for identifying abnormal use of a house according to an embodiment of the present application with reference to the accompanying drawings.
[0058] Figure 1 A flow chart of a model training method for identifying abnormal use of a house provided in an embodiment of the present application, wherein the method is applied to at least two devices in a federated learning system, and the federated learning system also includes a cloud server.
[0059] like Figure 1 As shown, the method comprises the following steps:
[0060] Step 101 : collect multiple candidate records, wherein each candidate record includes house address information and associated information associated with the house address information.
[0061] In the present embodiment, each of the two devices corresponds to a data source from which candidate records can be obtained, and the multiple candidate records collected by at least the two devices are from different data sources. Different data sources lead to different candidate records.
[0062] The house address information includes one or more of the corresponding province, city, district, street name, community name, building number, unit number, floor and room number, which is not limited in this embodiment. The associated information in the candidate record is used to characterize at least one of energy use, commodity consumption, house registration, household registration and industrial and commercial registration.
[0063] In this embodiment, the candidate records may be obtained from different data sources. Different data sources may lead to different association information associated with the house address information contained in the candidate records obtained from the data sources.
[0064] In an example of this embodiment, the data source includes a first data source related to energy use, a second data source related to commodity consumption, a third data source related to house registration, a fourth data source related to household registration, and a fifth data source related to industrial and commercial registration. The classification and quantity of the data sources are not limited in this embodiment.
[0065] As an example, in the candidate record provided by the first data source, the associated information includes but is not limited to one or more combinations of the following:
[0066] Periodic consumption of various energy sources;
[0067] The change in the periodic consumption of various energy sources relative to the set reference amount;
[0068] Average periodic consumption of various energy sources;
[0069] Maximum and / or minimum periodic consumption of various energy sources;
[0070] Median periodic consumption of each energy source.
[0071] Among them, various types of energy include electric energy, water energy, liquefied gas energy, etc.; the cycle can be set according to demand, for example, 1 year, 3 years or 5 years, which is not limited in this embodiment.
[0072] As an example, in the candidate records provided by the second data source, the associated information includes but is not limited to one or more combinations of the following:
[0073] Order information;
[0074] Tag data of the user who placed the order;
[0075] Social relationship data of the user who placed the order;
[0076] Portrait data of the user who placed the order.
[0077] Among them, the label data, social relationship data and portrait data of the ordering user reflect the user's background.
[0078] It should be noted that the order in this embodiment can come from a single online shopping platform in the selected area, or it can be a combination of multiple online platforms. The online platform can be a shopping platform or a food delivery platform, for example, JD.com, Taobao, and Tuanjie, which are not listed one by one in this embodiment.
[0079] Alternatively, online consumption data often also includes logistics and express delivery data. If there is no online consumption data, logistics and express delivery data, such as the average monthly express delivery volume, the average daily express delivery volume, the express delivery amount, the express delivery package size, etc., can also be used as consumption-related information and auxiliary modeling data to increase the data dimension.
[0080] As an example, in the candidate records provided by the third data source, the associated information includes but is not limited to one or more combinations of the following:
[0081] House type information;
[0082] House area information;
[0083] Information on housing planning and use;
[0084] Year of house construction;
[0085] Information on illegal housing construction;
[0086] Information about the neighborhood where the house is located.
[0087] The third data source may be obtained from a housing authority, a housing agency, etc. Optionally, the community information where the house is located includes one or more of the number of households in the community, the number of elevators, the number of floors of the house, the number of surveillance cameras, and the number of roads.
[0088] As an example, in the candidate records provided by the fourth data source, the associated information includes but is not limited to one or more combinations of the following:
[0089] Household information;
[0090] Permanent population information;
[0091] Rental population information.
[0092] The tenant population information may include the tenant’s age and rental period.
[0093] As an example, in the candidate records provided by the fifth data source, the associated information includes but is not limited to one or more combinations of the following:
[0094] Capital information of the registered entity;
[0095] Default information of the registered unit.
[0096] In this embodiment, since some houses can be used as both residential and commercial properties, fifth data related to industrial and commercial registration can be obtained to increase the information dimension contained in the house.
[0097] It should be noted that, in this embodiment, multiple candidate records can be obtained from the data sources corresponding to the at least two devices mentioned above.
[0098] Step 102: irreversibly encode the house address information of each candidate record to obtain the house address identifier corresponding to each candidate record.
[0099] In this embodiment, for each candidate record, the house address information contained in the candidate record is extracted, and the house address information is irreversibly encoded according to a preset encoding method, so that the irreversible code corresponding to the house address information can uniquely identify the house address information, that is, the house address identifier corresponding to each candidate record can be used to uniquely identify the house address information, thereby improving the uniqueness of each house address information, so that the candidate records corresponding to the same house address information can be aligned according to the unique house address information to determine the candidate records belonging to the same training sample.
[0100] Step 103, sending the house address identifier corresponding to each candidate record to the cloud server, so that the cloud server determines the training sample to which each candidate record belongs according to the house address identifier.
[0101] In this embodiment, at least two devices send the house address identifiers corresponding to the collected candidate records to the cloud server, and multiple training samples are determined in the cloud server, thereby protecting the privacy and security of the candidate record data collected by at least two devices. In other words, isolation between the candidate records collected by at least two devices is achieved, data leakage is avoided, and data security is improved while determining the training samples to which each candidate record belongs.
[0102] In this embodiment, at least two devices send the collected candidate records to the cloud server, and the cloud server aligns the house address identifiers corresponding to the candidate records, and divides at least two candidate records corresponding to the same house address identifier into the same data set, and determines the candidate records in the same data set as belonging to the same training sample. That is to say, each training sample has a corresponding house address identifier, and determines the training sample to which each candidate record belongs for each house address identifier, thereby increasing the amount of data information contained in the training samples, so as to improve the accuracy of subsequent model training.
[0103] Step 104, in response to the instruction information sent by the cloud server, determine the training sample to which each candidate record collected locally belongs.
[0104] The indication information is used to indicate the training sample to which the candidate record collected by the corresponding device belongs.
[0105] The training sample in this embodiment may be one or more. The more the number of training samples is, the better the effect of subsequent model training will be.
[0106] Step 105: Train the model nodes running locally in the federated learning model according to the training samples to which each candidate record belongs that are collected locally.
[0107] Each training sample is used to represent the associated information contained in two candidate records collected by at least two devices corresponding to the same address identifier.
[0108] In this embodiment, each device is locally configured with a corresponding node of the federated learning model, and different devices have different model nodes configured locally.
[0109] In one implementation of the embodiment of the present application, an unsupervised training method is used to train the model nodes running locally in the federated learning model. Specifically, each device performs model training on the model nodes running locally in the federated learning model based on the training samples belonging to each candidate record collected locally, and the model node training is completed when the loss function of the model node running locally in the federated learning model is minimized, so that the federated learning model node learns the correspondence between the associated information and the abnormal use of the house. That is to say, the model node of the federated learning model obtained by the training can be used to identify abnormal use of the house.
[0110] In another implementation method of the embodiment of the present application, a supervised training method is used to train the federated learning model. Specifically, according to the abnormal information marked by multiple candidate records, the abnormal information of the training samples is marked, and according to the multiple training samples marked with abnormal information, supervised training is performed on the nodes running locally of the federated learning model, so that the nodes running locally of the federated learning model learn the correspondence between the associated information and the abnormal usage of the house. That is to say, the model nodes of the federated learning model obtained by training can be used to identify abnormal usage of the house.
[0111] In the model training method of this embodiment, multiple candidate records are collected, and the house address information of each candidate record is irreversibly encoded to obtain the house address identifier corresponding to each candidate record, and the house address identifier corresponding to each candidate record is sent to the cloud server, so that the cloud server determines the training sample to which each candidate record belongs according to the house address identifier, and in response to the instruction information sent by the cloud server, determines the training sample to which each candidate record collected locally belongs, and trains the model nodes running locally in the federated learning model according to the training samples to which each candidate record collected locally belongs. Based on the house address identifier corresponding to each candidate record, data from multiple data sources are aligned to obtain the corresponding training samples, and the model nodes running locally in the federated learning model are trained based on the corresponding dimensional data of the training samples, thereby ensuring the data security of each device end and improving the training effect of each model node.
[0112] Based on the previous embodiment, Figure 2 A flowchart of another model training method for identifying abnormal usage of a house provided in this embodiment.
[0113] like Figure 2 As shown, the above step 101 may include the following steps:
[0114] Step 201, extracting house address information for each stored record.
[0115] In this embodiment, each device stores multiple records. Since each record contains house address information and associated information associated with the house address information, the house address information can be extracted from the records stored in each device. The house address information includes one or more of the corresponding province, city, district, street name, community name, building number, unit number, floor and room number, and each house address information has a corresponding house category.
[0116] Step 202, based on the house category to which the house address information belongs, select candidate records whose house category is residential from multiple stored records.
[0117] Among them, the housing categories include residential and commercial buildings, among which commercial buildings include office buildings, storefronts, etc.
[0118] In this embodiment, feature extraction is performed based on the house address information to determine the house category to which the house address information belongs. As an implementation method, in order to identify the house category, the real estate management department usually identifies the house category when generating the house address information to distinguish different house categories. As another implementation method, a corresponding relationship between the house address information and the corresponding house category is pre-established, so that the house category to which the house belongs can be determined based on the house address information.
[0119] It should be understood that in this embodiment, candidate records whose house type is residential are screened out, and candidate records whose house type is commercial are not processed. Normally, commercial houses are regularly inspected by relevant regulatory departments, and there are fewer cases of abnormal use.
[0120] Step 203, screening out candidate records based on the completeness of the house address information.
[0121] In this embodiment, the complete house address information includes the province, city, district, street community name, building number, unit number, floor and house number.
[0122] In this embodiment, based on the extracted house address information, it can also be determined whether the granularity of the house address information is up to the portal, that is, whether the house address information is complete and includes the province, city, district, street community name, building number, unit number, floor and house number. Then, the corresponding house address information set is determined based on the complete house address information, and for each complete house address information in the corresponding house address set, the candidate records corresponding to the complete house address information are screened out to delete the candidate records with incomplete house address information, and retain the complete candidate records corresponding to each house address information, thereby improving the accuracy of the candidate record determination.
[0123] In this embodiment, the house address information is extracted from each stored record, and according to the house category to which the house address information belongs, candidate records whose house category is residential are screened out from multiple stored records, thereby excluding commercial houses and improving the accuracy of determining the candidate records. Furthermore, based on the complete house address information, the candidate records are screened, thereby improving the accuracy of determining the candidate records.
[0124] Based on the above embodiments, Figure 3 A flowchart of another model training method for identifying abnormal usage of a house provided in this embodiment.
[0125] like Figure 3 As shown, the above step 105 may include the following steps:
[0126] Step 301 : Perform feature engineering processing on the associated information contained in each candidate record collected locally to obtain the attribute value of each set feature field in the corresponding training sample.
[0127] The feature engineering process includes at least one of null value filling, normalization, aggregation of multiple feature fields, and feature field generation.
[0128] In one implementation of this embodiment, for each candidate record collected locally, it is determined that the associated information contained therein does not conform to the data rules of the samples required for model training of the federated learning model, and feature engineering processing is performed on the corresponding associated information to obtain the attribute values of each set feature field in the corresponding training sample, wherein the feature engineering processing includes but is not limited to at least one of null value filling, normalization, aggregation of multiple feature fields and feature field generation, which is not limited in this embodiment.
[0129] It should be understood that feature engineering processing is not necessarily performed. In another implementation of this embodiment, for each candidate record collected locally, the associated information contained is determined. If it meets the data rules of the samples required for model training of the federated learning model, there is no need to perform feature engineering processing on the corresponding associated information, that is, the corresponding associated information that has not been processed is used as a training sample.
[0130] In this embodiment, other data processing methods for model training samples of the federated learning model are not limited in this embodiment.
[0131] Step 302: Input the attribute value of each set feature field in the training sample into the model node for training to obtain the model parameters of the model node.
[0132] In this embodiment, each device is locally configured with a corresponding node of the federated learning model, and different devices have different model nodes configured locally.
[0133] In this embodiment, for each electronic device, for the same training sample, the attribute value of the electronic device corresponding to the feature field set in the training sample is input into the model node corresponding to the electronic device for training to obtain the model parameters of the model node corresponding to the electronic device. Similarly, the model parameters obtained by training the corresponding model node for each electronic device can be obtained. The model node running locally is trained using the attribute values of different feature fields belonging to the same training sample, and the model parameters for training the corresponding model node based on different dimensions of a training sample are obtained.
[0134] Step 303, sending the model parameters to the cloud server, so that the cloud server aggregates the model parameters sent by each device to obtain the model parameters after the model nodes are aggregated.
[0135] Step 304, obtaining the aggregated model parameters of the model nodes from the cloud server, and updating the model parameters of the model nodes according to the aggregated model parameters of the model nodes.
[0136] In this embodiment, each device sends the model parameters corresponding to different data dimensions obtained from the training of the local model nodes to the cloud server. The cloud server aggregates the model parameters sent by each device to obtain the aggregated model parameters of the model nodes. As a possible implementation method, the cloud server determines the weights of the model parameters determined by each electronic device based on the loss function determined when at least two devices use the data of the corresponding dimensions in the corresponding training samples to train the model nodes, and performs weighted averaging based on the weights to determine the model parameters after the model nodes are aggregated.
[0137] Furthermore, each device obtains the model parameters after the model nodes are aggregated from the cloud server, and updates the model parameters of each model node according to the model parameters after the model nodes are aggregated. Through continuous iteration, the effect of model training is improved.
[0138] In the model training method of this embodiment, feature engineering processing is performed on the associated information contained in each candidate record collected locally to obtain the attribute values of each set feature field in the corresponding training sample, and the attribute values of each set feature field in the training sample are input into the model node for training to obtain the model parameters of the model node, and the model parameters are sent to the cloud server so that the cloud server aggregates the model parameters sent by each device to obtain the aggregated model parameters of the model node. It is realized that the locally running model nodes are trained by using the attribute values of different feature fields belonging to the same training sample, and the model parameters of different dimensions based on a training sample are obtained, and the model parameters are converged in the cloud. According to the converged model parameters, the direction of local model training is adjusted to gradually reduce the loss function, and multi-party joint training is realized to continuously adjust the loss function of the locally running model nodes, which not only avoids data leakage but also improves the training effect of each model node.
[0139] Based on the above embodiments, Figure 4 , the model training method for identifying abnormal house usage in this embodiment is described.
[0140] like Figure 4 As shown, the region to be queried is determined, and multiple candidate records are obtained from the data sources corresponding to the five devices in the region. The five data sources are respectively a first data source related to energy use, a second data source related to commodity consumption, a third data source related to house registration, a fourth data source related to household registration, and a fifth data source related to industrial and commercial registration. Among them, candidate records are provided for each data source, and the candidate records provided by each data source may be one or more, and each candidate record has corresponding house address information and associated information associated with the house address information. Among them, the associated information that can be included in the candidate records provided by different data sources can refer to the description in any of the above embodiments, and will not be repeated in this embodiment.
[0141] Then, based on the candidate records provided by each data source, the house address information is extracted, and based on the extracted house address information, the house category corresponding to the house address information is identified and determined, the house address information belonging to the residential category is screened out, and the candidate records corresponding to the residential category house address information are determined. Among them, based on the extracted house address information, it can also be determined whether the granularity contained in the house address information is up to the portal, that is, whether the house address information is complete and includes the province, city, district, street community name, building number, unit number, floor and house number. Then, based on the complete house address information, the corresponding house address information set is determined, and for each complete house address information in the corresponding house address set, the candidate records provided by each data source are screened to delete the candidate records with incomplete house address information, retain the complete candidate records corresponding to each house address information, and improve the accuracy of the candidate records corresponding to each house address information.
[0142] Further, the house address is irreversibly encoded to determine the house address identifier, wherein the determined house address identifier is a unique identifier, thereby determining the house address identifier corresponding to each candidate record.
[0143] Then, the house address identifier corresponding to each candidate record will be sent to the cloud server so that the cloud server can align according to the house address identifier. Based on the uniqueness of the house address identifier, the candidate records corresponding to the same house address identifier can be accurately determined to belong to the same data set, and multiple data sets can be obtained. The candidate records in the same data set are determined to belong to the same training sample, thereby determining the training sample to which each candidate record belongs. In response to the indication information sent by the cloud server, the training sample to which each candidate record collected locally belongs is determined.
[0144] Furthermore, feature engineering is performed on the associated information contained in at least two candidate records in each data set to obtain the attribute values of each set feature field in the corresponding training sample. The attribute values of each set feature field in the training sample are input into each model node of the federated learning model for training, so that the locally running model nodes are trained by using the attribute values of different feature fields belonging to the same training sample, and the model parameters of different dimensions based on a training sample are obtained. The model parameters are converged in the cloud, and the direction of local model training is adjusted according to the converged model parameters to gradually reduce the loss function, so that multi-party joint training is realized to continuously adjust the loss function of the locally running model nodes, which not only avoids data leakage, but also improves the training effect of each model node.
[0145] In order to implement the above embodiment, this embodiment provides another model training method for identifying abnormal use of a house, which is applied to a cloud server in a federated learning system, and the federated learning system also includes at least two devices.
[0146] Figure 5 A flow chart of another model training method for identifying abnormal housing use provided in an embodiment of the present application is shown as follows: Figure 5 As shown, the method comprises the following steps:
[0147] Step 501: Acquire house address identifiers corresponding to collected candidate records from at least two devices.
[0148] The execution subject of this embodiment is a cloud server.
[0149] In an example of this embodiment, the house address identifier is obtained by irreversibly encoding the house address information contained in the candidate records collected by the device to obtain the house address identifier corresponding to each candidate record. The house address identifier can be used to uniquely identify the corresponding house address information, that is, it can uniquely identify the corresponding candidate record.
[0150] Step 502: Determine the training sample to which each candidate record belongs according to the house address identifier.
[0151] In this embodiment, multiple training samples are determined in the cloud server, thereby protecting the privacy and security of candidate record data collected by at least two devices. In other words, isolation between candidate records collected by at least two devices is achieved, data leakage is avoided, and data security is improved.
[0152] In this embodiment, the cloud server aligns the house address identifiers corresponding to each candidate record, and determines that the candidate records corresponding to the same house address identifier belong to the same training sample, thereby aggregating the corresponding candidate records for each house address identifier to obtain the corresponding training sample. In other words, each training sample has a corresponding house address identifier, that is, the corresponding house address information. Through alignment and aggregation, the amount of data information contained in the training sample is increased, so as to improve the accuracy of subsequent model training.
[0153] Step 503, sending corresponding indication information to each device, where the indication information is used to indicate the training samples to which the candidate records collected by the corresponding device belong, so that the corresponding device uses the collected candidate records as training samples to train the model nodes running locally in the federated learning model.
[0154] In this embodiment, the cloud server sends corresponding indication information to each device, indicating the training samples to which the candidate records collected by each device belong.
[0155] For example, device 1 collects 2 candidate records and, based on the indication information, determines that candidate record 1 belongs to training sample 1 and candidate record 2 belongs to training sample 2; device 2 collects 2 candidate records and, based on the indication information, determines that candidate record 3 belongs to training sample 2 and candidate record 4 belongs to training sample 1.
[0156] Among them, the implementation method of at least two devices training the model nodes running locally of the federated learning model based on the training samples belonging to each candidate record collected locally can be as described in the above embodiment, the principle is the same, and it will not be repeated here.
[0157] In the model training method of the embodiment of the present application, the cloud server obtains the house address identifiers corresponding to the collected candidate records from at least two devices, determines the training samples to which each candidate record belongs based on the house address identifier, and sends corresponding indication information to each device, where the indication information is used to indicate the training samples to which the candidate records collected by the corresponding device belong, so that the corresponding device uses the collected candidate records as training samples to train the model nodes running locally in the federated learning model. Based on the house address identifiers corresponding to the candidate records, the cloud aligns and aggregates data from multiple data sources to obtain corresponding training samples, thereby improving the dimensionality of the information contained in the training samples and improving the privacy and security of the information in each dimensionality. Furthermore, each device trains the model nodes running locally in the federated learning model based on the determined training samples, thereby improving the training effect of each model node.
[0158] Based on the above embodiments, Figure 6 A flow chart of another model training method for identifying abnormal housing use provided in an embodiment of the present application is shown as follows: Figure 6 As shown, step 502 includes the following steps:
[0159] Step 601: divide at least two candidate records with matching house address identifiers into the same data set.
[0160] In this embodiment, each device corresponds to a data source from which candidate records can be obtained, and the multiple candidate records collected by at least two devices are from different data sources. Different data sources lead to different candidate records.
[0161] The associated information in the candidate record is used to characterize at least one of energy use, commodity consumption, house registration, household registration, and industrial and commercial registration.
[0162] In this embodiment, the candidate records may be obtained from different data sources. Different data sources may lead to different association information associated with the house address information contained in the candidate records obtained from the data sources.
[0163] In an example of this embodiment, the data source includes a first data source related to energy use, a second data source related to commodity consumption, a third data source related to house registration, a fourth data source related to household registration, and a fifth data source related to industrial and commercial registration. The data source in this embodiment is only an example, and the classification and quantity of the data source are not limited in this embodiment.
[0164] As an example, in the candidate record provided by the first data source, the associated information includes but is not limited to one or more combinations of the following:
[0165] Periodic consumption of various energy sources;
[0166] The change in the periodic consumption of various energy sources relative to the set reference amount;
[0167] Average periodic consumption of various energy sources;
[0168] Maximum and / or minimum periodic consumption of various energy sources;
[0169] Median periodic consumption of each energy source.
[0170] Among them, various types of energy include electric energy, water energy, liquefied gas energy, etc.; the cycle can be set according to demand, for example, 1 year, 3 years or 5 years, which is not limited in this embodiment.
[0171] As an example, in the candidate records provided by the second data source, the associated information includes but is not limited to one or more combinations of the following:
[0172] Order information;
[0173] Tag data of the user who placed the order;
[0174] Social relationship data of the user who placed the order;
[0175] Portrait data of the user who placed the order.
[0176] Among them, the label data, social relationship data and portrait data of the ordering user reflect the user's background.
[0177] It should be noted that the order in this embodiment can come from a single online shopping platform in the selected area, or it can be a combination of multiple online platforms. The online platform can be a shopping platform or a food delivery platform, for example, JD.com, Taobao, and Tuanjie, which are not listed one by one in this embodiment.
[0178] Alternatively, online consumption data often also includes logistics and express delivery data. If there is no online consumption data, logistics and express delivery data, such as the average monthly express delivery volume, the average daily express delivery volume, the express delivery amount, the express delivery package size, etc., can also be used as consumption-related information and auxiliary modeling data to increase the data dimension.
[0179] As an example, in the candidate records provided by the third data source, the associated information includes but is not limited to one or more combinations of the following:
[0180] House type information;
[0181] House area information;
[0182] Information on housing planning and use;
[0183] Year of house construction;
[0184] Information on illegal housing construction;
[0185] Information about the neighborhood where the house is located.
[0186] The third data source may be obtained from a housing authority, a housing agency, etc. Optionally, the community information where the house is located includes one or more of the number of households in the community, the number of elevators, the number of floors of the house, the number of surveillance cameras, and the number of roads.
[0187] As an example, in the candidate records provided by the fourth data source, the associated information includes but is not limited to one or more combinations of the following:
[0188] Household information;
[0189] Permanent population information;
[0190] Rental population information.
[0191] The tenant population information may include the tenant’s age and rental period.
[0192] As an example, in the candidate records provided by the fifth data source, the associated information includes but is not limited to one or more combinations of the following:
[0193] Capital information of the registered entity;
[0194] Default information of the registered unit.
[0195] In this embodiment, since some houses can be used as both residential and commercial properties, fifth data related to industrial and commercial registration can be obtained to increase the information dimension contained in the house.
[0196] It should be noted that, in this embodiment, multiple candidate records can be obtained from the data sources corresponding to the at least two devices mentioned above.
[0197] In this embodiment, two devices are used as an example for explanation, wherein the data source corresponding to device 1 is related to energy usage, and the data source corresponding to device 2 is related to commodity consumption. The corresponding relationship between the house address identifier, candidate record and data set is shown in Table 1 below.
[0198]
[0199] Table 1
[0200] For example, the house address identifier 1, the two candidate records contained in the corresponding data set 1 obtained by aggregation are candidate record 1 involving energy collected by device 1 and candidate record a involving energy collected by device 2, which realizes the aggregation of relevant data for the same house address information and improves the information dimension contained in each house address information.
[0201] In one implementation of the embodiment of the present application, in order to improve the accuracy of determining the same data set, the house address identifier includes a first part of corresponding coded characters and a second part of corresponding text. Thus, the above step 501 includes the following steps: for any two house address identifiers, based on the semantic similarity of the second part and the text matching of the first part, identify whether the two house address identifiers match.
[0202] Step 602: determine candidate records in the same data set as belonging to the same training sample.
[0203] Furthermore, candidate records in the same data set are determined to belong to the same training sample, thereby increasing the dimension of information contained in each training sample.
[0204] It should be noted that, for different house address information, the number of associated information contained in the two candidate records corresponding to the corresponding records may be the same or different, that is, the aggregated data sets may be the same or different, which is not limited in this embodiment. For example, when the associated information and the number are different, the house address identifier 1 corresponds to the associated information 1 of the electricity consumption and the water consumption. The house address identifier 2 corresponds to the associated information 2 of the electricity consumption, the water consumption and the gas consumption, as well as the change in the periodic consumption of electricity, water and gas relative to the set reference amount.
[0205] In the model training method of this embodiment, the house address information of candidate records originating from data sources corresponding to different devices is irreversibly encoded to obtain the house address identifier corresponding to each candidate record, thereby achieving the uniqueness of the house address identifier, and thus the candidate records corresponding to the same house address identifier are determined to belong to the same data set, so that when the corresponding data set is summarized for each house address information, the accuracy of the data set summary is improved.
[0206] In order to implement the above embodiment, Figure 7 A flowchart of a method for identifying abnormal use of a house provided in an embodiment of the present application.
[0207] like Figure 7 As shown, the method comprises the following steps:
[0208] Step 701, obtaining target records to be identified, wherein each target record includes house address information and associated information associated with the house address information.
[0209] In this embodiment, the target record to be identified may be obtained from different devices, and the target record may be one or more. In this embodiment, different devices may cause different associated information associated with the house address information contained in the obtained candidate record.
[0210] Each device corresponds to a data source from which candidate records can be obtained, and the multiple candidate records collected by at least two devices are from different data sources. Different data sources lead to different candidate records.
[0211] The house address information includes one or more of the corresponding province, city, district, street name, community name, building number, unit number, floor and room number, which is not limited in this embodiment. The associated information in the candidate record is used to characterize at least one of energy use, commodity consumption, house registration, household registration and industrial and commercial registration.
[0212] In this embodiment, the candidate records may be obtained from different data sources. Different data sources may lead to different association information associated with the house address information contained in the candidate records obtained from the data sources.
[0213] In an example of this embodiment, the data source includes a first data source related to energy use, a second data source related to commodity consumption, a third data source related to house registration, a fourth data source related to household registration, and a fifth data source related to industrial and commercial registration. The classification and quantity of data sources are not limited in this embodiment. As an example, in the candidate records provided by the first data source, the associated information includes but is not limited to one or more combinations of the following:
[0214] Periodic consumption of various energy sources;
[0215] The change in the periodic consumption of various energy sources relative to the set reference amount;
[0216] Average periodic consumption of various energy sources;
[0217] Maximum and / or minimum periodic consumption of various energy sources;
[0218] Median periodic consumption of each energy source.
[0219] Among them, various types of energy include electric energy, water energy, liquefied gas energy, etc.; the cycle can be set according to demand, for example, 1 year, 3 years or 5 years, which is not limited in this embodiment.
[0220] As an example, in the candidate records provided by the second data source, the associated information includes but is not limited to one or more combinations of the following:
[0221] Order information;
[0222] Tag data of the user who placed the order;
[0223] Social relationship data of the user who placed the order;
[0224] Portrait data of the user who placed the order.
[0225] Among them, the label data, social relationship data and portrait data of the ordering user reflect the user's background.
[0226] It should be noted that the order in this embodiment can come from a single online shopping platform in the selected area, or it can be a combination of multiple online platforms. The online platform can be a shopping platform or a food delivery platform, for example, JD.com, Taobao, and Tuanjie, which are not listed one by one in this embodiment.
[0227] Alternatively, online consumption data often also includes logistics and express delivery data. If there is no online consumption data, logistics and express delivery data, such as the average monthly express delivery volume, the average daily express delivery volume, the express delivery amount, the express delivery package size, etc., can also be used as consumption-related information and auxiliary modeling data to increase the data dimension.
[0228] As an example, in the candidate records provided by the third data source, the associated information includes but is not limited to one or more combinations of the following:
[0229] House type information;
[0230] House area information;
[0231] Information on housing planning and use;
[0232] Year of house construction;
[0233] Information on illegal housing construction;
[0234] Information about the neighborhood where the house is located.
[0235] The third data source may be obtained from a housing authority, a housing agency, etc. Optionally, the community information where the house is located includes one or more of the number of households in the community, the number of elevators, the number of floors of the house, the number of surveillance cameras, and the number of roads.
[0236] As an example, in the candidate records provided by the fourth data source, the associated information includes but is not limited to one or more combinations of the following:
[0237] Household information;
[0238] Permanent population information;
[0239] Rental population information.
[0240] The tenant population information may include the tenant’s age and rental period.
[0241] As an example, in the candidate records provided by the fifth data source, the associated information includes but is not limited to one or more combinations of the following:
[0242] Capital information of the registered entity;
[0243] Default information of the registered unit.
[0244] In this embodiment, since some houses can be used as both residential and commercial properties, fifth data related to industrial and commercial registration can be obtained to increase the information dimension contained in the house.
[0245] Step 702, input the associated information in the target record into the trained model node of the federated learning model to determine the abnormal usage of the house corresponding to the house address information.
[0246] Wherein, the model node.
[0247] The model nodes are Figure 1-Figure 6 The model training method described in any embodiment is used for training, and the training process of the federated learning model is not repeated in this embodiment.
[0248] In one implementation, the target record in this embodiment corresponds to the same house address information, wherein the target record may be multiple records obtained from the at least two data sources mentioned above, that is, the target record contains records from at least two data sources, thereby increasing the amount of information contained in the target record corresponding to the house address, thereby improving the accuracy of recognition when the federated learning model obtained through training is used for recognition.
[0249] Furthermore, in this embodiment, when it is identified that the house corresponding to the house address information is being used abnormally, the address and contact person of the house with abnormal use can be sent to a corresponding organization to resolve the abnormal use and improve safety.
[0250] It should be noted that, regarding the method of aggregating target records for the same house address information, reference can be made to the relevant explanations in the above embodiment, which will not be repeated in this embodiment.
[0251] In the method for identifying abnormal use of a house in this embodiment, the trained federated learning model is used to identify the target record to be identified, so as to determine whether the house corresponding to the house address information is used abnormally. The identification is performed based on multi-dimensional information, thereby improving the accuracy of abnormal identification.
[0252] In order to implement the above embodiment, the present application also proposes a model training device for identifying abnormal use of a house.
[0253] Figure 8 A schematic diagram of the structure of a model training device for identifying abnormal use of a house provided in an embodiment of the present application.
[0254] like Figure 8 As shown, the device includes: a first acquisition module 81, an encoding module 82, a sending module 83, a determination module 84 and a training module 85.
[0255] The first acquisition module 81 is used to collect multiple candidate records, wherein each candidate record includes house address information and associated information associated with the house address information.
[0256] The encoding module 82 is used to irreversibly encode the house address information of each candidate record to obtain the house address identifier corresponding to each candidate record.
[0257] The sending module 83 is used to send the house address identifier corresponding to each candidate record to the cloud server, so that the cloud server can determine the training sample to which each candidate record belongs according to the house address identifier.
[0258] The determination module 84 is used to determine the training sample to which each candidate record collected locally belongs in response to the instruction information sent by the cloud server.
[0259] The training module 85 is used to train the model nodes running locally in the federated learning model according to the training samples to which each candidate record belongs that are collected locally.
[0260] Further, as a possible implementation, the first acquisition module 81 is specifically configured to:
[0261] The house address information is extracted from each stored record; and according to the house category to which the house address information belongs, the candidate records whose house category is residential are screened out from the plurality of stored records.
[0262] As a possible implementation, the device further includes:
[0263] The screening module is used to screen out the candidate records according to the completeness of the house address information.
[0264] As a possible implementation, the training module 85 is specifically used for:
[0265] Performing feature engineering processing on the associated information contained in each candidate record collected locally to obtain attribute values corresponding to each set feature field in the training sample; wherein the feature engineering processing includes at least one of null value filling, normalization, aggregation of multiple feature fields, and feature field generation;
[0266] Inputting the attribute value of each set feature field in the training sample into the model node for training to obtain the model parameter of the model node;
[0267] Sending the model parameters to the cloud server, so that the cloud server aggregates the model parameters sent by each of the devices to obtain the aggregated model parameters of the model nodes;
[0268] The aggregated model parameters of the model nodes are obtained from the cloud server, and the model parameters of the model nodes are updated according to the aggregated model parameters of the model nodes.
[0269] As a possible implementation manner, the associated information in the candidate record is used to characterize at least one of energy use, commodity consumption, house registration, household registration, and industrial and commercial registration.
[0270] It should be noted that the above explanation of the method embodiment is also applicable to the device of this embodiment, and will not be repeated here.
[0271] In the model training device of this embodiment, multiple candidate records are obtained from at least two data sources, and the candidate records from different data sources are aligned according to the house address information contained therein to obtain multiple data sets, each of which contains at least two candidate records corresponding to the same house address information. Based on the multiple data sets, multiple corresponding training samples are generated, and each training sample is used to characterize the associated information contained in at least two candidate records in the corresponding data set. Based on the multiple training samples, the federated learning model is trained so that the federated learning model learns the corresponding relationship between the associated information and the abnormal use of the house. The training samples are obtained by aggregating data from multiple data sources to train the federated learning model so that the trained model can accurately and efficiently identify the abnormal use of the house.
[0272] In order to implement the above embodiment, this embodiment provides another model training device for identifying abnormal use of a house, such as Fig. 9 As shown, the device comprises:
[0273] A second acquisition module 91 is used to acquire the house address identifier corresponding to the collected candidate records from at least two of the devices;
[0274] An identification module 92, configured to determine the training sample to which each candidate record belongs according to the house address identifier;
[0275] The indication module 93 is used to send corresponding indication information to each of the devices, where the indication information is used to indicate the training samples to which the candidate records collected by the corresponding device belong, so that the corresponding device uses the collected candidate records as training samples to train the model nodes running locally in the federated learning model.
[0276] Furthermore, in an implementation of the embodiment of the present application, the identification module 92 is specifically used to:
[0277] At least two candidate records matching the house address identifiers are divided into the same data set; and the candidate records in the same data set are determined to belong to the same training sample.
[0278] In one implementation of the embodiment of the present application, the house address identifier includes a first part of corresponding coded characters and a second part of corresponding text; the above-mentioned recognition module 92 is further specifically used for:
[0279] For any two of the house address identifiers, whether the two house address identifiers match is identified based on the semantic similarity of the second part and the text matching of the first part.
[0280] It should be noted that the above explanation of the method embodiment is also applicable to the device of this embodiment, and will not be repeated here.
[0281] In the model training device of the embodiment of the present application, the cloud server obtains the house address identifiers corresponding to the collected candidate records from at least two devices, determines the training samples to which each candidate record belongs based on the house address identifier, and sends corresponding indication information to each device, where the indication information is used to indicate the training samples to which the candidate records collected by the corresponding device belong, so that the corresponding device uses the collected candidate records as training samples to train the model nodes running locally in the federated learning model. Based on the house address identifiers corresponding to the candidate records, the cloud aligns and aggregates data from multiple data sources to obtain corresponding training samples, thereby increasing the dimensionality of the information contained in the training samples and increasing the privacy and security of the information in each dimensionality. Furthermore, each device trains the model nodes running locally in the federated learning model based on the determined training samples, so that the trained model can accurately and efficiently identify abnormal usage of the house.
[0282] In order to implement the above embodiment, this embodiment provides another device for identifying abnormal use of a house.
[0283] Fig.10 A schematic diagram of the structure of a device for identifying abnormal use of a house provided in an embodiment of the present application.
[0284] like Fig.10 As shown, the device includes: a third acquisition module 110 and a processing module 111.
[0285] The third acquisition module 110 is used to acquire a target record to be identified; wherein the target record includes house address information and associated information associated with the house address information.
[0286] Processing module 111 is used to input the associated information in the target record into a trained model node of a federated learning model to determine the abnormal usage of the house corresponding to the house address information; wherein the model node is trained using any of the aforementioned embodiments of the model training method for identifying abnormal usage of a house.
[0287] It should be noted that the above explanation of the embodiment of the method for identifying abnormal use of a house is also applicable to the device of this embodiment, and the principles are the same, which will not be repeated here.
[0288] In the abnormal use identification device for a house in this embodiment, the trained federated learning model is used to identify the target record to be identified, so as to determine whether the house corresponding to the house address information is used abnormally. The identification is performed based on multi-dimensional information, thereby improving the accuracy of abnormal identification.
[0289] In order to implement the above-mentioned embodiment, the embodiment of the present application proposes a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the above-mentioned Figure 1-4 The model training method described in the method embodiment, or, implementing the above Figure 5-6 The model training method described in the method embodiment, or, implementing Figure 7 The method for identifying abnormal use of a house described in the embodiment.
[0290] In order to implement the above-mentioned embodiment, the embodiment of the present application proposes a non-temporary computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned Figure 1-4 The model training method described in the method embodiment, or, implementing the above Figure 5-6 The model training method described in the method embodiment, or, implementing Figure 7 The method for identifying abnormal use of a house described in the embodiment.
[0291] In order to implement the above embodiment, the present application embodiment provides a computer program product. When the instructions in the computer program product are executed by a processor, the above Figure 1-4 The model training method described in the method embodiment, or, implementing the above Figure 5-6 The model training method described in the method embodiment, or, implementing Figure 7 The method for identifying abnormal use of a house described in the embodiment.
[0292] In order to implement the above embodiment, Fig.11 A block diagram of an exemplary computer device provided for application embodiments. Fig.11 The computer device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.
[0293] like Fig.11 As shown, the computer device 12 is in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that connects various system components (including the system memory 28 and the processing unit 16).
[0294] The bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnection (PCI) bus.
[0295] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.
[0296] The memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Fig.11 not shown, usually called a "hard drive"). Although Fig.11 Not shown, a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a compact disc read only memory (CD-ROM), a digital versatile disc read only memory (DVD-ROM), or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present application.
[0297] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.
[0298] The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), one or more devices that enable a user to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed via an input / output (I / O) interface 22. In addition, the computer device 12 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the computer device 12 via a bus 18. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.
[0299] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the methods mentioned in the above embodiments.
[0300] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.
[0301] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
[0302] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.
[0303] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.
[0304] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0305] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.
[0306] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.
[0307] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.
Claims
1. A model training method for identifying abnormal housing use, characterized in that: The method is applied to at least two devices in a federated learning system, wherein the federated learning system further includes a cloud server, and each of the devices performs the following steps: Collecting multiple candidate records; wherein each candidate record includes house address information and associated information associated with the house address information, and the associated information in the candidate record is used to characterize at least one of energy use, commodity consumption, house registration, household registration, and industrial and commercial registration; Perform irreversible encoding of the house address information on each candidate record to obtain a house address identifier corresponding to each candidate record, wherein the house address identifier is used to uniquely identify the house address information; Sending each candidate record and the house address identifier corresponding to each candidate record to the cloud server, the cloud server aligns the house address identifiers corresponding to each candidate record, and divides at least two candidate records corresponding to the same house address identifier into the same data set to determine the training samples to which each candidate record belongs; In response to the instruction information sent by the cloud server, determining the training sample to which each of the candidate records collected locally belongs; The model nodes running locally in the federated learning model are trained according to the training samples to which each of the candidate records belongs that are collected locally.
2. The model training method according to claim 1, characterized in that: The collecting of multiple candidate records includes: Extracting the house address information from each stored record; According to the house category to which the house address information belongs, the candidate records whose house category is residential are screened out from the multiple stored records.
3. The model training method according to claim 2, characterized in that: The method further comprises: The candidate records are screened out according to the completeness of the house address information.
4. The model training method according to claim 1, characterized in that: The training of the model nodes running locally in the federated learning model according to the training samples to which the candidate records belong collected locally includes: Performing feature engineering processing on the associated information contained in each candidate record collected locally to obtain attribute values corresponding to each set feature field in the training sample; wherein the feature engineering processing includes at least one of null value filling, normalization, aggregation of multiple feature fields, and feature field generation; Inputting the attribute value of each set feature field in the training sample into the model node for training to obtain the model parameter of the model node; Sending the model parameters to the cloud server, so that the cloud server aggregates the model parameters sent by each of the devices to obtain the aggregated model parameters of the model nodes; The aggregated model parameters of the model nodes are obtained from the cloud server, and the model parameters of the model nodes are updated according to the aggregated model parameters of the model nodes.
5. A model training method for identifying abnormal housing use, characterized in that: A cloud server applied to a federated learning system, wherein the federated learning system further comprises at least two devices; and the method comprises: Receive multiple candidate records sent by at least two of the devices and house address identifiers corresponding to each of the candidate records, where the house address identifier is used to uniquely identify the house address information; wherein each of the candidate records includes house address information and associated information associated with the house address information, and the associated information in the candidate records is used to characterize at least one of energy use, commodity consumption, house registration, household registration, and industrial and commercial registration; irreversibly encode the house address information of each of the candidate records to obtain a house address identifier corresponding to each of the candidate records; Aligning the house address identifiers corresponding to the candidate records, and dividing at least two candidate records corresponding to the same house address identifier into the same data set, so as to determine the training samples to which the candidate records belong; Send corresponding indication information to each of the devices, where the indication information is used to indicate the training samples to which the candidate records collected by the corresponding device belong, so that the corresponding device uses the collected candidate records as training samples to train the model nodes running locally in the federated learning model.
6. The model training method according to claim 5, characterized in that: Determining the training sample to which each candidate record belongs according to the house address identifier includes: Dividing at least two candidate records matching the house address identifier into the same data set; Candidate records in the same data set are determined to belong to the same training sample.
7. The model training method according to claim 6, characterized in that: The house address identifier includes a first part of corresponding coded characters and a second part of corresponding text; the at least two candidate records matching the house address identifier are divided into the same data set, including: For any two of the house address identifiers, whether the two house address identifiers match is identified based on the semantic similarity of the second part and the text matching of the first part.
8. A method for identifying abnormal use of a house, characterized in that: The following steps are involved: Acquire a target record to be identified; wherein the target record includes house address information and associated information associated with the house address information; The associated information in the target record is input into a trained model node of a federated learning model to determine abnormal usage of the house corresponding to the house address information; wherein the model node is trained using the model training method described in any one of claims 1-7.
9. A model training device for identifying abnormal use of a house, characterized in that: include: A first acquisition module is used to collect multiple candidate records; each of the candidate records includes house address information and associated information associated with the house address information, and the associated information in the candidate records is used to characterize at least one of energy use, commodity consumption, house registration, household registration and industrial and commercial registration; An encoding module, used for irreversibly encoding the house address information of each candidate record to obtain a house address identifier corresponding to each candidate record, wherein the house address identifier is used to uniquely identify the house address information; A sending module, used for sending each candidate record and the house address identifier corresponding to each candidate record to a cloud server, wherein the cloud server aligns the house address identifiers corresponding to each candidate record, and divides at least two candidate records corresponding to the same house address identifier into the same data set, so as to determine the training samples to which each candidate record belongs; A determination module, configured to determine, in response to the instruction information sent by the cloud server, the training sample to which each of the candidate records collected locally belongs; The training module is used to train the model nodes running locally in the federated learning model according to the training samples to which each candidate record belongs collected locally.
10. A model training device for identifying abnormal use of a house, characterized in that: include: A second acquisition module is used to receive multiple candidate records sent by at least two devices and house address identifiers corresponding to each of the candidate records, wherein the house address identifier is used to uniquely identify the house address information; wherein each of the candidate records includes house address information and associated information associated with the house address information, and the associated information in the candidate record is used to characterize at least one of energy use, commodity consumption, house registration, household registration, and industrial and commercial registration; irreversibly encode the house address information of each of the candidate records to obtain the house address identifier corresponding to each of the candidate records; An identification module, used to align the house address identifiers corresponding to the candidate records, and classify at least two candidate records corresponding to the same house address identifier as belonging to the same data set, so as to determine the training samples to which the candidate records belong; An indication module is used to send corresponding indication information to each of the devices, wherein the indication information is used to indicate the training samples to which the candidate records collected by the corresponding device belong, so that the corresponding device uses the collected candidate records as training samples to train the model nodes running locally in the federated learning model.
11. A device for identifying abnormal use of a house, characterized in that: include: A third acquisition module is used to acquire a target record to be identified; wherein the target record includes house address information and associated information associated with the house address information; A processing module, used to input the associated information in the target record into a trained model node of a federated learning model to determine the abnormal usage of the house corresponding to the house address information; wherein the model node is trained using the model training method described in any one of claims 1-7.
12. A computer device, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the model training method as described in any one of claims 1 to 7, or implements the abnormal use of a house identification method as described in claim 8.
13. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements the model training method as described in any one of claims 1 to 7, or implements the abnormal use of a house identification method as described in claim 8.
14. A computer program product, comprising a computer program, which, when executed by a processor, implements the model training method as described in any one of claims 1 to 7, or implements the method for identifying abnormal use of a house as described in claim 8.
