A method for identifying a resource consumption abnormal object and a related device

Abstract

The application discloses a resource consumption abnormal object identification method and related device, which is applied to the field of artificial intelligence. By obtaining unlabeled samples and positive samples, determining target feature dimensions, inputting the unlabeled samples and the positive samples into a semi-supervised learning framework to iteratively train a classification model, obtaining multiple identification feature values corresponding to the unlabeled samples output by the classification model in the iterative training process, and determining resource consumption abnormal objects, the resource consumption abnormal object identification process under a small amount of positive samples is realized. In the training process, the positive samples in the unlabeled samples are continuously mined heuristically for supplementation and added to the next round of iteration, effectively solving the sample imbalance problem in the identification scene and improving the accuracy of resource consumption abnormal object identification.

Description

[0001] This application claims priority to Chinese Patent Application No. 202110154406.5, filed on February 4, 2021, entitled "A Method for Identifying Group Rental Housing Based on Semi-Supervised Learning and Related Device", the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of computer technology, and in particular to a method and apparatus for identifying objects with abnormal resource consumption. Background Technology

[0003] Due to issues such as housing renovations, complex wiring, and fire hazards, shared rental housing poses significant risks. Identifying shared rental housing based on limited data presents a challenge, as such shared rental housing is considered an example of abnormal resource consumption.

[0004] Generally, a machine learning-based method for identifying shared rental housing can be used. This involves using machine learning classification algorithms to fit shared rental housing data, effectively distinguishing between shared rental housing and non-shared rental housing samples.

[0005] However, the process of identifying shared rental housing based on machine learning requires a large amount of data support. Due to the limitations of the disclosed data, there are often only a small number of positive samples and a large amount of unlabeled data in the shared rental housing data, which reduces the accuracy and effectiveness of shared rental housing identification and affects the accuracy of identifying objects with abnormal resource consumption. Summary of the Invention

[0006] In view of this, this application provides a method for identifying objects with abnormal resource consumption, which can effectively improve the accuracy of identifying objects with abnormal resource consumption.

[0007] The first aspect of this application provides a method for identifying objects with abnormal resource consumption, which can be applied to a system or program in a terminal device that includes a function for identifying shared rental housing, specifically including:

[0008] The resource usage data of the target object set is obtained as unlabeled samples, and the resource usage data corresponding to the objects that have been verified as shared rental housing is called as positive samples. The number of positive samples is less than the number of unlabeled samples.

[0009] The positive sample is subjected to feature extraction based on at least one feature term to obtain the target feature dimension, wherein the feature term is set based on resource usage features at a preset time granularity;

[0010] The unlabeled samples and the positive samples are input into a semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension. During the iterative training process, unlabeled samples that meet the preset conditions are labeled as supplementary samples, and the supplementary samples are used to update the positive samples.

[0011] The classification model is used to obtain multiple identification feature values ​​corresponding to the unlabeled samples output during the iterative training process, and these identification feature values ​​are fused to determine the shared rental housing objects in the target object set.

[0012] Optionally, in some possible implementations of this application, obtaining resource usage data of the target object set as unlabeled samples includes:

[0013] Determine the candidate data corresponding to the target object set within the data statistics scope;

[0014] The candidate data is divided based on the preset time granularity to obtain granular data;

[0015] The resource feature items in the granular data are statistically analyzed to obtain the resource usage data corresponding to the target object set;

[0016] The resource usage data is preprocessed to obtain the unlabeled samples.

[0017] Optionally, in some possible implementations of this application, the preprocessing of the resource usage data to obtain the unlabeled samples includes:

[0018] Determine the smallest granularity in the preset time granularity;

[0019] The resource usage data is traversed based on the minimum granularity to determine missing items and negative value items;

[0020] The missing and negative values ​​are replaced by replacement values ​​to preprocess the resource usage data and obtain the unlabeled samples.

[0021] Optionally, in some possible implementations of this application, the preprocessing of the resource usage data to obtain the unlabeled samples includes:

[0022] Obtain the average value from the resource usage data;

[0023] Identify the prominent items in the resource usage data that exceed the mean;

[0024] The value of the salient item is replaced with the mean to preprocess the resource usage data to obtain the unlabeled sample.

[0025] Optionally, in some possible implementations of this application, the preprocessing of the resource usage data to obtain the unlabeled samples includes:

[0026] Determine the data correspondence in the resource usage data;

[0027] Extract outliers from the data correspondence;

[0028] The overlapping values ​​in the anomalies are filtered out to preprocess the resource usage data and obtain the unlabeled samples.

[0029] Optionally, in some possible implementations of this application, the step of extracting features from the positive samples based on at least one feature term to obtain the target feature dimension includes:

[0030] Determine the numerical features corresponding to the feature terms;

[0031] Based on the numerical features, feature extraction is performed on the positive samples to obtain numerical span information;

[0032] The target feature dimension is determined based on the numerical span information.

[0033] Optionally, in some possible implementations of this application, the method further includes:

[0034] The numerical features within a preset time range are correlated to obtain the fluctuation characteristics;

[0035] Based on the fluctuation characteristics, feature extraction is performed on the positive samples to obtain the feature fluctuation range;

[0036] The target feature dimension is determined based on the range of feature fluctuations.

[0037] Optionally, in some possible implementations of this application, the method further includes:

[0038] Determine the characteristic time period corresponding to the numerical feature;

[0039] Based on the characteristic time period, feature extraction is performed on the positive samples to obtain time period resource usage information;

[0040] The target feature dimension is determined based on the resource usage information for the specified time period.

[0041] Optionally, in some possible implementations of this application, the method further includes:

[0042] The adjacent numerical features are compared to obtain periodic features;

[0043] The positive samples are analyzed based on the periodic characteristics to obtain the characteristic period;

[0044] The target feature dimension is determined based on the feature period.

[0045] Optionally, in some possible implementations of this application, the step of inputting the unlabeled samples and the positive samples into a semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension includes:

[0046] A training set is generated based on the unlabeled samples and the positive samples;

[0047] The training set is input into the semi-supervised learning framework, and a portion of the unlabeled samples are randomly selected as negative samples.

[0048] The preset model is trained based on the positive samples and the negative samples to obtain the classification model;

[0049] The unlabeled samples that were not extracted are identified according to the classification model to obtain the identification feature value corresponding to each sample in the unlabeled samples that were not extracted.

[0050] The identification feature values ​​are filtered based on the preset conditions to extract supplementary samples from the unlabeled samples, and the positive samples are updated based on the supplementary samples.

[0051] The random sampling process is repeated to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension.

[0052] Optionally, in some possible implementations of this application, the method further includes:

[0053] Determine the specific data of resource usage data corresponding to the target object set under different data dimensions;

[0054] Obtain multiple predicted values ​​corresponding to the specific data respectively;

[0055] The multiple predicted values ​​are weighted and calculated to obtain the target feature value;

[0056] Based on the target feature values, the shared rental housing objects in the target object set are determined.

[0057] Optionally, in some possible implementations of this application, the target object set is a community user set, the resource usage data is electricity consumption, the positive samples are from an executable third-party platform, and the third-party platform is used to monitor the group rental housing objects.

[0058] A second aspect of this application provides a device for identifying objects with abnormal resource consumption, comprising:

[0059] The acquisition unit is used to acquire resource usage data of the target object set as unlabeled samples and call the resource usage data corresponding to the objects that have been verified as having abnormal resource consumption as positive samples, wherein the number of positive samples is less than the number of unlabeled samples;

[0060] An extraction unit is used to extract features from the positive sample based on at least one feature term to obtain a target feature dimension, wherein the feature term is set based on resource usage features at a preset time granularity.

[0061] The training unit is used to input the unlabeled samples and the positive samples into a semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension. During the iterative training process, unlabeled samples that meet preset conditions are labeled as supplementary samples, and the supplementary samples are used to update the positive samples.

[0062] The identification unit is used to acquire multiple identification feature values ​​corresponding to the unlabeled samples output by the classification model during the iterative training process, and to fuse them based on the identification feature values ​​to determine the resource consumption abnormal objects in the target object set.

[0063] Optionally, in some possible implementations of this application, the acquisition unit is specifically used to determine candidate data corresponding to the target object set within the data statistics range;

[0064] The acquisition unit is specifically used to divide the candidate data based on the preset time granularity to obtain granular data;

[0065] The acquisition unit is specifically used to statistically analyze the resource feature items in the granular data to obtain the resource usage data corresponding to the target object set.

[0066] The acquisition unit is specifically used to preprocess the resource usage data to obtain the unlabeled samples.

[0067] Optionally, in some possible implementations of this application, the acquisition unit is specifically used to determine the smallest granularity in the preset time granularity;

[0068] The acquisition unit is specifically used to traverse the resource usage data based on the smallest granularity to determine missing items and negative value items.

[0069] The acquisition unit is specifically used to call replacement values ​​to replace the missing items and the negative value items, so as to preprocess the resource usage data to obtain the unlabeled samples.

[0070] Optionally, in some possible implementations of this application, the acquisition unit is specifically used to acquire the average number in the resource usage data;

[0071] The acquisition unit is specifically used to determine the prominent items in the resource usage data that exceed the average number;

[0072] The acquisition unit is specifically used to replace the value of the salient item with the mean to preprocess the resource usage data to obtain the unlabeled sample.

[0073] Optionally, in some possible implementations of this application, the acquisition unit is specifically used to determine the data correspondence in the resource usage data;

[0074] The acquisition unit is specifically used to extract abnormal items from the data correspondence relationship;

[0075] The acquisition unit is specifically used to filter overlapping values ​​in the anomalies to preprocess the resource usage data to obtain the unlabeled samples.

[0076] Optionally, in some possible implementations of this application, the extraction unit is specifically used to determine the numerical features corresponding to the feature item;

[0077] The extraction unit is specifically used to extract features from the positive sample based on the numerical features to obtain numerical span information;

[0078] The extraction unit is specifically used to determine the target feature dimension based on the numerical span information.

[0079] Optionally, in some possible implementations of this application, the extraction unit is specifically used to correlate the numerical features within a preset time range to obtain fluctuation features;

[0080] The extraction unit is specifically used to extract features from the positive sample based on the fluctuation characteristics to obtain the feature fluctuation range;

[0081] The extraction unit is specifically used to determine the target feature dimension based on the feature fluctuation range.

[0082] Optionally, in some possible implementations of this application, the extraction unit is specifically used to determine the feature time period corresponding to the numerical feature;

[0083] The extraction unit is specifically used to extract features from the positive sample based on the characteristic time period to obtain time period resource usage information;

[0084] The extraction unit is specifically used to determine the target feature dimension based on the time period resource usage information.

[0085] Optionally, in some possible implementations of this application, the extraction unit is specifically used to compare adjacent numerical features to obtain periodic features;

[0086] The extraction unit is specifically used to analyze the positive sample based on the periodic features to obtain the feature period;

[0087] The extraction unit is specifically used to determine the target feature dimension based on the feature period.

[0088] Optionally, in some possible implementations of this application, the training unit is specifically used to generate a training set based on the unlabeled samples and the positive samples;

[0089] The training unit is specifically used to input the training set into the semi-supervised learning framework and randomly select a portion of the unlabeled samples as negative samples.

[0090] The training unit is specifically used to train a preset model based on the positive samples and the negative samples to obtain the classification model;

[0091] The training unit is specifically used to identify the unlabeled samples that have not been extracted according to the classification model, so as to obtain the identification feature value corresponding to each sample in the unlabeled samples.

[0092] The training unit is specifically used to filter the recognition feature values ​​based on the preset conditions, so as to extract supplementary samples from the unlabeled samples and update the positive samples based on the supplementary samples;

[0093] The training unit is specifically used to repeat the random sampling process to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension.

[0094] Optionally, in some possible implementations of this application, the identification unit is specifically used to determine the specific data of the resource usage data corresponding to the target object set under different data dimensions;

[0095] The identification unit is specifically used to acquire multiple predicted values ​​corresponding to the specific data.

[0096] The identification unit is specifically used to perform weighted calculations on the multiple predicted values ​​to obtain the target feature value;

[0097] The identification unit is specifically used to determine the resource consumption abnormal objects in the target object set based on the target feature values.

[0098] A third aspect of this application provides a computer device, comprising: a memory, a processor, and a bus system; the memory is used to store program code; the processor is used to execute the group rental housing identification method described in the first aspect or any one of the first aspects according to the instructions in the program code.

[0099] The fourth aspect of this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the group rental housing identification method described in the first aspect or any one of the first aspects.

[0100] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the shared rental housing identification method provided in the first aspect or various optional implementations thereof.

[0101] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0102] By acquiring resource usage data from the target object set as unlabeled samples and calling resource usage data corresponding to objects verified as having abnormal resource consumption as positive samples, the number of positive samples is less than the number of unlabeled samples. Then, feature extraction is performed on the positive samples based on at least one feature term to obtain the target feature dimension. The feature term is set based on resource usage features at a preset time granularity. The unlabeled samples and positive samples are then input into a semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension. During this iterative training process, unlabeled samples that meet preset conditions are labeled as supplementary samples, which are used to update the positive samples. Finally, multiple identification feature values ​​corresponding to the unlabeled samples output by the classification model during the iterative training process are obtained and fused based on the identification feature values ​​to determine the objects with abnormal resource consumption in the target object set. This enables the identification of abnormal resource consumption objects with a small number of positive samples. By using multiple target feature dimensions to extract the features of abnormal resource consumption objects, the sensitivity of the classification model to the features of abnormal resource consumption objects can be guaranteed. Furthermore, during the training process, positive samples in the unlabeled samples are continuously heuristically mined and added to the next iteration, effectively solving the problem of sample imbalance in the scenario of identifying abnormal resource consumption objects and improving the accuracy of identifying abnormal resource consumption objects. Attached Figure Description

[0103] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0104] Figure 1 Network architecture diagram for the operation of the shared rental housing identification system;

[0105] Figure 2 A flowchart illustrating the process architecture for identifying shared rental housing provided in this application embodiment;

[0106] Figure 3 A flowchart illustrating a method for identifying objects with abnormal resource consumption, provided in an embodiment of this application;

[0107] Figure 4 A schematic diagram illustrating a method for identifying objects with abnormal resource consumption, provided in an embodiment of this application.

[0108] Figure 5 A flowchart illustrating another method for identifying abnormal resource consumption objects provided in this application embodiment;

[0109] Figure 6 A flowchart illustrating another method for identifying abnormal resource consumption objects provided in this application embodiment;

[0110] Figure 7 A flowchart illustrating another method for identifying abnormal resource consumption objects provided in this application embodiment;

[0111] Figure 8 A schematic diagram illustrating a scenario for another method of identifying abnormal resource consumption objects provided in an embodiment of this application;

[0112] Figure 9 This is a schematic diagram of the structure of a device for identifying objects with abnormal resource consumption, provided in an embodiment of this application.

[0113] Figure 10 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application;

[0114] Figure 11 This is a schematic diagram of the structure of a server provided in an embodiment of this application. Detailed Implementation

[0115] This application provides a method and related apparatus for identifying objects with abnormal resource consumption. It can be applied to systems or programs in terminal devices that include a function for identifying shared rental housing. The method involves acquiring resource usage data from a target object set as unlabeled samples and calling resource usage data corresponding to objects verified as having abnormal resource consumption as positive samples, where the number of positive samples is less than the number of unlabeled samples. Then, feature extraction is performed on the positive samples based on at least one feature term to obtain the target feature dimension. The feature term is set based on resource usage features at a preset time granularity. The unlabeled samples and positive samples are input into a semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension. During this iterative training, unlabeled samples that meet preset conditions are labeled as supplementary samples, which are used to update the positive samples. Finally, multiple identification feature values ​​corresponding to the unlabeled samples output by the classification model during iterative training are obtained and fused based on these identification feature values ​​to determine the objects with abnormal resource consumption in the target object set. This enables the identification of abnormal resource consumption objects with a small number of positive samples. By using multiple target feature dimensions to extract the features of abnormal resource consumption objects, the sensitivity of the classification model to the features of abnormal resource consumption objects can be guaranteed. Furthermore, during the training process, positive samples in the unlabeled samples are continuously heuristically mined and added to the next iteration, effectively solving the problem of sample imbalance in the scenario of identifying abnormal resource consumption objects and improving the accuracy of identifying abnormal resource consumption objects.

[0116] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0117] First, some terms that may appear in the embodiments of this application will be explained.

[0118] Semi-supervised learning (SSL) is a key research area in pattern recognition and machine learning. It is a learning method that combines supervised and unsupervised learning. SSL uses a large amount of unlabeled data, as well as labeled data simultaneously, to perform pattern recognition tasks.

[0119] PU Learning (Positive-unlabeled Learning): A research direction in semi-supervised learning, which refers to training a binary classifier with only positive class data and unlabeled data.

[0120] Shared rental housing refers to a housing rental method in which a person is concentrated in a small area by changing the structure and layout of the house and dividing the rooms into several small rooms for rent by the room or by the bed.

[0121] It should be understood that the shared rental housing identification method provided in this application can be applied to systems or programs in terminal devices that include shared rental housing identification functions, such as rental applications. Specifically, the shared rental housing identification system can run on systems such as... Figure 1 In the network architecture shown, such as Figure 1 The diagram shown illustrates the network architecture of a shared rental housing identification system. As illustrated, this system can identify shared rental housing from multiple information sources. Specifically, the terminal sends corresponding housing information to the server via a trigger operation. The server then uses semi-supervised learning-based methods to identify shared rental housing based on resource usage data within the corresponding community, thereby determining the corresponding rental type. This can be understood as... Figure 1 The document shows various terminal devices, which can be computer devices. In real-world scenarios, more or fewer types of terminal devices may participate in the process of identifying shared rental housing. The specific number and types depend on the actual scenario and are not limited here. Figure 1 The image shows one server, but in real-world scenarios, multiple servers can be involved, with the specific number depending on the actual situation.

[0122] In this embodiment, the server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, and the terminal and server can be connected to form a blockchain network; this application does not impose any restrictions.

[0123] It is understandable that the aforementioned shared rental housing identification system can run on personal mobile terminals, such as rental housing applications, or on servers, or on third-party devices to provide shared rental housing identification and obtain the shared rental housing identification processing results from the information source. Specifically, the shared rental housing identification system can run as a program on the aforementioned devices, or as a system component of the aforementioned devices, or as a cloud service program. The specific operating mode depends on the actual scenario and is not limited here.

[0124] Shared rental housing poses significant risks due to issues such as renovations, complex wiring, and fire hazards. Identifying shared rental housing based on limited data presents a major challenge.

[0125] Therefore, artificial intelligence (AI) can be used to solve the above-mentioned problems. AI is the theory, methods, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. In other words, AI is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to have the functions of perception, reasoning, and decision-making.

[0126] Generally, a machine learning-based method for identifying shared rental housing can be used. This involves using machine learning classification algorithms to fit shared rental housing data, effectively distinguishing between shared rental housing and non-shared rental housing samples.

[0127] Machine learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

[0128] However, the process of identifying shared rental housing based on machine learning requires a large amount of data. Due to the limitations of the disclosed data, there are often only a small number of positive samples and a large amount of unlabeled data in the shared rental housing data, which reduces the accuracy and effectiveness of shared rental housing identification and affects the accuracy of shared rental housing identification.

[0129] To address the aforementioned issues, this application proposes a method for identifying objects with abnormal resource consumption. This method is applied to... Figure 2 In the process framework shown for identifying shared rental housing, such as Figure 2 The diagram shown is a flowchart of a process for identifying shared rental housing provided in an embodiment of this application. The user sends the corresponding housing information to the server through the interactive operation of the terminal. Then, the server calls the resource usage information in the area (community or unit) corresponding to the housing information to perform a semi-supervised framework identification process, thereby obtaining the housing type prediction result and providing corresponding feedback to the terminal.

[0130] It is understood that the method provided in this application can be a program written as processing logic in a hardware system, or it can be a device for identifying abnormal resource consumption objects, implemented through integration or external means. As one implementation, this shared rental housing identification device introduces the PU Learning semi-supervised learning method into the shared rental housing identification scenario. Based on daily and monthly electricity consumption data of residents in the community provided by the power grid company, a training set is constructed by randomly combining a small number of positive samples and a large number of unlabeled samples. Effective features for identifying shared rental housing are extracted, and a binary classification model is trained using the LightGBM algorithm. Positive samples from the unlabeled samples are heuristically mined and added to the next iteration, effectively solving the problem of imbalanced samples in the shared rental housing identification scenario and greatly improving the accuracy of the shared rental housing identification algorithm. This method can quickly output highly suspicious residents in shared rental housing, facilitating accurate door-to-door sampling and investigation by law enforcement personnel, and effectively managing the phenomenon of shared rental housing.

[0131] The solutions provided in this application relate to machine learning technology in artificial intelligence, and are specifically illustrated through the following embodiments:

[0132] Based on the above process architecture, the method for identifying shared rental housing in this application will be described below. Please refer to [link / reference]. Figure 3 , Figure 3 This is a flowchart illustrating a method for identifying abnormal resource consumption objects according to an embodiment of this application. This identification method can be executed by a terminal, a server, or both. The following description uses terminal execution as an example. This embodiment of the application includes at least the following steps:

[0133] 301. Obtain the resource usage data of the target object set as unlabeled samples, and call the resource usage data corresponding to the objects that have been verified as having abnormal resource consumption as positive samples.

[0134] In this embodiment, the number of positive samples is less than the number of unlabeled samples; that is, positive samples are a small amount of data, while unlabeled samples are a large amount of data. Resource usage data can be data reflecting the daily living expenses of residents, such as electricity consumption, water consumption, or network resource consumption.

[0135] In addition, the abnormal resource consumption object can be a tenant, a homeowner, a resource account, or other entity corresponding to resource consumption. In one possible scenario, since the resource consumption of shared rental housing users is different from that of ordinary residents and is likely to cause safety hazards, the abnormal resource consumption object in this embodiment can also be a shared rental housing user (object).

[0136] The following example illustrates the implementation of a system where the abnormal resource consumption targets are users of shared rental housing, and the resource usage data is electricity consumption. This involves collecting and integrating daily and monthly electricity consumption data of residents in the community unit provided by the power grid company, as well as a small number of positive samples of shared rental housing provided by relevant departments. The specific form of the abnormal resource consumption targets and the form of the resource usage data will vary depending on the actual scenario and are only used as examples here.

[0137] Specifically, resource usage data can be obtained by organizing data based on a certain time granularity. First, candidate data corresponding to the target object set within the data statistical scope is determined. Then, the candidate data is divided based on a preset time granularity to obtain granular data. Resource feature items in the granular data are then statistically analyzed to obtain resource usage data corresponding to the target object set. Finally, the resource usage data is preprocessed to obtain unlabeled samples. For example, if the data statistical scope is two years, the preset time granularity includes days and months, and resource feature items include total electricity consumption, off-peak electricity consumption, and peak electricity consumption, then the resource usage data includes daily granular data: the total monthly electricity consumption, average monthly electricity consumption, and off-peak monthly electricity consumption for residents over the past two years; and monthly granular data: the total daily electricity consumption, average daily electricity consumption, peak daily electricity consumption, off-peak daily electricity consumption, and peak daily electricity consumption for residents over the past two years. The specific data division method depends on the actual scenario and is not limited here.

[0138] Optionally, since outliers may exist during data statistics, a data preprocessing process can be performed. This involves determining the smallest granularity within a preset time frame; then, based on this smallest granularity, the resource usage data is traversed to identify missing and negative values; and finally, replacement values ​​are used to replace these missing and negative values, thus preprocessing the resource usage data to obtain unlabeled samples. For example, missing electricity values ​​and negative electricity values ​​in daily electricity consumption can be uniformly replaced with 0, thereby ensuring data integrity.

[0139] Optionally, the mean value in the resource usage data can be obtained; then, outstanding items exceeding the mean value can be identified; and finally, the mean value can be used to replace the values ​​of the outstanding items to preprocess the resource usage data and obtain unlabeled samples. For example, data with electricity consumption values ​​exceeding 100 times the average of the total electricity consumption of users can be filled with the mean value, thereby avoiding the influence of outlier data on the overall data.

[0140] Optionally, during preprocessing, the data correspondence in the resource usage data can be determined; then, outliers in the data correspondence can be extracted; and finally, overlapping values ​​in the outliers can be filtered to preprocess the resource usage data and obtain unlabeled samples. This is because changes in storage format may result in two power values ​​being stored at a given time point. Therefore, a power value can be randomly selected as the power value for the current time point, thus ensuring data correspondence.

[0141] Meanwhile, based on a small number of positively labeled samples of shared rental housing provided by relevant departments, the corresponding samples in the user electricity data are labeled as 1, and the remaining unlabeled samples are uniformly labeled as 0, thus constructing a shared rental housing identification dataset.

[0142] In one possible scenario, the target set is a set of users in the community, or it can be a user group with a larger (area) or smaller (building) scope, while the resource usage data is electricity consumption. The positive sample comes from an executable third-party platform, which is used to monitor the group rental housing.

[0143] 302. Extract features from positive samples based on at least one feature term to obtain the target feature dimension.

[0144] In this embodiment, the feature terms are set based on the resource usage characteristics of objects with abnormal resource consumption at a preset time granularity, that is, the resource usage characteristics of shared rental housing objects. This is because resource usage data changes over time, thus reflecting the characteristics of shared rental housing, and the target feature dimension is the feature dimension used to identify shared rental housing.

[0145] Specifically, the target feature dimension extracts a series of fine-grained features based on the differences in electricity usage patterns between shared rental users and normal users. In the subsequent model training process, feature learning is performed based on the target feature dimension, thereby improving the sensitivity of the classification model to shared rental data.

[0146] Optionally, the target feature dimension can be open-sourced as a numerical dimension. This involves first determining the numerical features corresponding to the feature items; then extracting features from positive samples based on these numerical features to obtain numerical span information; and finally determining the target feature dimension based on this numerical span information. This is because shared rental housing often involves renovations, increasing the number of residents in a single unit, which in turn increases the use of electrical appliances, leading to higher electricity consumption. Therefore, the corresponding target feature dimensions are: the user's average / valley electricity consumption ranking within the corresponding community; statistical features of average / valley usage; statistical features of the user's average / valley / total electricity consumption last year; and the number of users with average monthly consumption values ​​less than 50 and valley values ​​less than 10, etc. The specific numerical composition depends on the actual scenario and is not limited here.

[0147] Optionally, numerical fluctuations can also be determined based on numerical features. First, numerical features within a preset time range (e.g., one year of data) are correlated to obtain fluctuation characteristics. Then, features are extracted from positive samples based on these fluctuation characteristics to obtain the feature fluctuation range. Finally, the target feature dimension is determined based on the feature fluctuation range. This is because the turnover of people in shared rental housing is relatively high, and there are periods of vacancy in the housing, resulting in relatively large fluctuations in user electricity consumption. Therefore, the corresponding target feature dimensions are: the user's average / valley / total electricity consumption changes for corresponding months this year and last year; and the user's fluctuation characteristics for adjacent months last year (e.g., the difference in electricity consumption between January and February).

[0148] Optionally, group rental housing can be identified based on feature segments within numerical features. This involves first determining the corresponding time periods for the numerical features; then extracting features from positive samples based on these time periods to obtain resource usage information; and finally determining the target feature dimension based on this information. This approach considers that group rental housing users are more likely to be working professionals, who consume less electricity during the day, resulting in a greater difference between peak and off-peak electricity consumption. Specifically, the feature is the ratio of electricity consumption from 8 AM to 5 PM to electricity consumption from 6 PM to midnight.

[0149] Additionally, the characteristic segments in the numerical features may also be due to lease terminations, meaning that shared rental apartments may have vacancy periods during which users consume less electricity. Specific features include: the proportion of months with 0% electricity consumption, and the proportion of outliers in electricity consumption (e.g., electricity consumption less than the average value minus three times the variance).

[0150] Optionally, adjacent numerical features can be compared to obtain periodic features; then, positive samples can be analyzed based on the periodic features to obtain the feature period; and the target feature dimension can be determined based on the feature period. This is because ordinary residential electricity consumption has a certain periodicity and stability, and the electricity consumption values ​​of adjacent months have similarity characteristics, while the similarity of users in shared rental housing is unstable. Therefore, shared rental housing can be identified through periodic features.

[0151] It is understandable that, since the above target feature dimensions are extracted based on positive samples identified as group rental rooms, the representativeness of the feature dimensions for group rental rooms is guaranteed.

[0152] 303. Input unlabeled samples and positive samples into the semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension.

[0153] In this embodiment, unlabeled samples that meet preset conditions during iterative training are labeled as supplementary samples, which are used to update positive samples. The iterative training process can also be called a heuristic training process, as detailed below. Figure 4 The scene architecture shown is Figure 4 This is a schematic diagram illustrating another method for identifying abnormal resource consumption objects provided in this application embodiment. It shows that all unlabeled user samples are labeled with 1, and randomly divided into N parts. Each time, one part is combined with the detected user samples as a training set for training the LightGBM model (classification model). Then, predictions are made for the remaining N-1 parts, and the predicted positive samples are added to the training set for the next round of training. This process is repeated until the predicted probabilities of all unverified user samples are obtained.

[0154] It is understandable that one or more copies can be used each time, but in order to avoid too many negative samples affecting the recognition results, one copy can be selected for iterative labeling, and the specific number depends on the actual scenario.

[0155] Specifically, the training process described above involves first generating a training set based on unlabeled samples and positive samples; then inputting the training set into a semi-supervised learning framework, and randomly selecting some samples from the unlabeled samples as negative samples; training a pre-defined model based on the positive and negative samples to obtain a classification model; then identifying the unselected unlabeled samples according to the classification model to obtain the identification feature value corresponding to each sample in the unselected unlabeled samples; filtering the identification feature values ​​based on pre-defined conditions (feature value greater than a specific value or feature value at the top of the size sort) to extract supplementary samples from the unlabeled samples, and updating the positive samples based on the supplementary samples, for example, applying the classification model to unlabeled samples OOB (out of bag) not in the training set, recording their scores, and adding samples with higher probability values ​​to the positive samples; then repeating the random sampling process to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension, and finally obtaining N-1 predicted probability values ​​for each unlabeled sample that is suspected to be a group rental sample.

[0156] Understandably, the above classification model uses the relatively lightweight LightGBM model. Since the electricity consumption data of users in the community unit has time-series characteristics, a sequence model such as LSTM can also be used for binary classification to automatically extract features and capture the changes in user electricity consumption patterns, saving the process of manually designing statistical features. The specific classification model depends on the actual scenario.

[0157] 304. Obtain multiple identification feature values ​​corresponding to the unlabeled samples output by the classification model during iterative training, and fuse them based on the identification feature values ​​to determine the abnormal resource consumption objects in the target object set.

[0158] In this embodiment, based on step 303, N-1 predicted probability values ​​can be obtained for each unlabeled sample that is suspected to be a group rental sample. Therefore, the N-1 results predicted by the classification model (e.g., LightGBM binary classification model) during training can be fused by mean to output the probability that the unlabeled sample is suspected to be a group sample. Specifically, the following formula can be used:

[0159]

[0160] Here, X1 to Xn-1 are the N-1 predicted probability values ​​for each unlabeled sample that is suspected to be a shared rental housing sample, and then the predicted mean is calculated.

[0161] Therefore, based on the calculated predicted mean, the group rental housing objects in the target object set are judged. Specifically, unlabeled samples with a predicted mean greater than a preset value (e.g., 0.9) can be regarded as group rental housing objects, or the unlabeled samples in the target object set can be sorted from largest to smallest according to the predicted mean, and the top 5 unlabeled samples in the sorted set can be regarded as group rental housing objects. The specific method depends on the actual scenario.

[0162] As can be seen from the above embodiments, resource usage data of the target object set is obtained as unlabeled samples, and resource usage data corresponding to objects verified as having abnormal resource consumption is called as positive samples, wherein the number of positive samples is less than the number of unlabeled samples; then, feature extraction is performed on the positive samples based on at least one feature term to obtain the target feature dimension, the feature term being set based on resource usage features at a preset time granularity; and the unlabeled samples and positive samples are input into a semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension. During this iterative training process, unlabeled samples that meet preset conditions are labeled as supplementary samples, and the supplementary samples are used to update the positive samples; then, multiple identification feature values ​​corresponding to the unlabeled samples output by the classification model during the iterative training process are obtained, and fusion is performed based on the identification feature values ​​to determine the objects with abnormal resource consumption in the target object set. This enables the identification of abnormal resource consumption objects with a small number of positive samples. By using multiple target feature dimensions to extract the features of abnormal resource consumption objects, the sensitivity of the classification model to the features of abnormal resource consumption objects can be guaranteed. Furthermore, during the training process, positive samples in the unlabeled samples are continuously heuristically mined and added to the next iteration, effectively solving the problem of sample imbalance in the scenario of identifying abnormal resource consumption objects and improving the accuracy of identifying abnormal resource consumption objects.

[0163] The following section describes the identification process of shared rental housing based on electricity consumption, using a modular design as an example. Figure 5 As shown, Figure 5A flowchart illustrating another method for identifying abnormal resource consumption objects provided in this application embodiment. This application embodiment includes at least the following module execution steps:

[0164] 501. Data Acquisition Module.

[0165] In this embodiment, the data acquisition module is used to obtain daily and monthly granular electricity consumption data of community unit users provided by the power grid company, as well as a small number of positive samples of group rental housing provided by relevant departments; specifically including daily granular data: the total monthly electricity consumption, average monthly electricity consumption, and off-peak monthly electricity consumption of residents over the past two years; and monthly granular data: the total daily electricity consumption, average daily electricity consumption, peak daily electricity consumption, off-peak daily electricity consumption, and peak daily electricity consumption of residents over the past two years.

[0166] 502. Data Preprocessing Module.

[0167] In this embodiment, the data preprocessing module is used to process outliers in user electricity consumption data. It also uses a small number of positive samples from shared rental housing to label the data as 1, while labeling the remaining unlabeled samples as 0. Specifically, missing electricity values ​​and negative electricity values ​​can be replaced with 0. Data with electricity values ​​exceeding 100 times the average of total user electricity consumption is filled with the average. If a change in storage format results in two electricity values ​​being stored at a given time point, a random electricity value is selected as the current electricity value.

[0168] 503. Feature Engineering Module.

[0169] In this embodiment, the feature engineering module is used to extract a series of fine-grained features based on the differences in electricity consumption patterns between group rental users and regular users, starting from the group rental scenario; the specific feature dimensions are detailed in [link to feature engineering module]. Figure 3 The description of the illustrated embodiment will not be repeated here.

[0170] 504. Model Training Module.

[0171] In this embodiment, the model training module is used to train the LightGBM binary classification model based on the PU Learning training framework using the constructed dataset. First, a training set is created by randomly combining all positive samples and unlabeled samples. Then, a classifier is built using "bootstrap" samples, treating positive samples and unlabeled samples as positive and negative, respectively. The classifier is then applied to unlabeled samples (OOB) that are not in the training set, and their scores are recorded. Samples with higher probability values ​​are added to the positive samples. The above three steps are repeated, and finally, N-1 predicted probability values ​​are obtained for each unlabeled sample that is suspected to be a shared rental sample.

[0172] 505. Output Results Module.

[0173] In this embodiment, the output result module is used to fuse the prediction results of the LightGBM binary classification model during training and output the probability that an unlabeled sample is likely a group sample. That is, it fuses the mean of N-1 prediction results of the LightGBM binary classification model during training and outputs the probability that an unlabeled sample is likely a group sample.

[0174] This embodiment introduces the PU Learning semi-supervised learning method into the shared rental housing identification scenario. Based on daily and monthly electricity consumption data of residents in the community provided by the power grid company, a training set is constructed by randomly combining a small number of positive samples and a large number of unlabeled samples. Effective features for identifying shared rental housing are extracted, and a binary classification model is trained using the LightGBM algorithm. Positive samples in the unlabeled samples are heuristically mined and added to the next iteration, effectively solving the problem of sample imbalance in the shared rental housing identification scenario and greatly improving the accuracy of the shared rental housing identification algorithm. This method can quickly output highly suspicious residents in shared rental housing, facilitating accurate door-to-door sampling and investigation by law enforcement officers, and effectively managing the phenomenon of shared rental housing.

[0175] The above embodiments illustrate the process of identifying shared rental housing based on electricity consumption. In real-world scenarios, multi-dimensional identification can also be performed based on various resource usage data. This scenario is described below. Please refer to... Figure 6 , Figure 6 A flowchart illustrating another method for identifying abnormal resource consumption objects provided in this application embodiment. This application embodiment includes at least the following steps:

[0176] 601. Determine the specific data of resource usage data corresponding to the target object set under different data dimensions.

[0177] In this embodiment, specialized data under different data dimensions are classified and processed, meaning that data of different categories will not interfere with each other, thereby ensuring the accuracy of data processing.

[0178] 602. Electricity consumption data.

[0179] In this embodiment, the electricity consumption data refers to the electricity consumption data under each user account, which can be statistical data at the daily, monthly, or other time granularities.

[0180] 603. Water consumption data.

[0181] In this embodiment, the water consumption data refers to the water consumption data under each user account, which can be statistical data at the daily, monthly, or other time granularities.

[0182] 604. Network consumption data.

[0183] In this embodiment, the network consumption data refers to the network resource consumption data under each broadband account, which is statistically analyzed from the network routing to avoid data separation between different tenants.

[0184] 605. Obtain the first predicted value.

[0185] In this embodiment, the first predicted value is the result of using electricity consumption data as resource usage data. Figure 3 The predicted value obtained by the identification method in the illustrated embodiment.

[0186] 606. Obtain the second predicted value.

[0187] In this embodiment, the second predicted value is obtained by using water consumption data as resource usage data. Figure 3 The predicted value obtained by the identification method in the illustrated embodiment.

[0188] 607. Obtain the third predicted value.

[0189] In this embodiment, the third predicted value is the result of executing the resource usage data using network consumption data. Figure 3 The predicted value obtained by the identification method in the illustrated embodiment.

[0190] 608. Perform weighted calculations to determine the participants in the shared rental housing scheme.

[0191] In this embodiment, since the processes of electricity consumption, water consumption, and network consumption may have different degrees of correlation with the shared rental housing, different weight values ​​can be set for the final value calculation. For example, the weighted calculation can be performed with electricity consumption: water consumption: network consumption = 0.5:0.3:0.2, thereby ensuring the accuracy of the identification of shared rental housing objects.

[0192] The following explains the real-time notification function of the rental software in the terminal. Please refer to [link / reference]. Figure 7 , Figure 7 A flowchart illustrating another method for identifying abnormal resource consumption objects provided in this application embodiment. This application embodiment includes at least the following steps:

[0193] 701. Respond to the trigger operation on the rental interface to determine the target property.

[0194] In this embodiment, the triggering operation on the rental interface is the clicking operation on the property listing, which can be during the process of viewing details or during the process of making a phone call.

[0195] 702. Based on the target property, retrieve the resource usage information of the corresponding area to identify shared rental properties.

[0196] In this embodiment, for properties in areas without marked shared rental housing, since a semi-supervised learning identification method is used, resource usage information within the corresponding area of ​​the target property is required. Figure 3 The identification process in the illustrated embodiment will not be described in detail here. For properties in areas that have already been marked as shared rental housing, the identification result can be obtained simply by traversing the area based on the markings.

[0197] 703. Display prompts for target properties on the rental interface.

[0198] In this embodiment, as Figure 8 As shown, Figure 8 This is a schematic diagram of a scenario for another method of identifying abnormal resource consumption objects provided in this application embodiment. The diagram shows that after a user clicks on a property listing, the server is triggered to call resource usage data and identify shared rental properties to obtain a prediction result. If the prediction result is a shared rental property, a prompt element A1 of the target property listing will be displayed on the rental interface, that is, prompting the user that the property listing may be a shared rental property and needs to be checked on-site, thereby ensuring the credibility of the rental software.

[0199] To better implement the above-described solutions of the embodiments of this application, related apparatus for implementing the above solutions is also provided below. Please refer to... Figure 9 , Figure 9 This is a schematic diagram of a device for identifying objects with abnormal resource consumption, provided in an embodiment of this application. The identification device 900 includes:

[0200] The acquisition unit 901 is used to acquire resource usage data of the target object set as unlabeled samples, and call the resource usage data corresponding to the objects that have been verified as having abnormal resource consumption as positive samples, wherein the number of positive samples is less than the number of unlabeled samples;

[0201] Extraction unit 902 is used to extract features from the positive sample based on at least one feature term to obtain the target feature dimension, wherein the feature term is set based on resource usage features at a preset time granularity;

[0202] Training unit 903 is used to input the unlabeled samples and the positive samples into a semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension. During the iterative training process, unlabeled samples that meet preset conditions are labeled as supplementary samples, and the supplementary samples are used to update the positive samples.

[0203] The identification unit 904 is used to acquire multiple identification feature values ​​corresponding to the unlabeled samples output by the classification model during the iterative training process, and to fuse them based on the identification feature values ​​to determine the abnormal resource consumption objects in the target object set.

[0204] Optionally, in some possible implementations of this application, the acquisition unit 901 is specifically used to determine candidate data corresponding to the target object set within the data statistics range;

[0205] The acquisition unit 901 is specifically used to divide the candidate data based on the preset time granularity to obtain granular data;

[0206] The acquisition unit 901 is specifically used to statistically analyze the resource feature items in the granular data to obtain the resource usage data corresponding to the target object set;

[0207] The acquisition unit 901 is specifically used to preprocess the resource usage data to obtain the unlabeled sample.

[0208] Optionally, in some possible implementations of this application, the acquisition unit 901 is specifically used to determine the smallest granularity in the preset time granularity;

[0209] The acquisition unit 901 is specifically used to traverse the resource usage data based on the minimum granularity to determine missing items and negative value items.

[0210] The acquisition unit 901 is specifically used to call replacement values ​​to replace the missing items and the negative value items, so as to preprocess the resource usage data to obtain the unlabeled samples.

[0211] Optionally, in some possible implementations of this application, the acquisition unit 901 is specifically used to acquire the average number in the resource usage data;

[0212] The acquisition unit 901 is specifically used to determine the outstanding items in the resource usage data that exceed the average number;

[0213] The acquisition unit 901 is specifically used to replace the value of the salient item with the mean to preprocess the resource usage data to obtain the unlabeled sample.

[0214] Optionally, in some possible implementations of this application, the acquisition unit 901 is specifically used to determine the data correspondence in the resource usage data;

[0215] The acquisition unit 901 is specifically used to extract abnormal items from the data correspondence relationship;

[0216] The acquisition unit 901 is specifically used to filter overlapping values ​​in the anomaly items in order to preprocess the resource usage data to obtain the unlabeled samples.

[0217] Optionally, in some possible implementations of this application, the extraction unit 902 is specifically used to determine the numerical features corresponding to the feature item;

[0218] The extraction unit 902 is specifically used to extract features from the positive sample based on the numerical features to obtain numerical span information;

[0219] The extraction unit 902 is specifically used to determine the target feature dimension based on the numerical span information.

[0220] Optionally, in some possible implementations of this application, the extraction unit 902 is specifically used to correlate the numerical features within a preset time range to obtain fluctuation features;

[0221] The extraction unit 902 is specifically used to extract features from the positive sample based on the fluctuation characteristics to obtain the feature fluctuation range;

[0222] The extraction unit 902 is specifically used to determine the target feature dimension based on the feature fluctuation range.

[0223] Optionally, in some possible implementations of this application, the extraction unit 902 is specifically used to determine the feature time period corresponding to the numerical feature;

[0224] The extraction unit 902 is specifically used to extract features from the positive sample based on the characteristic time period to obtain time period resource usage information;

[0225] The extraction unit 902 is specifically used to determine the target feature dimension based on the time period resource usage information.

[0226] Optionally, in some possible implementations of this application, the extraction unit 902 is specifically used to compare adjacent numerical features to obtain periodic features;

[0227] The extraction unit 902 is specifically used to analyze the positive sample based on the periodic features to obtain the feature period;

[0228] The extraction unit 902 is specifically used to determine the target feature dimension based on the feature period.

[0229] Optionally, in some possible implementations of this application, the training unit 903 is specifically used to generate a training set based on the unlabeled samples and the positive samples;

[0230] The training unit 903 is specifically used to input the training set into the semi-supervised learning framework and randomly select a portion of the unlabeled samples as negative samples.

[0231] The training unit 903 is specifically used to train a preset model based on the positive samples and the negative samples to obtain the classification model;

[0232] The training unit 903 is specifically used to identify the unlabeled samples that have not been extracted according to the classification model, so as to obtain the identification feature value corresponding to each sample in the unlabeled samples.

[0233] The training unit 903 is specifically used to filter the recognition feature values ​​based on the preset conditions, so as to extract supplementary samples from the unlabeled samples and update the positive samples based on the supplementary samples;

[0234] The training unit 903 is specifically used to repeat the random sampling process to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension.

[0235] Optionally, in some possible implementations of this application, the identification unit 904 is specifically used to determine the specific data of the resource usage data corresponding to the target object set under different data dimensions;

[0236] The identification unit 904 is specifically used to acquire multiple predicted values ​​corresponding to the specific data;

[0237] The identification unit 904 is specifically used to perform weighted calculations on the multiple predicted values ​​to obtain the target feature value;

[0238] The identification unit 904 is specifically used to determine the resource consumption abnormal objects in the target object set based on the target feature value.

[0239] By acquiring resource usage data from the target object set as unlabeled samples and calling resource usage data corresponding to objects verified as having abnormal resource consumption as positive samples, the number of positive samples is less than the number of unlabeled samples. Then, feature extraction is performed on the positive samples based on at least one feature term to obtain the target feature dimension. The feature term is set based on resource usage features at a preset time granularity. The unlabeled samples and positive samples are then input into a semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension. During this iterative training process, unlabeled samples that meet preset conditions are labeled as supplementary samples, which are used to update the positive samples. Finally, multiple identification feature values ​​corresponding to the unlabeled samples output by the classification model during the iterative training process are obtained and fused based on the identification feature values ​​to determine the objects with abnormal resource consumption in the target object set. This enables the identification of abnormal resource consumption objects with a small number of positive samples. By using multiple target feature dimensions to extract the features of abnormal resource consumption objects, the sensitivity of the classification model to the features of abnormal resource consumption objects can be guaranteed. Furthermore, during the training process, positive samples in the unlabeled samples are continuously heuristically mined and added to the next iteration, effectively solving the problem of sample imbalance in the scenario of identifying abnormal resource consumption objects and improving the accuracy of identifying abnormal resource consumption objects.

[0240] This application also provides a terminal device, such as... Figure 10 The diagram shown is a structural schematic of another terminal device provided in an embodiment of this application. For ease of explanation, only the parts related to the embodiment of this application are shown. For specific technical details not disclosed, please refer to the method section of the embodiment of this application. The terminal can be any terminal device including mobile phones, tablets, personal digital assistants (PDAs), point-of-sale (POS) terminals, in-vehicle computers, etc. Taking a mobile phone as an example:

[0241] Figure 10 This is a block diagram illustrating a portion of the structure of a mobile phone related to the terminal provided in the embodiments of this application. (Reference) Figure 10 The mobile phone includes components such as a radio frequency (RF) circuit 1010, a memory 1020, an input unit 1030, a display unit 1040, a sensor 1050, an audio circuit 1060, a wireless fidelity (WiFi) module 1070, a processor 1080, and a power supply 1090. Those skilled in the art will understand that... Figure 10 The mobile phone structure shown does not constitute a limitation on the mobile phone and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0242] The following is combined with Figure 10 A detailed introduction to each component of a mobile phone:

[0243] The RF circuit 1010 can be used for receiving and transmitting signals during information transmission or calls. Specifically, it receives downlink information from the base station and processes it with the processor 1080; additionally, it transmits uplink data to the base station. Typically, the RF circuit 1010 includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low-noise amplifier (LNA), and a duplexer. Furthermore, the RF circuit 1010 can also communicate wirelessly with networks and other devices. The aforementioned wireless communication can use any communication standard or protocol, including but not limited to Global System for Mobile Communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, and Short Message Service (SMS).

[0244] The memory 1020 can be used to store software programs and modules. The processor 1080 executes various mobile phone functions and data processing by running the software programs and modules stored in the memory 1020. The memory 1020 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 1020 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0245] The input unit 1030 can be used to receive input numerical or character information, and generate key signal inputs related to user settings and function control of the mobile phone. Specifically, the input unit 1030 may include a touch panel 1031 and other input devices 1032. The touch panel 1031, also known as a touch screen, can collect touch operations performed by the user on or near it (such as operations performed by the user using a finger, stylus, or any suitable object or accessory on or near the touch panel 1031, as well as air touch operations within a certain range on the touch panel 1031), and drive the corresponding connection devices according to a pre-set program. Optionally, the touch panel 1031 may include two parts: a touch detection device and a touch controller. The touch detection device detects the user's touch position and the signal generated by the touch operation, and transmits the signal to the touch controller; the touch controller receives touch information from the touch detection device, converts it into touch point coordinates, sends it to the processor 1080, and can receive and execute commands sent by the processor 1080. Furthermore, the touch panel 1031 can be implemented using various types of sensors, including resistive, capacitive, infrared, and surface acoustic wave sensors. In addition to the touch panel 1031, the input unit 1030 may also include other input devices 1032. Specifically, these other input devices 1032 may include, but are not limited to, one or more of the following: a physical keyboard, function keys (such as volume control buttons, power buttons, etc.), a trackball, a mouse, and a joystick.

[0246] The display unit 1040 can be used to display information input by the user or information provided to the user, as well as various menus of the mobile phone. The display unit 1040 may include a display panel 1041, which may optionally be configured as a liquid crystal display (LCD), organic light-emitting diode (OLED), or similar form. Further, a touch panel 1031 may cover the display panel 1041. When the touch panel 1031 detects a touch operation on or near it, it transmits the information to the processor 1080 to determine the type of touch event. Subsequently, the processor 1080 provides corresponding visual output on the display panel 1041 according to the type of touch event. Although in Figure 10 In this embodiment, the touch panel 1031 and the display panel 1041 are two separate components to realize the input and output functions of the mobile phone. However, in some embodiments, the touch panel 1031 and the display panel 1041 can be integrated to realize the input and output functions of the mobile phone.

[0247] The mobile phone may also include at least one sensor 1050, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor. The ambient light sensor can adjust the brightness of the display panel 1041 according to the ambient light level, and the proximity sensor can turn off the display panel 1041 and / or backlight when the phone is moved to the ear. As a type of motion sensor, an accelerometer sensor can detect the magnitude of acceleration in various directions (generally three axes). When stationary, it can detect the magnitude and direction of gravity and can be used for applications that recognize the phone's posture (such as landscape / portrait switching, related games, magnetometer posture calibration), vibration recognition-related functions (such as pedometer, taps), etc. Other sensors that may be configured in the mobile phone, such as gyroscopes, barometers, hygrometers, thermometers, and infrared sensors, will not be described in detail here.

[0248] The audio circuit 1060, speaker 1061, and microphone 1062 provide an audio interface between the user and the mobile phone. The audio circuit 1060 converts the received audio data into electrical signals and transmits them to the speaker 1061, where the speaker 1061 converts them into sound signals for output. On the other hand, the microphone 1062 converts the collected sound signals into electrical signals, which are then received by the audio circuit 1060, converted into audio data, and then processed by the processor 1080 before being transmitted via the RF circuit 1010 to, for example, another mobile phone, or the audio data can be output to the memory 1020 for further processing.

[0249] WiFi is a short-range wireless transmission technology. Through the WiFi module 1070, mobile phones can help users send and receive emails, browse web pages, and access streaming media, providing users with wireless broadband internet access. Although Figure 10 The WiFi module 1070 is shown, but it is understood that it is not an essential component of a mobile phone and can be omitted as needed without changing the essence of the invention.

[0250] The processor 1080 is the control center of the mobile phone, connecting various parts of the phone through various interfaces and lines. It executes software programs and / or modules stored in the memory 1020, and calls data stored in the memory 1020 to perform various functions and process data, thereby performing overall detection of the phone. Optionally, the processor 1080 may include one or more processing units; optionally, the processor 1080 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the aforementioned modem processor may also not be integrated into the processor 1080.

[0251] The mobile phone also includes a power supply 1090 (such as a battery) that supplies power to various components. Optionally, the power supply can be logically connected to the processor 1080 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system.

[0252] Although not shown, mobile phones may also include a camera, Bluetooth module, etc., which will not be described in detail here.

[0253] In this embodiment of the application, the processor 1080 included in the terminal also has the function of performing the various steps of the page processing method described above.

[0254] This application also provides a server; please refer to [link / reference]. Figure 11 , Figure 11 This is a schematic diagram of a server structure provided in an embodiment of this application. The server 1100 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 1122 (e.g., one or more processors) and memory 1132, and one or more storage media 1130 (e.g., one or more mass storage devices) for storing application programs 1142 or data 1144. The memory 1132 and storage media 1130 can be temporary or persistent storage. The program stored in the storage media 1130 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the server. Furthermore, the CPU 1122 may be configured to communicate with the storage media 1130 and execute the series of instruction operations in the storage media 1130 on the server 1100.

[0255] Server 1100 may also include one or more power supplies 1126, one or more wired or wireless network interfaces 1150, one or more input / output interfaces 1158, and / or one or more operating systems 1141, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0256] The steps performed by the management device in the above embodiments can be based on this Figure 11 The server structure shown.

[0257] This application also provides a computer-readable storage medium storing instructions for identifying shared rental housing. When these instructions are executed on a computer, they cause the computer to perform the aforementioned actions. Figures 3 to 8 The steps performed by the shared rental housing identification device in the method described in the illustrated embodiment.

[0258] This application also provides a computer program product that includes instructions for identifying shared rental housing. When run on a computer, it causes the computer to perform the aforementioned actions. Figures 3 to 8 The steps performed by the shared rental housing identification device in the method described in the illustrated embodiment.

[0259] This application embodiment also provides a shared rental housing identification system, which may include... Figure 9 The shared rental housing identification device described in the embodiments, or Figure 10 The terminal device in the described embodiments, or Figure 11 The server described.

[0260] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0261] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0262] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0263] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0264] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a shared rental housing identification device, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0265] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for identifying abnormal resource consumption, characterized in that, include: The resource usage data of the target object set is obtained as unlabeled samples, and the resource usage data corresponding to the verified abnormal resource consumption objects is called as positive samples. The number of positive samples is less than the number of unlabeled samples. The target object set is the user group in the corresponding area. The resource usage data is data used to reflect the daily living expenses of users. The abnormal resource consumption objects are shared rental housing objects. The positive sample is subjected to feature extraction based on at least one feature term to obtain the target feature dimension. The feature term is set based on the resource usage characteristics of the resource consumption abnormal object under a preset time granularity. The unlabeled samples and the positive samples are input into a semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension. During the iterative training process, unlabeled samples that meet the preset conditions are labeled as supplementary samples, and the supplementary samples are used to update the positive samples. The classification model outputs multiple identification feature values ​​corresponding to the unlabeled samples during the iterative training process, and performs mean fusion based on the identification feature values ​​to determine the resource consumption abnormal objects in the target object set. The step of inputting the unlabeled samples and the positive samples into a semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension includes: A training set is generated based on the unlabeled samples and the positive samples; The training set is input into the semi-supervised learning framework, and a portion of the unlabeled samples are randomly selected as negative samples. The preset model is trained based on the positive samples and the negative samples to obtain the classification model; The classification model is used to identify the unlabeled samples that were not extracted, so as to obtain the identification feature value corresponding to each sample in the unlabeled samples. The identification feature values ​​are filtered based on the preset conditions to extract supplementary samples from the unlabeled samples, and the positive samples are updated based on the supplementary samples. The random sampling process is repeated to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension.

2. The method according to claim 1, characterized in that, The acquisition of resource usage data of the target object set as unlabeled samples includes: Determine the candidate data corresponding to the target object set within the data statistics scope; The candidate data is divided based on the preset time granularity to obtain granular data; The resource feature items in the granular data are statistically analyzed to obtain the resource usage data corresponding to the target object set; The resource usage data is preprocessed to obtain the unlabeled samples.

3. The method according to claim 2, characterized in that, The preprocessing of the resource usage data to obtain the unlabeled samples includes: Determine the smallest granularity in the preset time granularity; The resource usage data is traversed based on the minimum granularity to determine missing items and negative value items; The missing and negative values ​​are replaced by replacement values ​​to preprocess the resource usage data and obtain the unlabeled samples.

4. The method according to claim 2, characterized in that, The preprocessing of the resource usage data to obtain the unlabeled samples includes: Obtain the average number from the resource usage data; Identify the prominent items in the resource usage data that exceed the mean; The value of the salient item is replaced with the mean to preprocess the resource usage data to obtain the unlabeled sample.

5. The method according to claim 2, characterized in that, The preprocessing of the resource usage data to obtain the unlabeled samples includes: Determine the data correspondence in the resource usage data; Extract outliers from the data correspondence; The overlapping values ​​in the anomalies are filtered out to preprocess the resource usage data and obtain the unlabeled samples.

6. The method according to claim 1, characterized in that, The step of extracting features from the positive samples based on at least one feature term to obtain the target feature dimension includes: Determine the numerical features corresponding to the feature terms; Based on the numerical features, feature extraction is performed on the positive samples to obtain numerical span information; The target feature dimension is determined based on the numerical span information.

7. The method according to claim 6, characterized in that, The method further includes: The numerical features within a preset time range are correlated to obtain the fluctuation characteristics; Based on the fluctuation characteristics, feature extraction is performed on the positive samples to obtain the feature fluctuation range; The target feature dimension is determined based on the range of feature fluctuations.

8. The method according to claim 6, characterized in that, The method further includes: Determine the characteristic time period corresponding to the numerical feature; Based on the characteristic time period, feature extraction is performed on the positive samples to obtain time period resource usage information; The target feature dimension is determined based on the resource usage information for the specified time period.

9. The method according to claim 6, characterized in that, The method further includes: The adjacent numerical features are compared to obtain periodic features; The positive samples are analyzed based on the periodic characteristics to obtain the characteristic period; The target feature dimension is determined based on the feature period.

10. The method according to any one of claims 1-9, characterized in that, The method further includes: Determine the specific data of resource usage data corresponding to the target object set under different data dimensions; Obtain multiple predicted values ​​corresponding to the specific data respectively; The multiple predicted values ​​are weighted and calculated to obtain the target feature value; Based on the target feature values, identify the resource consumption anomaly objects in the target object set.

11. The method according to claim 1, characterized in that, The target object set includes a set of community users, the resource usage data includes electricity consumption, and the positive samples come from an executable third-party platform used to monitor the shared rental housing objects.

12. A device for identifying objects with abnormal resource consumption, characterized in that, include: The acquisition unit is used to acquire resource usage data of the target object set as unlabeled samples and call the resource usage data corresponding to the verified abnormal resource consumption objects as positive samples. The number of positive samples is less than the number of unlabeled samples. The target object set is a user group in the corresponding area. The resource usage data is data used to reflect the daily living expenses of users. The abnormal resource consumption objects are shared rental housing objects. An extraction unit is used to extract features from the positive sample based on at least one feature term to obtain a target feature dimension, wherein the feature term is set based on resource usage features at a preset time granularity. The training unit is used to input the unlabeled samples and the positive samples into a semi-supervised learning framework to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension. During the iterative training process, unlabeled samples that meet preset conditions are labeled as supplementary samples, and the supplementary samples are used to update the positive samples. The identification unit is used to acquire multiple identification feature values ​​corresponding to the unlabeled samples output by the classification model during the iterative training process, and to perform mean fusion based on the identification feature values ​​to determine the resource consumption abnormal objects in the target object set. The training unit is specifically used to generate a training set based on the unlabeled samples and the positive samples; The training unit is specifically used to input the training set into the semi-supervised learning framework and randomly select a portion of the unlabeled samples as negative samples. The training unit is specifically used to train a preset model based on the positive samples and the negative samples to obtain the classification model; The training unit is specifically used to identify the unlabeled samples that have not been extracted according to the classification model, so as to obtain the identification feature value corresponding to each sample in the unlabeled samples. The training unit is specifically used to filter the recognition feature values ​​based on the preset conditions, so as to extract supplementary samples from the unlabeled samples and update the positive samples based on the supplementary samples; The training unit is specifically used to repeat the random sampling process to iteratively train the classification model in the semi-supervised learning framework based on the target feature dimension.

13. The apparatus according to claim 12, characterized in that, The acquisition unit is specifically used to determine the candidate data corresponding to the target object set within the data statistics range; The acquisition unit is specifically used to divide the candidate data based on the preset time granularity to obtain granular data; The acquisition unit is specifically used to statistically analyze the resource feature items in the granular data to obtain the resource usage data corresponding to the target object set. The acquisition unit is specifically used to preprocess the resource usage data to obtain the unlabeled samples.

14. The apparatus according to claim 13, characterized in that, The acquisition unit is specifically used to determine the smallest granularity in the preset time granularity; The acquisition unit is specifically used to traverse the resource usage data based on the smallest granularity to determine missing items and negative value items. The acquisition unit is specifically used to call replacement values ​​to replace the missing items and the negative value items, so as to preprocess the resource usage data to obtain the unlabeled samples.

15. The apparatus according to claim 13, characterized in that, The acquisition unit is specifically used to acquire the average number in the resource usage data; The acquisition unit is specifically used to determine the prominent items in the resource usage data that exceed the average number; The acquisition unit is specifically used to replace the value of the salient item with the mean to preprocess the resource usage data to obtain the unlabeled sample.

16. A computer device, characterized in that, The computer device includes a processor and memory: The memory is used to store program code; the processor is used to execute the method for identifying abnormal resource consumption objects according to any one of claims 1 to 11, based on the instructions in the program code.

17. A computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method for identifying abnormal resource consumption objects as described in any one of claims 1 to 11.

18. A computer program product, characterized in that, The computer program product includes instructions that, when executed on a computer device, cause the computer to perform the method for identifying abnormal resource consumption objects as described in any one of claims 1 to 11.

Images