Abnormal commodity object identification method, device and storage medium

Abstract

The embodiment of the present invention provides a method, device and storage medium for identifying abnormal commodity objects, by obtaining historical information related to multiple commodity objects to be identified from a database; determining the target feature information of each commodity object to be identified based on the historical information; obtaining the commodity objects to be identified on the convex hull boundary corresponding to the multiple commodity objects to be identified as abnormal commodity objects based on the target feature information of each commodity object to be identified and a preset convex hull model; and outputting the abnormal commodity objects to perform corresponding processing on the abnormal commodity objects. The embodiment of the present invention obtains the target feature information of each commodity object to be identified, and processes it in combination with a preset convex hull model, so that the abnormal commodity objects are located on the convex hull boundary, thereby accurately identifying the abnormal commodity objects, discovering a large number of abnormal commodities with commonalities, and does not rely on the personal experience and business understanding of the decision maker, thereby improving the accuracy of abnormal commodity object identification.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of big data, and in particular to a method, device and storage medium for identifying abnormal commodity objects. Background Art

[0002] In e-commerce websites, it is very important to mine abnormal products, such as cheating products, slow-selling products, and products with abnormal prices, and then handle the abnormal products accordingly to maintain the normal operation of the e-commerce website.

[0003] The existing methods for discovering abnormal products are mainly outlier analysis methods in the field of machine learning or single-dimensional sorting methods. There are two main outlier analysis methods: one is a clustering-based method that detects outliers by examining the relationship between objects and clusters; the other is a classification-based method that trains a classification model that can distinguish between normal data and outliers and detects outliers based on the classification model. In the single-dimensional sorting method, the decision maker usually determines the characteristics of one dimension of the product and sorts all the products based on the characteristics of this dimension. The products ranked first or last are considered abnormal products.

[0004] Among the methods for discovering abnormal products in the existing technologies, the outlier analysis method often has low accuracy and is unable to discover a large number of abnormal products with common characteristics; the single-dimensional sorting method also has low accuracy and relies heavily on the decision maker's personal experience and business understanding, which is one-sided and may lead to the omission of abnormal products. Summary of the invention

[0005] The embodiments of the present invention provide a method, a device and a storage medium for identifying abnormal commodity objects, so as to improve the accuracy in the process of mining abnormal commodities.

[0006] In a first aspect, an embodiment of the present invention provides a method for identifying abnormal commodity objects, comprising:

[0007] Acquire historical information related to multiple commodity objects to be identified from a database;

[0008] Determine target feature information of each commodity object to be identified according to the historical information;

[0009] According to the target feature information of each commodity object to be identified and the preset convex hull model, the commodity objects to be identified on the convex hull boundary corresponding to the multiple commodity objects to be identified are obtained as abnormal commodity objects;

[0010] The abnormal commodity object is outputted to perform corresponding processing on the abnormal commodity object.

[0011] Optionally, for each commodity object to be identified, initial feature information of the commodity object to be identified is extracted from the historical information;

[0012] Acquire auxiliary feature information according to the initial feature information;

[0013] The initial feature information and the auxiliary feature information are used as target feature information of the commodity object to be identified.

[0014] Optionally, the acquiring auxiliary feature information according to the initial feature information includes:

[0015] Determining a regression target for the commodity object to be identified;

[0016] Performing linear regression based on the initial feature information and the regression target to obtain an objective function between the regression target and the initial feature information;

[0017] According to the positive and negative coefficients corresponding to the initial feature information in the objective function, the initial feature information is grouped to obtain a first feature information group with positive coefficients and a second feature information group with negative coefficients;

[0018] Auxiliary feature information is acquired according to the first feature information group, the second feature information group, and the type information of the abnormal commodity object to be identified.

[0019] Optionally, acquiring the auxiliary feature information according to the first feature information group, the second feature information group, and the type information of the abnormal commodity object to be identified includes:

[0020] Constructing an initial feature vector according to the first feature information group and the second feature information group;

[0021] Determine one of the first feature information group and the second feature information group that is related to the type information of the abnormal commodity object to be identified as a target feature information group, and the other group as a constraint feature information group;

[0022] The dimension related to the constraint feature information group in the initial feature vector is set to 0 to obtain a first auxiliary feature vector; and the positive direction of the coordinate axis of each dimension related to the constraint feature information group is taken as a second auxiliary feature vector;

[0023] Accordingly, the step of obtaining the commodity objects to be identified on the convex hull boundary corresponding to the plurality of commodity objects to be identified according to the target feature information of each commodity object to be identified and the preset convex hull model includes:

[0024] The initial feature vector, the first auxiliary feature vector, the second auxiliary feature vector and the origin of the coordinate system of each commodity object to be identified are input into the preset convex hull model, the convex hulls corresponding to the multiple commodity objects to be identified are obtained, and the commodity objects to be identified on the boundary of the convex hull are obtained.

[0025] Optionally, the step of determining one of the first feature information group and the second feature information group that is related to the type information of the to-be-identified abnormal commodity object as a target feature information group and the other as a constraint feature information group includes:

[0026] If the type information of the abnormal commodity object to be identified is a first type of abnormal commodity object with an abnormally high order conversion rate, the first feature information group is used as the target feature information group, and the second feature information group is used as the constraint feature information group; or

[0027] If the type information of the abnormal commodity object to be identified is the second type of abnormal commodity object with abnormally low order conversion rate, the second feature information group is used as the target feature information group, and the first feature information group is used as the constraint feature information group.

[0028] Optionally, determining the regression target of the commodity object includes:

[0029] Obtain relevant indicators of the regression target of the commodity object and weights corresponding to the relevant indicators;

[0030] Obtaining a weighted indicator according to the relevant indicators and their corresponding weights;

[0031] The weighted index is normalized as a regression target of the commodity object.

[0032] Optionally, the relevant indicators of the regression target include at least two of sales volume information, sales revenue information, and complaint volume information.

[0033] Optionally, performing linear regression according to the initial feature information and the regression target to obtain an objective function between the regression target and the initial feature information includes:

[0034] A linear regression is performed on the initial feature information and the regression target through a preset logistic regression model to obtain an objective function between the regression target and the initial feature information.

[0035] One embodiment of the above invention has the following advantages or beneficial effects: by obtaining the target feature information of each commodity object to be identified and processing it in combination with a preset convex hull model, the abnormal commodity object is located on the convex hull boundary, so that the abnormal commodity object can be accurately identified, a large number of common abnormal commodities can be found, and it does not rely on the personal experience and business understanding of the decision maker, thereby improving the accuracy of abnormal commodity object identification. In addition, based on the different types of information of the abnormal commodity objects to be identified, different types of abnormal commodity objects can be identified by constructing different target feature information, that is, the abnormal commodities can be classified, so that different types of abnormal commodity objects can be processed accordingly.

[0036] In a second aspect, an embodiment of the present invention provides a device for identifying abnormal commodity objects, including:

[0037] An acquisition module, used to acquire historical information related to multiple commodity objects to be identified from a database;

[0038] A feature extraction module, used to determine target feature information of each commodity object to be identified according to the historical information;

[0039] An identification module, configured to obtain, according to target feature information of each commodity object to be identified and a preset convex hull model, commodity objects to be identified on the convex hull boundary corresponding to the plurality of commodity objects to be identified as abnormal commodity objects;

[0040] The output module is used to output the abnormal commodity object so as to perform corresponding processing on the abnormal commodity object.

[0041] Optionally, when determining the target feature information of each commodity object to be identified according to the historical information, the feature extraction module is used to:

[0042] For each commodity object to be identified, extracting initial feature information of the commodity object to be identified from the historical information;

[0043] Acquire auxiliary feature information according to the initial feature information;

[0044] The initial feature information and the auxiliary feature information are used as target feature information of the commodity object to be identified.

[0045] Optionally, when acquiring the auxiliary feature information according to the initial feature information, the feature extraction module is used to:

[0046] Determining a regression target for the commodity object to be identified;

[0047] Performing linear regression based on the initial feature information and the regression target to obtain an objective function between the regression target and the initial feature information;

[0048] According to the positive and negative coefficients corresponding to the initial feature information in the objective function, the initial feature information is grouped to obtain a first feature information group with positive coefficients and a second feature information group with negative coefficients;

[0049] Auxiliary feature information is acquired according to the first feature information group, the second feature information group, and the type information of the abnormal commodity object to be identified.

[0050] Optionally, when the feature extraction module acquires the auxiliary feature information according to the first feature information group, the second feature information group, and the type information of the abnormal commodity object to be identified, it is used to:

[0051] constructing an initial feature vector according to the first feature information group and the second feature information group;

[0052] Determine one of the first feature information group and the second feature information group that is related to the type information of the abnormal commodity object to be identified as a target feature information group, and the other group as a constraint feature information group;

[0053] The dimension related to the constraint feature information group in the initial feature vector is set to 0 to obtain a first auxiliary feature vector; and the positive direction of the coordinate axis of each dimension related to the constraint feature information group is taken as a second auxiliary feature vector;

[0054] Accordingly, when the identification module obtains the commodity objects to be identified on the convex hull boundary corresponding to the plurality of commodity objects to be identified according to the target feature information of each commodity object to be identified and the preset convex hull model, it is used to:

[0055] The initial feature vector, the first auxiliary feature vector, the second auxiliary feature vector and the origin of the coordinate system of each commodity object to be identified are input into the preset convex hull model, the convex hulls corresponding to the multiple commodity objects to be identified are obtained, and the commodity objects to be identified on the boundary of the convex hull are obtained.

[0056] Optionally, when determining that one of the first feature information group and the second feature information group that is related to the type information of the to-be-identified abnormal commodity object is used as the target feature information group and the other is used as the constraint feature information group, the feature extraction module is used to:

[0057] If the type information of the abnormal commodity object to be identified is a first type of abnormal commodity object with an abnormally high order conversion rate, the first feature information group is used as the target feature information group, and the second feature information group is used as the constraint feature information group; or

[0058] If the type information of the abnormal commodity object to be identified is the second type of abnormal commodity object with abnormally low order conversion rate, the second feature information group is used as the target feature information group, and the first feature information group is used as the constraint feature information group.

[0059] Optionally, when determining the regression target of the commodity object, the feature extraction module is used to:

[0060] Obtain relevant indicators of the regression target of the commodity object and weights corresponding to the relevant indicators;

[0061] Obtaining a weighted indicator according to the relevant indicators and their corresponding weights;

[0062] The weighted index is normalized as a regression target of the commodity object.

[0063] Optionally, the relevant indicators of the regression target include at least two of sales volume information, sales revenue information, and complaint volume information.

[0064] Optionally, when performing linear regression according to the initial feature information and the regression target to obtain an objective function between the regression target and the initial feature information, the feature extraction module is used to:

[0065] A linear regression is performed on the initial feature information and the regression target through a preset logistic regression model to obtain an objective function between the regression target and the initial feature information.

[0066] In a third aspect, an embodiment of the present invention provides an electronic device, including: a memory, a processor;

[0067] Wherein, the memory is used to store the processor executable instructions;

[0068] The processor is configured to:

[0069] Acquire historical information related to multiple commodity objects to be identified from a database;

[0070] Determine target feature information of each commodity object to be identified according to the historical information;

[0071] According to the target feature information of each commodity object to be identified and the preset convex hull model, the commodity objects to be identified on the convex hull boundary corresponding to the multiple commodity objects to be identified are obtained as abnormal commodity objects;

[0072] The abnormal commodity object is outputted to perform corresponding processing on the abnormal commodity object.

[0073] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method described in the first aspect.

[0074] The method, device and storage medium for identifying abnormal commodity objects provided by the embodiments of the present invention obtain historical information related to multiple commodity objects to be identified from a database; determine the target feature information of each commodity object to be identified based on the historical information; obtain the commodity objects to be identified on the convex hull boundary corresponding to the multiple commodity objects to be identified as abnormal commodity objects based on the target feature information of each commodity object to be identified and a preset convex hull model; output the abnormal commodity objects to perform corresponding processing on the abnormal commodity objects. The embodiments of the present invention obtain the target feature information of each commodity object to be identified, and process it in combination with a preset convex hull model so that the abnormal commodity objects are located on the convex hull boundary, thereby accurately identifying abnormal commodity objects, discovering a large number of abnormal commodities with commonalities, and not relying on the personal experience and business understanding of the decision maker, thereby improving the accuracy of identifying abnormal commodity objects. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0076] Figure 1 A schematic diagram of an application scenario of the method for identifying abnormal commodity objects provided by an embodiment of the present invention;

[0077] Figure 2 A flow chart of a method for identifying abnormal commodity objects provided by an embodiment of the present invention;

[0078] Figure 3 A flow chart of a method for identifying abnormal commodity objects provided by another embodiment of the present invention;

[0079] Figure 4 A flow chart of a method for identifying abnormal commodity objects provided by another embodiment of the present invention;

[0080] Figure 5 A schematic diagram of a coordinate system for an example of a method for identifying abnormal commodity objects provided by an embodiment of the present invention;

[0081] Figure 6 A schematic diagram of a convex hull formed by a convex hull model in a method for identifying abnormal commodity objects provided by an embodiment of the present invention;

[0082] Figure 7 A schematic diagram of a convex hull formed by a convex hull model in a method for identifying abnormal commodity objects provided by another embodiment of the present invention;

[0083] Figure 8 A structural diagram of a device for identifying abnormal commodity objects provided by an embodiment of the present invention;

[0084] Fig. 9 A structural diagram of an electronic device for executing a method for identifying abnormal commodity objects provided by an embodiment of the present invention.

[0085] The above drawings show clear embodiments of the present disclosure, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present disclosure in any way, but to illustrate the concepts of the present disclosure to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0086] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0087] First, the terms involved in the embodiments of the present invention are explained:

[0088] Convex hull model: refers to a real vector space V. For a given set X, the intersection S of all convex sets containing X is called the convex hull of X. The convex hull of X can be constructed using the convex combination of all points (X1,...Xn) in X. The convex hull model is an algorithm used to discover the convex hull;

[0089] Logistic regression: logistic regression, also known as logistic regression analysis, is a generalized linear regression analysis model, commonly used in data mining, automatic disease diagnosis, economic forecasting and other fields.

[0090] In the prior art, the discovery of abnormal products is mainly based on outlier analysis methods in the field of machine learning or single-dimensional sorting methods.

[0091] Among them, there are two main methods for outlier analysis. One is a clustering-based method, which detects outliers by examining the relationship between objects and clusters. In other words, an outlier is an object that belongs to a small sparse cluster or does not belong to any cluster. For classification-based methods, if there are class labels in the training data, it can be regarded as a classification problem. The idea of ​​this problem is generally to train a classification model that can distinguish between "normal data" and outliers. Usually a one-class model is used, that is, a classifier that only describes the normal class is constructed, so that samples that do not belong to the normal class are outliers. Only using the normal class to detect outliers can detect new outliers that are not close to the outliers in the training set; in this way, when a new outlier comes in, as long as it is within the decision boundary of the normal class, it is a normal point, and outside the decision boundary, it is an outlier. The construction of the decision boundary can refer to the support vector machine (SVM).

[0092] Among the outlier analysis methods, clustering-based methods are often less accurate and cannot detect a large number of common cheating or abnormal products. In addition, they cannot classify the abnormal products found, such as cheating products and slow-moving products. In e-commerce websites, different types of cheating products usually need to be handled differently. For example, cheating products need to be punished, and slow-moving products need to find out the reasons for their slow sales and then deal with or support them. Classification-based methods can find abnormal products when feature engineering is perfect, but classification methods are particularly dependent on sample annotation and cannot find newly emerging abnormal products.

[0093] The method based on single-dimensional sorting usually requires the decision maker to determine the characteristics of one dimension of the product and sort all the products based on the characteristics of this dimension. The products ranked first or last are considered abnormal products. The method based on single-dimensional sorting relies heavily on the personal experience and business understanding of the decision maker. It considers too few dimensions and is one-sided, which will lead to the omission of abnormal products. In particular, for the complex e-commerce environment, only using a single dimension or a few dimensions cannot fully consider various situations and accurately discover abnormal products.

[0094] In view of the above technical problems of the prior art, the embodiment of the present invention takes into account the complex conditions of e-commerce websites, considers multi-dimensional target feature information for multiple commodity objects to be identified, and proposes a new method for identifying abnormal commodity objects using the convex hull model. By obtaining the target feature information of each commodity object to be identified and processing it in combination with a preset convex hull model, the abnormal commodity object is located on the convex hull boundary, so that the abnormal commodity object can be accurately identified, and a large number of common abnormal commodities can be found. In addition, the convex hull model does not rely on the annotation of samples, and can also find newly emerging abnormal commodities, nor does it rely on the personal experience and business understanding of decision makers, thereby improving the accuracy of abnormal commodity object identification. In addition, based on the different types of information of the abnormal commodity objects to be identified, different types of abnormal commodity objects can be identified by constructing different target feature information, that is, the abnormal commodities can be classified, so that different types of abnormal commodity objects can be processed accordingly.

[0095] An optional application scenario of the embodiment of the present invention is as follows Figure 1 As shown, the scenario includes an execution server 10, a database 11, and a terminal device 12, wherein the database 11 may be a database of an e-commerce website, and the server 10 may obtain historical information related to a plurality of commodity objects to be identified from the database 11, and then obtain target feature information of each commodity object to be identified, identify abnormal commodity objects according to the target feature information of each commodity object to be identified and a preset convex hull model, and output the abnormal commodity objects to the terminal device 12, so that the decision maker on the terminal device 12 side may perform corresponding processing on the abnormal commodity objects. Optionally, the server 10 may be a server, or a server cluster consisting of a plurality of servers, or a cloud computing service center.

[0096] The following specific embodiments are used to describe in detail the technical solutions of the present invention and how the technical solutions of the present application solve the above-mentioned technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present invention will be described below in conjunction with the accompanying drawings.

[0097] Figure 2 This is a flow chart of a method for identifying abnormal commodity objects provided by an embodiment of the present invention. This embodiment provides a method for identifying abnormal commodity objects, and the execution subject may be Figure 1 The specific steps of the method are as follows:

[0098] S201: Acquire historical information related to a plurality of commodity objects to be identified from a database.

[0099] In this embodiment, there may be abnormal commodity objects among multiple commodity objects on the e-commerce website, or there may also be abnormal commodity objects among multiple commodity objects to be identified in other application scenarios, wherein the abnormal commodity objects may be commodities with abnormally high order conversion rates compared to other commodity objects, such as cheating commodities, and cheating commodities may be commodities with abnormally high order volumes but abnormally low exposure volumes, prices, etc. compared to other commodity objects, or commodities with abnormally low order conversion rates compared to other commodity objects, such as unsalable commodities. When it is necessary to find abnormal commodity objects from multiple commodity objects on the e-commerce website, or when it is necessary to identify abnormal commodity objects from multiple commodity objects to be identified in other scenarios, historical information related to the commodity objects to be identified may be obtained from the database, such as historical order information, attention collection information, evaluation information, click volume information, etc., and of course, this embodiment is not limited to the above historical information.

[0100] S202: Determine target feature information of each commodity object to be identified according to the historical information.

[0101] In this embodiment, the target feature information of each commodity object to be identified can be determined from the historical information related to multiple commodity objects to be identified, wherein the target feature information may include initial feature information directly extracted from the historical information, such as commodity exposure conversion rate, commodity click volume, commodity order volume, commodity sales volume, number of commodity good reviews, number of commodity attention collections, commodity user repurchase rate, sales volume of the store where the commodity is located, store rating, store order volume, store sales volume, click-through rate of the category where the commodity is located, exposure conversion rate, category order volume, number of category good reviews, category sales volume, the proportion of commodity to category click volume, the proportion of commodity to category sales volume, the proportion of commodity to category sales volume, commodity exposure, commodity return rate, etc. It should be noted that the initial feature information can try to select the interactive features between the commodity and the consumer; in addition, the target feature information may also include auxiliary feature information obtained based on the initial feature information, that is, feature information obtained after a certain processing of the initial feature information. Of course, the target feature information in this embodiment may not be limited to the above-mentioned feature information, and will not be repeated here one by one.

[0102] S203 , according to the target feature information of each commodity object to be identified and a preset convex hull model, obtaining commodity objects to be identified on the convex hull boundary corresponding to the plurality of commodity objects to be identified as abnormal commodity objects.

[0103] In this embodiment, the target feature information of each commodity object to be identified is input into a preset convex hull model to obtain the convex hulls corresponding to the multiple commodity objects to be identified, wherein the convex hull is the minimum convex set containing the target feature information of all commodity objects to be identified in the vector space. In this embodiment, it is considered that if one or some of the target feature information of a commodity object to be identified is significantly higher or lower than that of other commodity objects, then its target feature information is located on the boundary of the convex hull in the vector space (especially the corner points of the convex hull boundary), and thus the abnormal commodity object can be obtained according to the convex hull boundary.

[0104] Of course, the commodity objects to be identified near the convex hull boundary may also be taken as abnormal commodity objects, for example, commodity objects to be identified whose distance from the corner point of the convex hull boundary is within a predetermined range.

[0105] S204: Output the abnormal commodity object to perform corresponding processing on the abnormal commodity object.

[0106] In this embodiment, after the abnormal commodity object is determined, the abnormal commodity object can be output, for example, displayed on a display device, or sent to a terminal device. Specifically, relevant information of the abnormal commodity object can be output, and then corresponding processing methods can be taken for the abnormal commodities. For example, cheating commodities can be rectified, removed from the shelves, or punished, and the reasons for the slow sales of slow-selling commodities can be found to deal with or support them.

[0107] The method for identifying abnormal commodity objects provided in this embodiment obtains historical information related to multiple commodity objects to be identified from a database; determines the target feature information of each commodity object to be identified based on the historical information; obtains the commodity objects to be identified on the convex hull boundary corresponding to the multiple commodity objects to be identified as abnormal commodity objects based on the target feature information of each commodity object to be identified and a preset convex hull model; and outputs the abnormal commodity objects to perform corresponding processing on the abnormal commodity objects. This embodiment obtains the target feature information of each commodity object to be identified, and processes it in combination with a preset convex hull model so that the abnormal commodity objects are located on the convex hull boundary, thereby accurately identifying abnormal commodity objects, discovering a large number of abnormal commodities with commonalities, and does not rely on the personal experience and business understanding of the decision maker, thereby improving the accuracy of identifying abnormal commodity objects.

[0108] Based on any of the above embodiments, Figure 3 As shown, the step of determining target feature information of each commodity object to be identified according to the historical information in S202 includes:

[0109] S301 : For each commodity object to be identified, extract initial feature information of the commodity object to be identified from the historical information.

[0110] In this embodiment, the target feature information of the commodity object to be identified includes the initial feature information and auxiliary feature information of the commodity object to be identified, wherein the initial feature information can be extracted from the historical information of the commodity object to be identified, and the initial feature information includes at least two items, and the initial feature information may include but is not limited to: commodity exposure conversion rate, commodity click volume, commodity order volume, commodity sales, number of product positive reviews, number of product attention and collection, commodity user repurchase rate, sales volume of the store where the commodity is located, store rating, store order volume, store sales, click-through rate of the category to which the commodity belongs, exposure conversion rate, category order volume, number of category positive reviews, category sales, the proportion of commodity click volume in the category, the proportion of commodity sales volume in the category, the proportion of commodity sales volume in the category, commodity exposure, commodity return rate, etc. It should be noted that the initial feature information can try to select the interaction characteristics between the commodity and the consumer.

[0111] S302: Acquire auxiliary feature information according to the initial feature information.

[0112] In this embodiment, for a certain commodity object, after obtaining its initial feature information, auxiliary feature information can be obtained after a certain processing of the initial feature information, so as to avoid misjudgment when only the initial feature information is input into the preset convex hull model, that is, some normal commodity objects are identified as commodity objects on the convex hull boundary, and thus are considered to be abnormal commodity objects.

[0113] Specifically, Figure 4 As shown, the step S302 of acquiring auxiliary feature information according to the initial feature information may include:

[0114] S3021: Determine the regression target of the commodity object to be identified.

[0115] In this embodiment, a regression target of a commodity object to be identified is first determined, so as to construct an objective function between initial feature information and the regression target through linear regression, wherein the regression target can be a specific indicator of the commodity object to be identified, such as sales volume information, sales amount information, complaint volume information, etc. Of course, the regression target can also be the weighted sum of multiple specific indicators, and the regression target can be selected according to needs.

[0116] In an optional embodiment, the relevant indicators of the regression target of the commodity object and the weights corresponding to the relevant indicators can be obtained; the weighted indicators can be obtained according to the relevant indicators and their corresponding weights; the weighted indicators are normalized as the regression target of the commodity object. In this embodiment, linear weighting is adopted, and corresponding weights can be assigned according to the attention or importance of the relevant indicators, and then linear combinations are performed to obtain the optimal regression target. In addition, the relevant indicators can include but are not limited to at least two of sales volume information, sales information, and complaint volume information.

[0117] S3022. Perform linear regression based on the initial feature information and the regression target to obtain an objective function between the regression target and the initial feature information.

[0118] In this embodiment, after the regression target is obtained, a linear regression processing process can be performed. Optionally, a linear regression can be performed on the initial feature information and the regression target through a preset logistic regression model to obtain an objective function between the regression target and the initial feature information.

[0119] S3023. Group the initial feature information according to the positive or negative coefficients corresponding to the initial feature information in the objective function to obtain a first feature information group with positive coefficients and a second feature information group with negative coefficients.

[0120] In this embodiment, since the objective function is a linear function, that is, each initial feature information corresponds to a coefficient, for a commodity object x to be identified, the initial feature information can be grouped according to the positive or negative coefficient corresponding to each initial feature information, and the first feature information group is the one with a positive coefficient, which is recorded as F x1 , which has a positive impact on the regression target, and the coefficient is negative for the second feature information group, denoted by F x2 , which has a negative impact on the regression target.

[0121] S3024. Obtain auxiliary feature information according to the first feature information group, the second feature information group, and the type information of the abnormal commodity object to be identified.

[0122] In this embodiment, the type information of the abnormal commodity object to be identified can be a first type of abnormal commodity object with an abnormally high order conversion rate, such as a cheating commodity, or a second type of abnormal commodity object with an abnormally low order conversion rate, such as a slow-moving commodity; when the first type of abnormal commodity object needs to be identified, the type information of the abnormal commodity object to be identified is determined to be a first type of abnormal commodity object, and when the second type of abnormal commodity object needs to be identified, the type information of the abnormal commodity object to be identified is determined to be a second type of abnormal commodity object. In this embodiment, different auxiliary feature information can be constructed for different types of abnormal commodity object information to be identified.

[0123] Specifically, an initial feature vector (F x1 , F x2 ); determine one of the first feature information group and the second feature information group that is related to the type information of the abnormal commodity object to be identified as the target feature information group, and the other group as the constraint feature information group; set the dimension related to the constraint feature information group in the initial feature vector to 0 to obtain the first auxiliary feature vector; and take the positive direction of the coordinate axis of each dimension related to the constraint feature information group as the second auxiliary feature vector.

[0124] For example, in one example, if the type information of the abnormal commodity object to be identified is the first type of abnormal commodity object with an abnormally high order conversion rate, the first feature information group F x1 As the target feature information group, the second feature information group F x2 As the constraint feature information group; then at this time, the initial feature vector (F x1 , F x2 ) x2 The relevant dimension is set to 0, and the first auxiliary feature vector (F x1 , 0); and take the constraint feature information group F x2 The positive direction of the coordinate axis of each dimension is used as the second auxiliary feature vector, which can be selected as F x2 The point of infinity in the positive direction of the coordinate axis of each dimension is related.

[0125] Specifically, suppose there are s commodity objects to be identified. For commodity object x, its initial feature vector (F x1 , F x2 ), whose dimension is m+n, where F x1 is an n-dimensional vector (n is the number of features in the first feature information group), F x2 is an m-dimensional vector (m is the number of features in the first feature information group), and for s commodity objects to be identified, there are s initial feature vectors; the first auxiliary feature vector (F x1 , 0) is also m+n in dimension. For s commodities to be identified, there are s initial feature vectors. The dimension of the second auxiliary feature vector is F x2 The positive direction of the coordinate axis of each related dimension is also m+n, and there are m in total.

[0126] In another example, if the type information of the abnormal commodity object to be identified is the second type of abnormal commodity object with abnormally low order conversion rate, the second feature information group F x2 As the target feature information group, the first feature information group F x1As the constraint feature information group; then at this time, the initial feature vector (F x1 , F x2 ) x1 The relevant dimension is set to 0, and the first auxiliary eigenvector (0, F x2 ); and take the constraint feature information group F x1 The positive direction of the coordinate axis of each dimension is used as the second auxiliary feature vector, which can be selected as F x1 The point of infinity in the positive direction of the coordinate axis of each dimension is related.

[0127] Specifically, suppose there are s commodity objects to be identified. For commodity object x, its initial feature vector (F x1 , F x2 ), whose dimension is m+n, where F x1 is an n-dimensional vector (n is the number of features in the first feature information group), F x2 is an m-dimensional vector (m is the number of features in the first feature information group). For s commodity objects to be identified, there are s initial feature vectors; the first auxiliary feature vector (0, F x2 ) dimension is also m+n. For s commodity objects to be identified, there are s initial feature vectors; the second auxiliary feature vector is F x1 The positive direction of the coordinate axis of each related dimension is also m+n, and the total number is n.

[0128] S303: Use the initial feature information and the auxiliary feature information as target feature information of the commodity object to be identified.

[0129] In this embodiment, the initial feature information and the auxiliary feature information are used together as the target feature information of the commodity object to be identified, that is, the target feature information of the commodity object to be identified may specifically include the above-mentioned initial feature vector, the first auxiliary feature vector and the second auxiliary feature vector. For the above-mentioned example of identifying the first type of abnormal commodity objects, there are s initial feature vectors, s first auxiliary feature vectors and m second auxiliary feature vectors for s commodity objects to be identified, for a total of 2s+m; for the above-mentioned example of identifying the second type of abnormal commodity objects, there are s initial feature vectors, s first auxiliary feature vectors and n second auxiliary feature vectors for s commodity objects to be identified, for a total of 2s+n.

[0130] Correspondingly, when obtaining the commodity objects to be identified on the convex hull boundary corresponding to the plurality of commodity objects to be identified according to the target feature information of each commodity object to be identified and the preset convex hull model, the initial feature vector, the first auxiliary feature vector, the second auxiliary feature vector and the origin of the coordinate system of each commodity object to be identified are input into the preset convex hull model, the convex hull corresponding to the plurality of commodity objects to be identified is obtained, and the commodity objects to be identified on the convex hull boundary are obtained. Of course, the input into the preset convex hull model may also include the origin of the coordinate system.

[0131] As an example, for ease of understanding, assume that the initial feature information only includes two dimensions: order volume and exposure volume, such as Figure 5 As shown, the target feature information of the commodity object to be identified can be represented as a point in the coordinate system. Since this embodiment believes that commodities that are significantly higher or lower than other commodities in certain features, certain feature combinations, and certain feature combination ratios are likely to be abnormal commodities, A and B are likely to be abnormal commodities, because A's order conversion rate is particularly high, that is, the order volume is particularly high, and the exposure is particularly low, so it may be a cheating commodity. B's order conversion rate is particularly low, that is, the order volume is particularly low, and the exposure is particularly high, so it may be a slow-moving commodity. In the two-dimensional case, such abnormal commodities can be easily found through intuitive methods. When the number of features increases, it is necessary to find them through the method of this embodiment. The following is based on Figure 5 The two-dimensional example shown is for illustration purposes only.

[0132] For the case of identifying the first type of abnormal commodity objects, since the positive direction of the coordinate axis of each dimension related to the second feature information group, that is, the positive direction of the coordinate axis of the exposure amount, is taken into account, the convex hull formed after being processed by the preset convex hull model is as follows: Figure 6 As shown, the upper and lower rays intersect at positive infinity to form a closed convex hull. At this time, only the product object A is located on the boundary of the convex hull. In this embodiment, the product object A located on the boundary of the convex hull and the products close to the product A (for example, within a preset distance range) are regarded as the first type of abnormal product objects, that is, abnormal product objects with abnormally high order conversion rates such as cheating products.

[0133] As another example, based on the above example, for the case of identifying the second type of abnormal commodity objects, since the positive direction of the coordinate axis of each dimension related to the first feature information group, that is, the positive direction of the coordinate axis of the order quantity, is considered, the convex hull formed after being processed by the preset convex hull model is as follows: Figure 7As shown, the upper and lower rays intersect at positive infinity to form a closed convex hull. At this time, only the product object B is located on the boundary of the convex hull. In this embodiment, the product object B located on the boundary of the convex hull and the products close to the product B (for example, the distance is within a preset distance range) are regarded as the first type of abnormal product objects, that is, abnormal product objects with abnormally high order conversion rates such as cheating products.

[0134] The method for identifying abnormal commodity objects provided in the above embodiment obtains the target feature information of each commodity object to be identified, and processes it in combination with a preset convex hull model so that the abnormal commodity object is located on the convex hull boundary, thereby accurately identifying the abnormal commodity object, discovering a large number of abnormal commodities with common characteristics, and not relying on the personal experience and business understanding of the decision maker, thereby improving the accuracy of identifying abnormal commodity objects. In addition, based on the different types of information of the abnormal commodity objects to be identified, different target feature information can be constructed to identify different types of abnormal commodity objects, that is, the abnormal commodities can be classified, so that different types of abnormal commodity objects can be processed accordingly.

[0135] Figure 8 The structure diagram of the device for identifying abnormal commodity objects provided by the embodiment of the present invention. The device for identifying abnormal commodity objects provided by this embodiment can execute the processing flow provided by the method for identifying abnormal commodity objects, such as Figure 8 As shown, the abnormal commodity object identification device 80 includes an acquisition module 81 , a feature extraction module 82 , an identification module 83 , and an output module 84 .

[0136] An acquisition module 81 is used to acquire historical information related to a plurality of commodity objects to be identified from a database;

[0137] A feature extraction module 82 is used to determine target feature information of each commodity object to be identified according to the historical information;

[0138] The identification module 83 is used to obtain the commodity objects to be identified on the convex hull boundary corresponding to the plurality of commodity objects to be identified as abnormal commodity objects according to the target feature information of each commodity object to be identified and the preset convex hull model;

[0139] The output module 84 is used to output the abnormal commodity object so as to perform corresponding processing on the abnormal commodity object.

[0140] On the basis of the above embodiment, when determining the target feature information of each commodity object to be identified according to the historical information, the feature extraction module 82 is used to:

[0141] For each commodity object to be identified, extracting initial feature information of the commodity object to be identified from the historical information;

[0142] Acquire auxiliary feature information according to the initial feature information;

[0143] The initial feature information and the auxiliary feature information are used as target feature information of the commodity object to be identified.

[0144] Based on any of the above embodiments, when the feature extraction module 82 obtains the auxiliary feature information according to the initial feature information, it is used to:

[0145] Determining a regression target for the commodity object to be identified;

[0146] Performing linear regression based on the initial feature information and the regression target to obtain an objective function between the regression target and the initial feature information;

[0147] According to the positive and negative coefficients corresponding to the initial feature information in the objective function, the initial feature information is grouped to obtain a first feature information group with positive coefficients and a second feature information group with negative coefficients;

[0148] Auxiliary feature information is acquired according to the first feature information group, the second feature information group, and the type information of the abnormal commodity object to be identified.

[0149] Based on any of the above embodiments, when the feature extraction module 82 obtains the auxiliary feature information according to the first feature information group, the second feature information group, and the type information of the abnormal commodity object to be identified, it is used to:

[0150] constructing an initial feature vector according to the first feature information group and the second feature information group;

[0151] Determine one of the first feature information group and the second feature information group that is related to the type information of the abnormal commodity object to be identified as a target feature information group, and the other group as a constraint feature information group;

[0152] The dimension related to the constraint feature information group in the initial feature vector is set to 0 to obtain a first auxiliary feature vector; and the positive direction of the coordinate axis of each dimension related to the constraint feature information group is taken as a second auxiliary feature vector;

[0153] Accordingly, when the identification module 83 obtains the commodity objects to be identified on the convex hull boundary corresponding to the plurality of commodity objects to be identified according to the target feature information of each commodity object to be identified and the preset convex hull model, it is used to:

[0154] The initial feature vector, the first auxiliary feature vector, the second auxiliary feature vector and the origin of the coordinate system of each commodity object to be identified are input into the preset convex hull model, the convex hulls corresponding to the multiple commodity objects to be identified are obtained, and the commodity objects to be identified on the boundary of the convex hull are obtained.

[0155] Based on any of the above embodiments, when determining one of the first feature information group and the second feature information group that is related to the type information of the abnormal commodity object to be identified as the target feature information group and the other as the constraint feature information group, the feature extraction module 82 is used to:

[0156] If the type information of the abnormal commodity object to be identified is a first type of abnormal commodity object with an abnormally high order conversion rate, the first feature information group is used as the target feature information group, and the second feature information group is used as the constraint feature information group; or

[0157] If the type information of the abnormal commodity object to be identified is the second type of abnormal commodity object with abnormally low order conversion rate, the second feature information group is used as the target feature information group, and the first feature information group is used as the constraint feature information group.

[0158] Based on any of the above embodiments, when determining the regression target of the commodity object, the feature extraction module 82 is used to:

[0159] Obtain relevant indicators of the regression target of the commodity object and weights corresponding to the relevant indicators;

[0160] Obtaining a weighted indicator according to the relevant indicators and their corresponding weights;

[0161] The weighted index is normalized as a regression target of the commodity object.

[0162] Based on any of the above embodiments, the relevant indicators of the regression target include at least two of sales volume information, sales amount information, and complaint volume information.

[0163] Based on any of the above embodiments, when the feature extraction module 82 performs linear regression according to the initial feature information and the regression target to obtain the objective function between the regression target and the initial feature information, it is used to:

[0164] A linear regression is performed on the initial feature information and the regression target through a preset logistic regression model to obtain an objective function between the regression target and the initial feature information.

[0165] The abnormal commodity object identification device provided in the embodiment of the present invention can be specifically used to perform the above Figure 2-4The specific functions of the provided method embodiment will not be repeated here.

[0166] The device for identifying abnormal commodity objects provided by the embodiment of the present invention obtains historical information related to multiple commodity objects to be identified from a database; determines the target feature information of each commodity object to be identified based on the historical information; obtains the commodity objects to be identified on the convex hull boundary corresponding to the multiple commodity objects to be identified as abnormal commodity objects based on the target feature information of each commodity object to be identified and a preset convex hull model; and outputs the abnormal commodity objects to perform corresponding processing on the abnormal commodity objects. This embodiment obtains the target feature information of each commodity object to be identified, and processes it in combination with a preset convex hull model so that the abnormal commodity objects are located on the convex hull boundary, thereby accurately identifying abnormal commodity objects, discovering a large number of abnormal commodities with commonalities, and does not rely on the personal experience and business understanding of the decision maker, thereby improving the accuracy of abnormal commodity object identification.

[0167] Fig. 9 The electronic device provided by the embodiment of the present invention can execute the processing flow provided by the embodiment of the method for identifying abnormal commodity objects, such as Fig. 9 As shown, the electronic device 90 includes a memory 91, a processor 92, a computer program and a communication interface 93; wherein the computer program is stored in the memory 91 and is configured so that the processor 92 executes the method for identifying abnormal commodity objects described in the above embodiment.

[0168] Fig. 9 The electronic device of the illustrated embodiment can be used to execute the technical solution of the above-mentioned method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0169] In addition, this embodiment further provides a computer-readable storage medium on which a computer program is stored. The computer program is executed by a processor to implement the method for identifying abnormal commodity objects described in the above embodiment.

[0170] In the several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0171] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0172] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0173] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0174] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0175] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are to be regarded as exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0176] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for identifying abnormal commodity objects, characterized in that: include: Acquire historical information related to multiple commodity objects to be identified from a database; Determine target feature information of each commodity object to be identified according to the historical information; The target feature information includes initial feature information, or the target feature information includes the initial feature information and auxiliary feature information; the initial feature information is feature information related to the interaction between the user and the commodity determined from the historical information; the auxiliary feature information is obtained based on the initial feature information and the type information of the abnormal commodity object to be identified; According to the target feature information of each commodity object to be identified and the preset convex hull model, the commodity objects to be identified on the convex hull boundary corresponding to the multiple commodity objects to be identified are obtained as abnormal commodity objects; The abnormal commodity object is outputted to perform corresponding processing on the abnormal commodity object.

2. The method according to claim 1, characterized in that The step of determining target feature information of each commodity object to be identified according to the historical information includes: For each commodity object to be identified, extracting initial feature information of the commodity object to be identified from the historical information; Acquire auxiliary feature information according to the initial feature information; The initial feature information and the auxiliary feature information are used as target feature information of the commodity object to be identified.

3. The method according to claim 2, characterized in that The acquiring of auxiliary feature information according to the initial feature information comprises: Determining a regression target for the commodity object to be identified; Performing linear regression based on the initial feature information and the regression target to obtain an objective function between the regression target and the initial feature information; According to the positive and negative coefficients corresponding to the initial feature information in the objective function, the initial feature information is grouped to obtain a first feature information group with positive coefficients and a second feature information group with negative coefficients; Auxiliary feature information is acquired according to the first feature information group, the second feature information group, and the type information of the abnormal commodity object to be identified.

4. The method according to claim 3, characterized in that: The acquiring the auxiliary feature information according to the first feature information group, the second feature information group, and the type information of the abnormal commodity object to be identified includes: constructing an initial feature vector according to the first feature information group and the second feature information group; Determine one of the first feature information group and the second feature information group that is related to the type information of the abnormal commodity object to be identified as a target feature information group, and the other group as a constraint feature information group; The dimension related to the constraint feature information group in the initial feature vector is set to 0 to obtain a first auxiliary feature vector; and the positive direction of the coordinate axis of each dimension related to the constraint feature information group is taken as a second auxiliary feature vector; Accordingly, the step of obtaining the commodity objects to be identified on the convex hull boundary corresponding to the plurality of commodity objects to be identified according to the target feature information of each commodity object to be identified and the preset convex hull model includes: The initial feature vector, the first auxiliary feature vector, the second auxiliary feature vector and the origin of the coordinate system of each commodity object to be identified are input into the preset convex hull model, the convex hulls corresponding to the multiple commodity objects to be identified are obtained, and the commodity objects to be identified on the boundary of the convex hull are obtained.

5. The method according to claim 4, characterized in that The step of determining one of the first feature information group and the second feature information group that is related to the type information of the abnormal commodity object to be identified as a target feature information group and the other as a constraint feature information group includes: If the type information of the abnormal commodity object to be identified is a first type of abnormal commodity object with an abnormally high order conversion rate, the first feature information group is used as the target feature information group, and the second feature information group is used as the constraint feature information group; or If the type information of the abnormal commodity object to be identified is the second type of abnormal commodity object with abnormally low order conversion rate, the second feature information group is used as the target feature information group, and the first feature information group is used as the constraint feature information group.

6. The method according to any one of claims 3 to 5, characterized in that: The step of determining the regression target of the commodity object includes: Obtain relevant indicators of the regression target of the commodity object and weights corresponding to the relevant indicators; Obtaining a weighted indicator according to the relevant indicators and their corresponding weights; The weighted index is normalized as a regression target of the commodity object.

7. The method according to claim 6, characterized in that The relevant indicators of the regression target include at least two of sales volume information, sales revenue information, and complaint volume information.

8. The method according to any one of claims 3 to 5, characterized in that: The performing linear regression according to the initial feature information and the regression target to obtain an objective function between the regression target and the initial feature information includes: A linear regression is performed on the initial feature information and the regression target through a preset logistic regression model to obtain an objective function between the regression target and the initial feature information.

9. A device for identifying abnormal commodity objects, characterized in that: include: An acquisition module, used to acquire historical information related to multiple commodity objects to be identified from a database; A feature extraction module, configured to determine target feature information of each commodity object to be identified according to the historical information; the target feature information includes initial feature information, or the target feature information includes the initial feature information and auxiliary feature information; the initial feature information is feature information related to the interaction between the user and the commodity determined from the historical information; the auxiliary feature information is obtained based on the initial feature information and the type information of the abnormal commodity object to be identified; An identification module, configured to obtain, according to target feature information of each commodity object to be identified and a preset convex hull model, commodity objects to be identified on the convex hull boundary corresponding to the plurality of commodity objects to be identified as abnormal commodity objects; The output module is used to output the abnormal commodity object so as to perform corresponding processing on the abnormal commodity object.

10. The device according to claim 9, characterized in that When determining the target feature information of each commodity object to be identified according to the historical information, the feature extraction module is used to: For each commodity object to be identified, extracting initial feature information of the commodity object to be identified from the historical information; Acquire auxiliary feature information according to the initial feature information; The initial feature information and the auxiliary feature information are used as target feature information of the commodity object to be identified.

11. The device according to claim 10, characterized in that When the feature extraction module obtains the auxiliary feature information according to the initial feature information, it is used to: Determining a regression target for the commodity object to be identified; Performing linear regression based on the initial feature information and the regression target to obtain an objective function between the regression target and the initial feature information; According to the positive and negative coefficients corresponding to the initial feature information in the objective function, the initial feature information is grouped to obtain a first feature information group with positive coefficients and a second feature information group with negative coefficients; Auxiliary feature information is acquired according to the first feature information group, the second feature information group, and the type information of the abnormal commodity object to be identified.

12. The device according to claim 11, characterized in that When the feature extraction module acquires the auxiliary feature information according to the first feature information group, the second feature information group, and the type information of the abnormal commodity object to be identified, it is used to: constructing an initial feature vector according to the first feature information group and the second feature information group; Determine one of the first feature information group and the second feature information group that is related to the type information of the abnormal commodity object to be identified as a target feature information group, and the other group as a constraint feature information group; The dimension related to the constraint feature information group in the initial feature vector is set to 0 to obtain a first auxiliary feature vector; and the positive direction of the coordinate axis of each dimension related to the constraint feature information group is taken as a second auxiliary feature vector; Accordingly, when the identification module obtains the commodity objects to be identified on the convex hull boundary corresponding to the plurality of commodity objects to be identified according to the target feature information of each commodity object to be identified and the preset convex hull model, it is used to: The initial feature vector, the first auxiliary feature vector, the second auxiliary feature vector and the origin of the coordinate system of each commodity object to be identified are input into the preset convex hull model, the convex hulls corresponding to the multiple commodity objects to be identified are obtained, and the commodity objects to be identified on the boundary of the convex hull are obtained.

13. The device according to claim 12, characterized in that When determining one of the first feature information group and the second feature information group that is related to the type information of the abnormal commodity object to be identified as the target feature information group and the other as the constraint feature information group, the feature extraction module is used to: If the type information of the abnormal commodity object to be identified is a first type of abnormal commodity object with an abnormally high order conversion rate, the first feature information group is used as the target feature information group, and the second feature information group is used as the constraint feature information group; or If the type information of the abnormal commodity object to be identified is the second type of abnormal commodity object with abnormally low order conversion rate, the second feature information group is used as the target feature information group, and the first feature information group is used as the constraint feature information group.

14. An electronic device, characterized in that: include: Memory, processor; Wherein, the memory is used to store the processor executable instructions; The processor is configured to: Acquire historical information related to multiple commodity objects to be identified from a database; Determine target feature information of each commodity object to be identified according to the historical information; the target feature information includes initial feature information, or the target feature information includes the initial feature information and auxiliary feature information; the initial feature information is feature information related to the interaction between the user and the commodity determined from the historical information; the auxiliary feature information is obtained based on the initial feature information and the type information of the abnormal commodity object to be identified; According to the target feature information of each commodity object to be identified and the preset convex hull model, the commodity objects to be identified on the convex hull boundary corresponding to the multiple commodity objects to be identified are obtained as abnormal commodity objects; The abnormal commodity object is outputted to perform corresponding processing on the abnormal commodity object.

15. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method for identifying abnormal commodity objects according to any one of claims 1 to 8.

Images