Order risk assessment model training method and device and order risk assessment method and device

Abstract

The invention provides an order risk assessment model training method and device and an order risk assessment method and device, and the order risk assessment model training method comprises the steps: determining the operation sequence data of a historical order and the information change data of a target user based on the historical order of the target user; according to different data modes, the operation sequence data, the information change data and the order data of the historical orders are embedded into an order risk assessment model; and enabling the order risk assessment model to output a risk assessment result of the historical order based on the operation sequence data, the information change data and order data of the historical order, and adjusting model parameters of the order risk assessment model based on the risk assessment result and a real risk tag of the historical order to train the order risk assessment model. According to the embodiment, the order risk assessment model can better understand and capture user preferences, so that the accuracy of risk assessment and the applicability of the order risk assessment model are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to an order risk assessment model training, an order risk assessment method and a device. Background Art

[0002] By conducting risk assessment on orders, effective early warning and intervention can be carried out before risks occur. However, existing order risk identification models usually adopt a rule-based approach, which uses the order characteristics, device characteristics or user's Internet Protocol address of the order to be evaluated based on preset rules to conduct risk assessment on the order to be evaluated. Such an approach may lead to misjudgment, low accuracy of risk assessment, and poor identification effect on risk order types that are not included in the preset rules. Summary of the invention

[0003] In view of this, the embodiments of the present disclosure at least provide an order risk assessment model training, an order risk assessment method and a device, which can improve the accuracy of order risk assessment and improve the applicability of the order risk assessment model.

[0004] In a first aspect, the present disclosure provides an order risk assessment model training method, including:

[0005] Based on the historical orders of the target user, determine the operation sequence data of the historical orders and the information change data of the target user;

[0006] According to different data modalities, the operation sequence data, information change data and order data of historical orders are embedded into the order risk assessment model; wherein, the methods of embedding data of different modalities into the order risk assessment model are different;

[0007] The order risk assessment model is made to output the risk assessment results of historical orders based on the operation sequence data, information change data and order data of historical orders. Based on the risk assessment results and the real risk labels of the historical orders, the model parameters of the order risk assessment model are adjusted to train the order risk assessment model.

[0008] Optionally, determine the operation sequence data of the historical order, including:

[0009] The operation sequence data of the historical orders are generated according to each operation event in the process of the target user placing an order for the historical orders and the timestamp information of each operation event.

[0010] Optionally, determining the information change data of the target user includes:

[0011] Based on the historical orders of the target user, obtain the order information change records of the target user in multiple dimensions; the dimensions include the recipient dimension and the Internet Protocol address dimension of the target user;

[0012] Based on the order information change records, determine the target user's information change data.

[0013] Optionally, the data modality includes categorical modality, numerical modality and text modality; according to different data modalities, the operation sequence data, information change data and order data of historical orders are embedded into the order risk assessment model, including:

[0014] For the category modal data in the operation sequence data, information change data, and order data of historical orders, an embedding matrix is ​​created according to the number of categories of the category modal data to embed the category modal data in the operation sequence data, information change data, and order data of historical orders into the order risk assessment model; wherein the embedding matrix is ​​initialized using a normal distribution;

[0015] For the numerical modal data in the operation sequence data, the information change data and the order data of the historical orders, the numerical modal data are normalized, and the normalized numerical modal data are input into the order risk assessment model;

[0016] For the text modal data in the operation sequence data, information change data, and order data of historical orders, word segmentation is performed on the text modal data. Based on the word segmentation results, a text embedding vector corresponding to the text modality is generated, and the text embedding vector is embedded into the order risk assessment model.

[0017] Optionally, before the order risk assessment model is enabled to output the risk assessment result of the historical order based on the operation sequence data, the information change data and the order data of the historical order, the method further includes:

[0018] According to the preset weight value, the operation sequence data, information change data and order data of historical orders are fused, so that the order risk assessment model outputs the risk assessment results of the historical orders based on the fused operation sequence data, information change data and order data of historical orders; wherein the preset weight value is determined based on the order time of the historical orders.

[0019] In a second aspect, the present disclosure provides an order risk assessment method, including:

[0020] Determine the operation sequence data of the target user's order to be evaluated;

[0021] Inputting the operation sequence data and order data of the order to be evaluated into a pre-trained order risk assessment model, so that the order risk assessment model outputs a risk assessment result of the order to be evaluated;

[0022] Among them, the order risk assessment model is trained based on the order risk assessment model training method of any one of claims 1-5.

[0023] In a third aspect, the present disclosure provides an order risk assessment model training device, including:

[0024] A first determination module is used to determine the operation sequence data of the historical orders and the information change data of the target user based on the historical orders of the target user;

[0025] An embedding module, used to embed the operation sequence data, information change data and order data of historical orders into the order risk assessment model according to different data modalities; wherein the method of embedding data of different modalities into the order risk assessment model is different;

[0026] The training module is used to enable the order risk assessment model to output the risk assessment results of historical orders based on the operation sequence data, information change data and order data of historical orders, and adjust the model parameters of the order risk assessment model based on the risk assessment results and the real risk labels of the historical orders to train the order risk assessment model.

[0027] In a fourth aspect, an embodiment of the present disclosure provides an order risk assessment device, including:

[0028] The second determination module is used to determine the operation sequence data of the order to be evaluated by the target user;

[0029] An evaluation module, used for inputting the operation sequence data and order data of the order to be evaluated into a pre-trained order risk evaluation model, so that the order risk evaluation model outputs a risk evaluation result of the order to be evaluated;

[0030] Among them, the order risk assessment model is trained based on the order risk assessment model training method of any one of claims 1-5.

[0031] In a fifth aspect, an embodiment of the present disclosure further provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the computer device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the above-mentioned first aspect, or the steps in any optional implementation of the first aspect are performed.

[0032] In a sixth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned first aspect, or any optional implementation of the first aspect are executed.

[0033] In a seventh aspect, an embodiment of the present disclosure further provides a computer program product, including a computer program, which implements the method of any of the above embodiments when executed by a processor.

[0034] Any of the above aspects or any of the implementation methods of the recognition of any aspect, by determining the operation sequence data of the historical user and the information change data of the target user based on the historical orders of the target user; thus, when evaluating the order, in addition to being able to evaluate based on the order characteristics of the current order to be evaluated like the conventional order evaluation method, it can also be combined with the historical behavior data of the user in a long period, which helps the model to better understand and capture the user's behavior habits, thereby improving the accuracy of recognition. For example, the target user needs to frequently buy items for relatives located in other geographical locations, resulting in frequent changes in information such as the recipient's address. If the conventional method is used, the risk probability of the current order may be considered high because the order address does not match the current user's Internet Protocol address. The model trained by the order risk assessment model training method provided by the embodiment of the present disclosure can be combined with the user's behavior habits in a long period, and based on these behavior habits, the current order is risk assessed, so that the order in the above situation will not be misjudged. In addition, the order risk assessment model training method provided by the embodiment of the present disclosure can also fuse data of different modes, which can not only extract time series features, but also effectively fuse multimodal features such as text. In addition, the prior art usually relies on preset rules when evaluating risky orders, and is unable to discover risks that do not exist in the rules. The order risk assessment model trained by the disclosed embodiment can combine the order's own characteristics with the user's behavior data over a long period of time, thereby better understanding the user's behavior patterns and demands for different items, thereby improving the ability to identify new risky behaviors, and thus improving the applicability of the order risk identification model. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following is a brief introduction to the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used together with the specification to illustrate the technical solutions of the present disclosure. It should be understood that the following drawings only illustrate certain embodiments of the present disclosure and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can also be obtained based on these drawings without creative work.

[0036] Figure 1 A flow chart of an order risk assessment model training method provided by an embodiment of the present disclosure is shown;

[0037] Figure 2A schematic diagram of order risk assessment model training provided by an embodiment of the present disclosure is shown;

[0038] Figure 3 A schematic diagram of historical order data provided by an embodiment of the present disclosure is shown;

[0039] Figure 4 A flow chart of an order risk assessment method provided by an embodiment of the present disclosure is shown;

[0040] Figure 5 A schematic diagram of an order risk assessment model training device provided by an embodiment of the present disclosure is shown;

[0041] Figure 6 A schematic diagram of an order risk assessment device provided by an embodiment of the present disclosure is shown;

[0042] Figure 7 An exemplary system architecture is shown in which embodiments of the present disclosure may be applied;

[0043] Figure 8 A schematic diagram of the structure of a computer system of a terminal device or a server used to implement an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0044] It should be noted that in the technical solution of the present invention, the collection, use, storage, sharing and transfer of user personal information involved are in compliance with the provisions of relevant laws and regulations, and it is necessary to inform the user and obtain the user's consent or authorization. When applicable, the user's personal information is de-identified and / or anonymized and / or encrypted.

[0045] The raising of the above problems and the solutions are the results obtained by the inventor after practice and careful research. The discovery process of the above problems and the solutions proposed for the above problems should be the contributions made by the inventor to the present disclosure during the disclosure process.

[0046] The technical solutions in the present disclosure will be clearly and completely described below in conjunction with the drawings in the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The components of the present disclosure generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the present disclosure claimed for protection, but merely represents the selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present disclosure.

[0047] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0048] To facilitate understanding of this embodiment, first, a detailed introduction is given to an order risk assessment model training and an order risk assessment method disclosed in the embodiment of the present disclosure. The execution subject of the order risk assessment model training and the order risk assessment method provided in the embodiment of the present disclosure is generally a computer device with a certain computing capability, and the computer device includes, for example: a terminal device or a server or other processing device, and the terminal device can be a user equipment (User Equipment, UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (Personal Digital Assistant, PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. In some possible implementations, the order risk assessment model training and the order risk assessment method can be implemented by a processor calling a computer-readable instruction stored in a memory.

[0049] See also Figure 1 As shown, it is a flow chart of the order risk assessment model training method provided by the embodiment of the present disclosure, and the method includes S101-S103, wherein:

[0050] S101: Based on the historical orders of the target user, determine the operation sequence data of the historical orders and the information change data of the target user.

[0051] In this step, the historical orders of the target user within a certain period of time can be obtained and arranged in ascending order according to the order time to form an order flow. Then, based on the order in the order flow, the operation sequence data of the historical orders and the information change data of the target user can be determined.

[0052] In the embodiments of the present disclosure, Figure 2 As shown, data in multiple dimensions can be extracted from historical orders, such as clickstream data in the traffic dimension; item name, item category, etc. in the item dimension; delivery address, user Internet Protocol address, etc. in the address dimension; payment method in the payment dimension; and order time in the time dimension.

[0053] In a specific implementation, the target user's historical orders in the past six months can be selected, and a maximum order threshold can be set. For example, when the target user's historical orders in the past six months are greater than 200, the subsequent steps can be performed based on the user's past 200 historical orders. In addition, a minimum order threshold can also be set. For example, when the number of historical orders of the user in the past six months is less than 100, the acquisition can continue until the required number of historical orders is obtained. The embodiments of the present disclosure do not make specific limitations on this. In actual applications, it can be selected according to the actual situation to achieve its function.

[0054] In the disclosed embodiment, the clickstream data of the target user's historical orders in the flow dimension can be determined based on each operation event. Specifically, the operation sequence data of the historical orders is determined, including: generating the operation sequence data of the historical orders according to each operation event in the process of the target user placing an order for the historical orders and the timestamp information of each operation event.

[0055] In specific implementation, when a historical order occurs, the event tracking tool can be used to capture various events related to the historical order, such as item browsing, adding to the shopping cart, filling in information, payment, etc. In addition, the timestamp information of each operation event can be recorded, and then this information can be integrated according to the order dimension, thus forming the click stream data of the historical order. Then the click stream data of the historical order can be saved, so that when training the order risk assessment model, the operation sequence data of the target user's historical order can be obtained from the corresponding storage location.

[0056] In another possible implementation, the user's click, slide and scroll behaviors on the page can also be visualized by heat map technology. For example, using relevant tools, a heat map can be generated to show the locations where users click most frequently on the website. By combining the heat map with the time series and superimposing the heat maps in different time periods, the operation sequence data of the target user's historical orders can also be obtained.

[0057] It should be noted that the above method for determining the operation sequence data of the target user's historical orders is only illustrated as an example of a feasible implementation method of the embodiment of the present disclosure, and does not constitute an improper limitation on the present invention. In practical applications, it can be set according to actual conditions. The embodiment of the present disclosure does not make specific limitations on this, and the ability to achieve its functions shall prevail.

[0058] like Figure 3As shown, in addition to the operation sequence data of the flow dimension, data of different dimensions such as item dimension, address dimension, payment dimension, and time dimension can also be extracted from historical orders. The order data of the item dimension can include item name, third-level category to which the item belongs, item purchase quantity, order type, order status, etc.; the address dimension can include the delivery address, Internet protocol address of the ordering user, consignee, and delivery contact information; the payment dimension can include the payment type; and the time dimension can include the order time. Among the above data, there are some data associated with the target user and can change with different orders, such as the delivery address and consignee of the address dimension. For different historical orders, these data may be different. For example, the payment method of the payment dimension, for some orders, the target user may choose mobile payment, while for some orders, the target user may also choose company transfer and other methods. Therefore, when obtaining the data of the above dimensions, the information change data associated with the target user can also be integrated based on these data, such as how many times the delivery address has changed, how many times the ordering Internet protocol address has changed, etc. Specifically, determining the target user's information change data includes: based on the target user's historical orders, obtaining the target user's order information change records in multiple dimensions; the dimensions include the receiver dimension and the target user's Internet Protocol address dimension; based on the order information change records, determining the target user's information change data.

[0059] For example, it is assumed that the information change data of user A needs to be determined based on the historical orders of user A. First, in the item dimension, based on the historical orders of user A, it can be found that he has purchased a variety of goods, including electronic products, furniture and clothing. For example, user A purchased a laptop in January 2023 and a winter coat in March 2023. These data help the order risk assessment module to better understand the purchase preferences of user A. In the address dimension, the delivery address of user A in January 2023 was "A Road, Chaoyang District, Beijing", and in March 2023, its delivery address was changed to "B Road, Haidian District, Beijing". From this, we can determine that in the historical orders, the delivery address of user A has changed once. In the payment dimension, user A chose credit card payment in January 2023, and chose ordinary online payment in March 2023. From this, it can be determined that the payment type of user A has changed once in the historical orders. Combined with the above data, the information change data of user A in the historical orders can be obtained.

[0060] It should be noted that the above-mentioned information change data is only used as an example of a possible implementation method of the present disclosure. In addition to the change data mentioned above, the consignee's contact information, the Internet Protocol address of the ordering user, the consignee, etc. can all be integrated into the information change data to help the order risk assessment model better understand the behavioral habits of the target users.

[0061] S102: According to different data modes, the operation sequence data, information change data and order data of historical orders are embedded into the order risk assessment model.

[0062] In this step, the data of each dimension mentioned above may contain data of different modes, and the methods of embedding data of different modes into the order risk assessment model may be different.

[0063] like Figure 3 As shown, the item name in the item dimension and the delivery address in the address dimension can be data in text mode; the item purchase quantity in the item dimension, the number of delivery address changes in the address dimension, the number of changes in the ordering user's Internet Protocol address, the number of changes in the consignee, the number of changes in the consignee's contact information, and the time interval between orders can be data in numerical mode; the operation sequence data in the flow dimension, the item category, order type, order status, payment method, etc. in the item dimension can be data in category mode. In order to embed the above data of different modes into the order risk identification model, different embedding methods can be used according to the different characteristics of different modal data.

[0064] In an embodiment of the present disclosure, data modalities include categorical modalities, numerical modalities, and textual modalities; according to different data modalities, operation sequence data, information change data, and order data of historical orders are embedded into an order risk assessment model, including: for categorical modal data in the operation sequence data, information change data, and order data of historical orders, an embedding matrix is ​​created according to the number of categories of the categorical modal data to embed the categorical modal data in the operation sequence data, information change data, and order data of historical orders into the order risk assessment model; wherein the embedding matrix is ​​initialized using a normal distribution; for numerical modal data in the operation sequence data, information change data, and order data of historical orders, the numerical modal data is normalized, and the normalized numerical modal data is input into the order risk assessment model; for textual modal data in the operation sequence data, information change data, and order data of historical orders, word segmentation is performed on the textual modal data, and a text embedding vector corresponding to the textual modality is generated based on the word segmentation result, and the text embedding vector is embedded into the order risk assessment model.

[0065] In the specific implementation, for category modal data, we can first count the number of different values ​​of each category feature, for example, there may be 100 categories to which an item belongs, and there may be 3 payment methods. Then, we can create an embedding matrix based on the number of each category, and its dimension can be (number of categories, embedding vector dimension). The embedding matrix can use the normal distribution to initialize the weights, that is, when constructing the embedding layer, the elements in the embedding matrix are randomly generated in the initial state, and these random values ​​obey the normal distribution, so that most weight values ​​will not be too large or too small, which can make the model training process more stable. Then, for the value of each category feature, it can be mapped to a low-dimensional dense vector by searching the embedding matrix, so that it can be better embedded in the order risk assessment model.

[0066] In a specific implementation, for the data of numerical modes, in order to ensure that the ranges of data of different numerical modes are similar, the data of numerical modes can be normalized to accelerate the convergence of model loss. For example, it can be scaled to the range of [0, 1], or standardized so that all are 0 and the standard deviation is 1. The normalized data of numerical modes can be directly used as model input without further embedding.

[0067] In a specific implementation, for data in textual modality, a pre-trained text embedding model can be used to convert the text into a vector representation to capture the semantic information in the textual data. For example, the data in the textual modality can be input into a pre-trained text embedding model, which can generate high-dimensional dense vectors based on the context of the text to represent the semantic information of the textual data. Alternatively, the textual data can be segmented and combed to generate an embedding vector for each segmentation or phrase, and then the sentence can be average pooled or other processing methods can be performed to obtain a vector representation of the entire textual data. The text embedding model here can be selected according to the actual situation, and the embodiments of the present disclosure do not specifically limit this.

[0068] It should be noted that the above method of embedding data of different modes into the order risk assessment model according to different modes is only used as an example of a possible implementation method in the embodiment of the present disclosure, and does not constitute an improper limitation of the present invention. In practical applications, a suitable embedding method can be selected according to actual conditions, and the embodiment of the present disclosure does not make specific limitations on this, and the function can be realized.

[0069] S103: The order risk assessment model is configured to output risk assessment results of historical orders based on the operation sequence data, information change data, and order data of historical orders, and the model parameters of the order risk assessment model are adjusted based on the risk assessment results and the real risk labels of the historical orders to train the order risk assessment model.

[0070] In this step, if Figure 2As shown in the figure, after embedding data of different dimensions and modalities into the order risk assessment model, these data can be fused and deeply embedded to obtain the final model embedding features.

[0071] In the embodiment of the present disclosure, before the order risk assessment model is enabled to output the risk assessment result of the historical order based on the operation sequence data, the information change data and the order data of the historical order, the following is further included:

[0072] According to the preset weight value, the operation sequence data, information change data and order data of historical orders are fused, so that the order risk assessment model outputs the risk assessment results of the historical orders based on the fused operation sequence data, information change data and order data of historical orders; wherein the preset weight value is determined based on the order time of the historical orders.

[0073] In a specific implementation, the operation sequence data, information change data and order data of historical orders can be fused. Since the aforementioned steps obtain the data of historical orders in the time dimension, when performing data fusion, the attenuation weight during fusion can be determined according to the order time of the historical orders, thereby ensuring sensitivity to changes in user interests. For example, there are two historical orders, one 2 months from the current time and the other 3 days from the current time. At this time, the weight of the data corresponding to the historical orders with an order time of 2 months from now can be set lower, and the weight of the data corresponding to the historical orders with an order time of 3 days from now can be set higher, so as to realize the dynamic change of perception of user preferences. Alternatively, according to actual conditions, the weight of the data corresponding to the historical orders with a longer time from now can be set higher. The embodiment of the present disclosure does not make specific limitations on this, and can be set according to actual conditions and needs. Then, the feature vector obtained after fusion can be subjected to high-level feature extraction, and the risk assessment results of the user's historical orders can be output.

[0074] In the disclosed embodiment, during training, two-dimensional labels can be constructed from multiple dimensions such as scalper orders, online gambling orders, and wool-pulling orders, and multi-label training can be performed through the model. The final model is used to conduct risk assessment on historical orders. Exemplarily, scalper orders are usually high-net-worth categories, and their delivery addresses change frequently, and there are abnormal identifications or remote orders in the delivery addresses; online gambling orders are generally virtual scenes such as phone bill recharges, and there are characteristics such as batch orders and concentrated Internet Protocol addresses; wool-pulling orders are generally low-value categories, which are usually manifested as frequent collection of merchant or site activity discounts. In a specific implementation, the risk scores corresponding to historical orders can be output for different types of risk orders, and then the one with the highest score is used as the risk assessment result of the historical order. For example, for a historical order, the model indicates that the risk score of the scalper order is 96 points, the risk score of the online gambling order is 56 points, and the risk score of other orders is no more than 20 points. At this time, the scalper order can be used as the risk assessment result of the historical order. After obtaining the risk assessment results of the order risk assessment model, the model parameters can be adjusted based on the real risk labels of historical orders and the risk assessment results output by the model to train the final order risk assessment model.

[0075] See also Figure 4 FIG. 4 is a flowchart of an order risk assessment method provided by an embodiment of the present disclosure, and the method includes S401 to S402, wherein:

[0076] S401: Determine the operation sequence data of the order to be evaluated of the target user.

[0077] In this step, after the order risk assessment model is trained through the above steps, it can be applied in the order risk assessment scenario.

[0078] In the embodiment of the present disclosure, when performing a risk assessment on the target user's current order to be evaluated, the operation sequence data of the order to be evaluated can be determined first, including events such as browsing, adding to shopping cart, payment, etc. associated with the order to be evaluated and corresponding timestamp information. The determination method is similar to the method for determining the operation sequence data of historical orders described above, and will not be repeated here.

[0079] S402: Inputting the operation sequence data and order data of the order to be evaluated into a pre-trained order risk assessment model, so that the order risk assessment model outputs a risk assessment result of the order to be evaluated.

[0080] In the embodiment of the present disclosure, after the operation sequence data of the order to be evaluated is determined, it can be input together with the order data of the order to be evaluated into the order risk assessment model trained through steps S101 to S103, so that the model outputs the risk assessment result of the order to be evaluated. It should be noted that Figure 3The several risk assessment results are only used as examples of possible implementation methods of the embodiments of the present disclosure and do not constitute improper limitations on the present invention. In practical applications, they can be adjusted according to actual conditions, and the embodiments of the present disclosure do not make specific limitations on this.

[0081] According to a third aspect of the embodiment of the present disclosure, Figure 5 As shown, an order risk assessment model training device 500 is provided, comprising:

[0082] The first determination module 501 is used to determine the operation sequence data of the historical orders and the information change data of the target user based on the historical orders of the target user;

[0083] An embedding module 502 is used to embed the operation sequence data, information change data and order data of historical orders into the order risk assessment model according to different data modalities; wherein the method of embedding data of different modalities into the order risk assessment model is different;

[0084] The training module 503 is used to enable the order risk assessment model to output the risk assessment results of historical orders based on the operation sequence data, information change data and order data of historical orders, and adjust the model parameters of the order risk assessment model based on the risk assessment results and the actual risk labels of the historical orders to train the order risk assessment model.

[0085] Optionally, the first determining module 501 is specifically configured to:

[0086] The operation sequence data of the historical orders are generated according to each operation event in the process of the target user placing an order for the historical orders and the timestamp information of each operation event.

[0087] Optionally, the first determining module 501 is specifically configured to:

[0088] Based on the historical orders of the target user, obtain the order information change records of the target user in multiple dimensions; the dimensions include the recipient dimension and the Internet Protocol address dimension of the target user;

[0089] Based on the order information change records, determine the target user's information change data.

[0090] Optionally, the data modality includes categorical modality, numerical modality and textual modality;

[0091] The embedding module 502 is specifically used for:

[0092] For the category modal data in the operation sequence data, information change data, and order data of historical orders, an embedding matrix is ​​created according to the number of categories of the category modal data to embed the category modal data in the operation sequence data, information change data, and order data of historical orders into the order risk assessment model; wherein the embedding matrix is ​​initialized using a normal distribution;

[0093] For the numerical modal data in the operation sequence data, the information change data and the order data of the historical orders, the numerical modal data are normalized, and the normalized numerical modal data are input into the order risk assessment model;

[0094] For the text modal data in the operation sequence data, information change data, and order data of historical orders, word segmentation is performed on the text modal data. Based on the word segmentation results, a text embedding vector corresponding to the text modality is generated, and the text embedding vector is embedded into the order risk assessment model.

[0095] Optionally, the training module 503 is specifically used for:

[0096] According to the preset weight value, the operation sequence data, information change data and order data of historical orders are fused, so that the order risk assessment model outputs the risk assessment results of the historical orders based on the fused operation sequence data, information change data and order data of historical orders; wherein the preset weight value is determined based on the order time of the historical orders.

[0097] According to the fourth aspect of an embodiment of the present disclosure, an electronic device for order risk assessment model training and order risk assessment is provided, comprising: one or more processors; a storage device for storing one or more programs, and when the one or more programs are executed by one or more processors, the one or more processors implement the method provided by the first aspect of the embodiment of the present invention.

[0098] According to a fifth aspect of an embodiment of the present disclosure, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method provided by the first aspect of the embodiment of the present disclosure is implemented.

[0099] According to a sixth aspect of an embodiment of the present invention, there is provided a computer program product, including a computer program, which implements the method of any of the above embodiments when executed by a processor.

[0100] According to a seventh aspect of the embodiment of the present disclosure, Figure 6 As shown, an order risk assessment device 600 is provided, comprising:

[0101] The second determination module 601 is used to determine the operation sequence data of the order to be evaluated of the target user;

[0102] An evaluation module 602, used to input the operation sequence data and order data of the order to be evaluated into a pre-trained order risk evaluation model, so that the order risk evaluation model outputs a risk evaluation result of the order to be evaluated;

[0103] Among them, the order risk assessment model is trained based on the order risk assessment model training method of any one of claims 1-5.

[0104] Figure 7 An exemplary system architecture 700 is shown to which the order risk assessment model training, order risk assessment method or order risk assessment model training, and order risk assessment apparatus implemented by the present disclosure can be applied.

[0105] like Figure 7 As shown, system architecture 700 may include terminal devices 701, 702, 703, network 704 and server 705. Network 704 is used to provide a medium for communication links between terminal devices 701, 702, 703 and server 705. Network 704 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0106] The user can use the terminal devices 701, 702, 703 to interact with the server 705 through the network 704 to receive or send messages, etc. Various communication client applications can be installed on the terminal devices 701, 702, 703, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only for example).

[0107] The terminal devices 701 , 702 , and 703 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0108] The server 705 may be a server that provides various services, such as a backend management server (for example only) that provides support for shopping websites browsed by users using the terminal devices 701, 702, and 703. The backend management server may process the received order risk assessment model training and order risk assessment requests, and feed back the processing results (for example only) to the terminal device.

[0109] It should be noted that the order risk assessment model training and order risk assessment method provided in the embodiment of the present invention are generally executed by the server 705, and accordingly, the order risk assessment model training and order risk assessment device are generally set in the server 705. The order risk assessment model training and order risk assessment method provided in the embodiment of the present invention can also be executed by the terminal devices 701, 702, and 703, and accordingly, the order risk assessment model training and order risk assessment device can be set in the terminal devices 701, 702, and 703.

[0110] It should be understood that Figure 7 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0111] Reference below Figure 8 , which shows a schematic diagram of the structure of a computer system 800 of a terminal device suitable for implementing an embodiment of the present invention. Figure 8 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0112] like Figure 8 As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage part 808 to a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the system 800 are also stored. The CPU 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0113] The following components are connected to the I / O interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, a modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed, so that a computer program read therefrom is installed into the storage section 808 as needed.

[0114] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the central processing unit (CPU) 801, the above-mentioned functions defined in the system of the present invention are executed.

[0115] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0116] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0117] The modules involved in the embodiments of the present invention may be implemented in software or hardware. The modules described may also be set in a processor. For example, a processor includes a first determination module, an embedding module, and a training module, wherein the names of these modules do not constitute a limitation on the modules themselves in certain cases. For example, the first determination module may also be described as "a module for determining the operation sequence data of historical orders and the information change data of target users based on historical orders of target users".

[0118] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device implements the following method: based on the historical orders of the target user, determine the operation sequence data of the historical orders and the information change data of the target user; according to different data modalities, embed the operation sequence data, information change data and order data of the historical orders into the order risk assessment model; wherein the method of embedding data of different modalities into the order risk assessment model is different; the order risk assessment model is made to output the risk assessment results of the historical orders based on the operation sequence data, information change data and order data of the historical orders, and adjust the model parameters of the order risk assessment model based on the risk assessment results and the real risk labels of the historical orders to train the order risk assessment model.

[0119] Finally, it should be noted that the above embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure is described in detail with reference to the above embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above embodiments within the technical scope disclosed in the present disclosure, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for training an order risk assessment model, characterized in that: include: Based on the historical orders of the target user, determining the operation sequence data of the historical orders and the information change data of the target user; According to different data modalities, the operation sequence data, the information change data and the order data of the historical orders are embedded into the order risk assessment model; wherein the method of embedding data of different modalities into the order risk assessment model is different; The order risk assessment model is enabled to output risk assessment results of the historical orders based on the operation sequence data, the information change data and the order data of the historical orders; and the model parameters of the order risk assessment model are adjusted based on the risk assessment results and the real risk labels of the historical orders to train the order risk assessment model.

2. The method according to claim 1, characterized in that Determine the operation sequence data of the historical order, including: The operation sequence data of the historical order is generated according to each operation event in the process of the target user placing an order on the historical order and the timestamp information of each operation event.

3. The method according to claim 1, characterized in that Determining the information change data of the target user includes: Based on the historical orders of the target user, obtaining order information change records of the target user in multiple dimensions; the dimensions include a recipient dimension and an Internet Protocol address dimension of the target user; Based on the order information change record, determine the information change data of the target user.

4. The method according to claim 1, characterized in that: The data modalities include category modalities, numerical modalities, and text modalities; according to different data modalities, the operation sequence data, information change data, and order data of the historical orders are embedded into the order risk assessment model, including: For the category modal data in the operation sequence data, the information change data, and the order data of the historical orders, an embedding matrix is ​​created according to the number of categories of the category modal data to embed the category modal data in the operation sequence data, the information change data, and the order data of the historical orders into the order risk assessment model; wherein the embedding matrix is ​​initialized using a normal distribution; For the numerical modal data in the operation sequence data, the information change data and the order data of the historical orders, normalizing the numerical modal data, and inputting the normalized numerical modal data into the order risk assessment model; For the text modality data in the operation sequence data, information change data and order data of the historical orders, word segmentation processing is performed on the text modality data, and based on the word segmentation processing result, a text embedding vector corresponding to the text modality is generated, and the text embedding vector is embedded into the order risk assessment model.

5. The method according to claim 1, characterized in that Before the order risk assessment model is made to output the risk assessment result of the historical order based on the operation sequence data, the information change data and the order data of the historical order, the method further includes: According to the preset weight value, the operation sequence data, the information change data and the order data of the historical order are fused, so that the order risk assessment model outputs the risk assessment result of the historical order based on the fused operation sequence data, the information change data and the order data of the historical order; wherein the preset weight value is determined based on the order time of the historical order.

6. An order risk assessment method, characterized in that: include: Determine the operation sequence data of the target user's order to be evaluated; Inputting the operation sequence data and order data of the order to be evaluated into a pre-trained order risk assessment model, so that the order risk assessment model outputs a risk assessment result of the order to be evaluated; Wherein, the order risk assessment model is trained based on the order risk assessment model training method described in any one of claims 1-5.

7. An order risk assessment model training device, characterized in that: include: A first determination module, configured to determine, based on the historical orders of the target user, the operation sequence data of the historical orders and the information change data of the target user; An embedding module, used to embed the operation sequence data, information change data and order data of the historical orders into the order risk assessment model according to different data modalities; wherein the method of embedding data of different modalities into the order risk assessment model is different; A training module is used to enable the order risk assessment model to output the risk assessment results of the historical orders based on the operation sequence data, the information change data and the order data of the historical orders, and to adjust the model parameters of the order risk assessment model based on the risk assessment results and the real risk labels of the historical orders to train the order risk assessment model.

8. An order risk assessment method, characterized in that: include: The second determination module is used to determine the operation sequence data of the order to be evaluated by the target user; An evaluation module, used for inputting the operation sequence data and order data of the order to be evaluated into a pre-trained order risk evaluation model, so that the order risk evaluation model outputs a risk evaluation result of the order to be evaluated; Wherein, the order risk assessment model is trained based on the order risk assessment model training method described in any one of claims 1-5.

9. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

10. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

11. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images