Data processing method, system, and device

Abstract

A data processing method, system and device are provided. The method includes: extracting user-related data, merchant-related data, environment-related data and corresponding delay durations in historical orders; identifying delay duration labels according to the delay duration intervals in which the delay durations fall; performing feature encoding on the extracted data to obtain feature vectors; training a first preset machine learning model based on the feature vectors and the delay duration labels; in response to a delay insurance package request, extracting multiple data of a delivery order; performing feature encoding on the extracted multiple data to obtain feature vectors, and inputting the encoded feature vectors into the first preset machine learning model to obtain the probabilities that the delay durations of the delivery order fall into multiple delay duration intervals; and determining the compensation indicators of multiple delay insurance packages based on the obtained probabilities and the insured amounts and premiums of multiple delay insurance grades, where each delay insurance package includes multiple delay insurance grades corresponding to multiple delay duration intervals.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing, and more particularly to data processing methods, systems, and devices. Background Art

[0002] The rapid development of the Internet and the continuous innovation of insurance products have provided good opportunities for the combination of various industries and the insurance industry. For example, the combination of the air transportation industry and the insurance industry, the combination of the medical industry and the insurance industry, etc. In recent years, the takeout model has become the consumption norm of modern office workers. The demand for takeout delivery has been continuously expanding, and the takeout market has become increasingly booming. However, due to factors such as a sharp increase in delivery orders, limited service capabilities of merchants themselves, insufficient delivery personnel, traffic jams, and bad weather, many problems of takeout delays have occurred. Therefore, in the case where the takeout delay problem has seriously affected the user experience, as one of the combined products of the takeout industry and the insurance industry, the advent of the takeout delay insurance takes into account the actual pain points of users and can improve the user experience in the takeout scenario. Summary of the Invention

[0003] In view of the above situation, the present disclosure provides a data processing system, device, and method.

[0004] In a first aspect, according to an embodiment of the present disclosure, there is provided a data processing method, including: extracting data related to a user, data related to a merchant, data related to an environment, and a corresponding delay duration from historical orders; identifying a corresponding delay duration label according to a delay duration interval identifier in a plurality of delay duration intervals into which the delay duration falls; performing feature encoding on the data related to the user, the data related to the merchant, and the data related to the environment to obtain corresponding feature vectors; training a first preset machine learning model based on the feature vectors and the delay duration label; in response to receiving a delay insurance package request corresponding to a delivery order, extracting a plurality of data of the delivery order, including data related to the current user, data related to the current merchant, and data related to the current environment; performing feature encoding on the plurality of data extracted from the delivery order to obtain corresponding feature vectors, and inputting the encoded feature vectors into the trained first preset machine learning model to obtain the probability that the delay duration of the delivery order estimated by the first preset machine learning model falls into a plurality of delay duration intervals; determining claim indicators for a plurality of delay insurance packages based on the respective probabilities that the estimated delay duration of the delivery order falls into the plurality of delay duration intervals and the insured amounts and premiums of a plurality of delay insurance levels corresponding to the plurality of delay duration intervals, wherein each of the plurality of delay insurance packages includes a plurality of delay insurance levels corresponding to the plurality of delay duration intervals.

[0005] Second aspect, according to an embodiment of the present disclosure, there is provided a data processing system, including: a first extraction device configured to extract user-related data, merchant-related data, environment-related data, and corresponding delay durations from historical orders; an identification device configured to identify a corresponding delay duration label according to a delay duration interval among a plurality of delay duration intervals in which the delay duration falls; an encoding device configured to perform feature encoding on the user-related data, the merchant-related data, and the environment-related data to obtain corresponding feature vectors; a training device configured to train a first preset machine learning model based on the feature vectors and the delay duration labels; a second extraction device configured to, in response to receiving a delay insurance package request corresponding to a delivery order, extract a plurality of data of the delivery order, including user-related data, merchant-related data, and environment-related data of the current time; a probability acquisition device configured to perform feature encoding on the plurality of data extracted from the delivery order to obtain corresponding feature vectors, and input the encoded feature vectors into the trained first preset machine learning model to obtain probabilities that the delay duration of the delivery order estimated by the first preset machine learning model falls into a plurality of delay duration intervals; a determination device configured to determine compensation indicators for a plurality of delay insurance packages based on the respective probabilities that the estimated delay duration of the delivery order falls into the plurality of delay duration intervals and the insured amounts and premiums of a plurality of delay insurance grades corresponding to the plurality of delay duration intervals, wherein each of the plurality of delay insurance packages includes a plurality of delay insurance grades corresponding to the plurality of delay duration intervals.

[0006] Third aspect, according to an embodiment of the present disclosure, there is provided a computer storage medium storing computer-executable instructions that, when run by a processor, execute a data processing method. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 is a block diagram showing a data processing system according to an embodiment of the present disclosure.

[0008] Figure 2 is a block diagram showing a data processing system according to another embodiment of the present disclosure.

[0009] Figure 3 is a block diagram showing a data processing system according to still another embodiment of the present disclosure.

[0010] Figure 4 is a block diagram showing a data processing system according to still another embodiment of the present disclosure.

[0011] Figure 5 is a schematic diagram showing a feature vector used to establish a first preset machine learning model.

[0012] Figure 6 It is a schematic diagram showing the feature vectors used to establish the second preset machine learning model.

[0013] Figure 7 It is a flowchart describing the data processing method for establishing the first preset machine learning model according to an embodiment of the present disclosure.

[0014] Figure 8 It is a flowchart showing the data processing method according to an embodiment of the present disclosure.

[0015] Figure 9 It is a flowchart showing the data processing method according to another embodiment of the present disclosure.

[0016] Figure 10 It is a flowchart describing feature encoding.

[0017] Figure 11 It is a block diagram showing the data processing device according to the present disclosure. Detailed implementation manners

[0018] Now, the technical solutions in the present disclosure will be clearly and completely described in conjunction with specific embodiments, and examples of the present disclosure are illustrated in detail in the drawings. Although the present disclosure will be described in conjunction with specific embodiments, it will be understood that it is not intended to limit the present disclosure to the described embodiments. On the contrary, it is intended to cover modifications, variations, and equivalents included in the spirit and scope of the present disclosure as defined by the appended claims. It should be noted that the method steps described herein can be implemented by any functional block or functional arrangement, and any functional block or functional arrangement can be implemented as a physical entity or a logical entity, or a combination of both.

[0019] Note that the examples to be introduced next are only specific examples and are not intended to limit the embodiments of the present disclosure to the specific shapes, hardware, connection relationships, steps, numerical values, conditions, data, sequences, etc. shown and described. Those skilled in the art can use the concepts of the present disclosure to construct more embodiments not mentioned in this specification by reading this specification.

[0020] Generally, when placing a takeaway order, a user can purchase takeaway delay insurance and pay a certain premium. If the takeaway delivery is delayed, the insurance company will pay a certain indemnity amount (i.e., the insured amount). For a certain takeaway order, there may be multiple candidate takeaway delay insurance packages, and the insured amount and premium of each delay insurance package may be different. Due to the many uncertainties in the delivery and delay time of takeaway delivery orders, therefore, selecting and recommending one or more suitable delay insurance packages to the user is crucial for enhancing the user's purchase intention and balancing the capital and indemnity indicators.

[0021] Figure 1The block diagram of a data processing system according to an embodiment of the present disclosure is shown.

[0022] Referring to Figure 1 , a data processing system 10 according to an embodiment of the present disclosure includes a delay module 101 and a recommendation module 102.

[0023] The delay module 101 further includes a delay prediction module 1011 and a first preset machine learning model 1012.

[0024] In response to receiving a delay insurance package request corresponding to a delivery order, the delay prediction module 1011 inputs user features, merchant features, and current environmental features related to the delivery order into the first preset machine learning model 1012 to estimate the probability that the delivery order falls into multiple delay duration intervals.

[0025] The recommendation module 102 can determine the compensation indicators of multiple delay insurance packages and recommend a certain or certain delay insurance packages based on the probability that the delivery order falls into multiple delay duration intervals estimated by the first preset machine learning model 1012 in the delay module 101 and the information of multiple delay insurance packages.

[0026] The first preset machine learning model 1012 is established by using machine learning algorithms with user features, merchant features, environmental features related to historical orders, and corresponding delay durations.

[0027] Next, the methods executed by each module in the Figure 1 data processing system 10 will be specifically described.

[0028] Referring to Figure 7 to describe the data processing method for establishing the first preset machine learning model 1012 according to an embodiment of the present disclosure.

[0029] The method includes: S701, extracting data related to users, data related to merchants, data related to the environment, and corresponding delay durations in historical orders.

[0030] S702, labeling the corresponding delay duration label according to a delay duration interval identifier in one of the multiple delay duration intervals into which the delay duration falls.

[0031] S703, performing feature encoding on the data related to users, the data related to merchants, and the data related to the environment to obtain corresponding feature vectors.

[0032] Referring to Figure 10 , Figure 10 The flowchart of feature encoding is described.

[0033] For example, user features and merchant features can be obtained from user-related data and merchant-related data through the following steps: filtering out user-related data and merchant-related data from historical orders or delivery orders; normalizing, taking the natural logarithm, taking the square root, and performing first encoding on the user-related data to obtain user features; normalizing, taking the natural logarithm, taking the square root, and performing first encoding on the merchant-related data to obtain merchant features.

[0034] Normalize, take the natural logarithm, take the square root, and perform first encoding on the environment-related data to obtain environment features. The environment-related data is obtained through the following steps: extracting the location code of the destination of the order initiated by the user and the location code of the merchant from the user-related data and the merchant-related data, obtaining the route from the merchant to the destination based on the location code of the destination of the order initiated by the user and the location code of the merchant, and obtaining environment-related data related to the route based on the route.

[0035] The first encoding includes: determining the number of values of a type of feature; constructing the number of encodings including that number of bits, such that each encoding corresponds to a value of the type of feature, and only one bit in each encoding is 1 and the remaining bits are 0, and the number of encodings are mutually exclusive.

[0036] Specifically, since the various data collected from orders is messy, may carry various non-numeric special symbols, and has a wide variety of formats and cannot be well recognized, it is considered to perform feature encoding on the data to obtain a feature vector with a more unified format. First, normalize the messy data to make it a decimal between (0, 1) or (1, 1). Map the data to the range of 0 to 1 for processing, so that the data can be processed more conveniently and quickly. Then, perform natural logarithm and square root operations.

[0037] Then, perform the first encoding. An N-bit status register is used to encode N states. Each state has N independent register bits, and at any given time, only one of them is valid. Specifically, for a feature with P categories (or values) (where P is a positive integer), it is always possible to construct P bits to represent it uniquely. For example, for a user's gender feature, if it is determined that the number of values of this gender feature is 2: male and female, an encoding consisting of two bits can be constructed, and the number of values of the encoding is also 2, which are: "male" → 10, "female" → 01. That is, each encoding corresponds to a value of this type of feature, and only one bit in each encoding is 1, and the rest of the bits are 0. These two encodings 10 and 01 are mutually exclusive. For a type of feature, if it has m possible values (where m is a positive integer), m encodings including m bits are constructed. For example, a feature with 6 possible values is encoded as 000001, 000010, 000100, 001000, 010000, 100000. Only one bit in the m bits is 1, and the other bits are all 0. Moreover, these encodings are mutually exclusive, and only one bit in each encoding is activated to 1. Therefore, after encoding, the data becomes sparse and discrete, capable of handling non-continuous numerical features. On the other hand, the features are also expanded. For example, gender itself is a feature, and after the above encoding, it becomes two features: male and female. And the discrete features are extended to the Euclidean space. A certain value corresponds to a certain point in the Euclidean space, and the distance calculation between features becomes more reasonable.

[0038] The above user-related data may include the geographical location of the destination of the order initiated by the user, the historical number of delays at the geographical location, the average delay duration, the number of insurance purchases and claim settlement times of the user, the number of displays of the user, the order amount, the user's delay insurance package, the number of orders initiated by the user, and so on.

[0039] The above merchant-related data may include the geographical location of the merchant, the meal preparation speed of the above merchant, the delivery method of the merchant, the order volume of the merchant, the number of visits to the merchant, the historical number of delays of the merchant, the average delay duration of the merchant, the number of insurance purchases and claim settlement times for the merchant, and so on.

[0040] The above data related to the route environment may include weather conditions, traffic conditions, time information, etc. during the delivery of the order. Here, the time information may include date information, week information, moment information, etc. For example, date information can help consider whether it is a festival, what season, summer vacation or winter vacation, etc. Week information can help consider whether it is a working day or a weekend, etc. Moment information can help consider whether it is the peak period of the day, whether it is the late-night snack time, etc.

[0041] Combine user features, merchant features, environmental features, and delay duration labels into a feature vector. The feature vector is equivalent to expanding the features into a Euclidean space, so a certain value corresponds to a certain point in the space. In this way, the feature vector obtained by feature encoding is more easily applied to machine learning algorithms that require similarity calculation or distance calculation. The "features" and "feature vectors" mentioned in this article are used interchangeably.

[0042] Figure 5 It is a schematic diagram showing the feature vector used to establish the first preset machine learning model.

[0043] Multiple delay duration labels correspond to multiple delay duration intervals. For example, label 0 indicates no delay, label 1 indicates a delay of 0 - 5 minutes, label 2 indicates a delay of 6 - 10 minutes, label 3 indicates a delay of more than 11 minutes, etc.

[0044] For example, for a historical order, its delay duration is 3 minutes, falling into the delay duration interval of 0 - 5 minutes, and its user features, merchant features, and environmental features are user feature 1, merchant feature 1, and environmental feature 1 respectively. Then the label of this historical order is label 1.

[0045] The above-mentioned feature vectors of user features, merchant features, environmental features, and delay duration labels are obtained for establishing the first preset machine learning model.

[0046] Return Figure 7 , S704, train the first preset machine learning model based on the feature vector and the delay duration label. Thus, the first preset machine learning model 1012 is established.

[0047] In this way, the first preset machine learning model is trained based on the feature vectors of many historical orders. The machine learning algorithms used for training can adopt various classification algorithms, such as gradient regression tree algorithm, deep learning algorithm, etc. The first preset machine learning model 1012 can learn the involved data and algorithms offline using machine learning principles.

[0048] In this way, the first preset machine learning model using various features (various user features, various merchant features, various environmental features) of historical orders and delay duration labels can more accurately reflect the comprehensive information of historical orders, more accurately simulate the relationship between various features of historical orders and the delay duration, so as to more accurately predict the probability of the delay duration of new delivery orders in the future.

[0049] In addition, the delay prediction of general machine learning only predicts whether there is a delay or how long the delay is, while the embodiments of the present disclosure divide the delay duration into multiple delay duration intervals, and the first preset machine learning model established by using the delay duration labels corresponding to each delay duration interval can predict the probability that the delay duration falls into each delay interval, which not only saves the calculation cost but also improves the prediction accuracy.

[0050] Next, refer to Figure 8 to describe the data processing method performed by the delay estimation module 1011 and the recommendation module 102 according to the embodiments of the present disclosure by using the first preset machine learning model 1012.

[0051] S705. In response to receiving a delay insurance package request corresponding to a delivery order, the delay estimation module 1011 extracts a plurality of data of the delivery order, including data related to the current user, data related to the current merchant, and data related to the current environment.

[0052] For example, a user initiates a delivery order on a food delivery platform, which is a food delivery from a certain merchant to a certain destination. Since the delivery has not been made yet, it is not known whether the delivery order will be delayed in reaching the destination at this time. If the user is worried that the delivery will be delayed, or the food delivery platform hopes that the user will purchase its delay insurance package, the user can initiate or the food delivery platform can directly initiate a delay insurance package request corresponding to the delivery order. Then, a plurality of features are extracted from the delivery order.

[0053] S706. The delay estimation module 1011 performs feature encoding on the plurality of data extracted from the delivery order to obtain corresponding feature vectors, and inputs the encoded feature vectors into the trained first preset machine learning model to obtain the probabilities that the delay duration of the delivery order estimated by the first preset machine learning model falls into multiple delay duration intervals.

[0054] As before, when receiving the feature vectors including user features, merchant features, and current environmental features input from the delay estimation module 1011, the first preset machine learning model 1012 determines the probabilities that the delivery order falls into multiple delay duration intervals according to the classification algorithm. For example, the probability of falling into the delay duration interval of 0 - 5 minutes is 60%, the probability of falling into the delay duration interval of 6 - 10 minutes is 30%, and the probability of falling into the delay duration interval of more than 11 minutes is 10%, and sends the determined probabilities of the multiple delay duration intervals to the recommendation module 102.

[0055] S707. The recommendation module 102 determines the compensation indicators of multiple delay insurance packages based on the respective probabilities of the estimated delay duration of the delivery order falling into multiple delay duration intervals and the sum of the insured amounts and premiums of multiple delay insurance levels corresponding to the multiple delay duration intervals, where each of the multiple delay insurance packages includes multiple delay insurance levels corresponding to the multiple delay duration intervals.

[0056] Each of the multiple delay insurance packages includes at least one delay insurance level corresponding to at least one delay duration interval. For example, Delay Insurance Package 1 includes a delay insurance level for a delay duration interval of 0 to 5 minutes, a delay insurance level for a delay duration interval of 6 to 10 minutes, and a delay insurance level for a delay duration interval of more than 11 minutes. Similarly, Delay Insurance Packages 2, 3, and 4 also include these delay insurance levels respectively.

[0057] For the delay insurance levels of each delay insurance package, the corresponding premiums and compensation insured amounts may be different. For example, the premium of Delay Insurance Package 1 is 5 yuan, and the compensation insured amount for the delay insurance level of 0 to 5 minutes of delay is 2 yuan, the compensation insured amount for the delay insurance level of 6 to 10 minutes of delay is 5 yuan, and the compensation insured amount for the delay insurance level of more than 11 minutes of delay is 10 yuan. The premium of Delay Insurance Package 2 is 6 yuan, and the compensation insured amount for the delay insurance level of 0 to 5 minutes of delay is 1 yuan, the compensation insured amount for the delay insurance level of 6 to 10 minutes of delay is 6 yuan, and the compensation insured amount for the delay insurance level of more than 11 minutes of delay is 12 yuan. The premium of Delay Insurance Package 3 is 2 yuan, and it only compensates the insured amount of 10 yuan for the delay insurance level of more than 11 minutes of delay. The premium of Delay Insurance Package 4 is 2 yuan, and the compensation insured amount for the delay insurance level of 6 to 10 minutes of delay is 1 yuan, and the compensation insured amount for the delay insurance level of more than 11 minutes of delay is 8 yuan, etc.

[0058] For example, the estimated probabilities of multiple delay duration intervals received from the first preset machine learning model 1012 in the delay module 101 are: the probability of a delay of 0 to 5 minutes is 60%, the probability of a delay of 6 to 10 minutes is 30%, and the probability of a delay of more than 11 minutes is 10%.

[0059] The prior art usually determines the one with the highest probability as the judgment result by the probability of falling into each classification. For example, it may be determined at this time that the delivery order will be delayed for 0 - 5 minutes. However, the present invention can use all the estimated probabilities of falling into each classification to calculate the compensation indicators of a certain delay insurance package.

[0060] For example, for a delay insurance package, the recommendation module 102 can use the weighted sum of the respective probabilities that the estimated delay duration of the delivery order falls into multiple delay duration intervals and the insured amounts of the corresponding multiple delay insurance levels in the delay insurance package to obtain the expected value of the insured amount of the delay insurance package, and then divide it by the premium of the delay insurance package to calculate the compensation index of the delay insurance package.

[0061] For example, the compensation index result of the determined delay insurance package 1 is as follows:

[0062] The compensation index of delay insurance package 1 is: (60% × 2 + 30% × 5 + 10% × 10) / 5 = 74%;

[0063] Similarly, the compensation index results of the determined delay insurance packages 2 - 4 are as follows respectively:

[0064] The compensation index of delay insurance package 2 is: (60% × 1 + 30% × 6 + 10% × 12) / 6 = 60%;

[0065] The compensation index of delay insurance package 3 is: (10% × 10) / 2 = 50%;

[0066] The compensation index of delay insurance package 4 is: (30% × 1 + 10% × 8) / 2 = 55%.

[0067] In this way, the compensation index of the delay insurance package calculated according to the embodiments of the present disclosure can more accurately reflect the influence of the respective probabilities falling into multiple delay duration intervals on the compensation index, so that the obtained compensation index of the delay insurance package is more accurate.

[0068] Then, as Figure 1 shown, the recommendation module 102 can recommend suitable delay insurance packages to the user from packages 1, 2,... N (N is a positive integer) in the package pool considering the compensation indexes of each delay insurance package. For example, the 3 delay insurance packages 1, 2, and 4 with the highest compensation indexes. It should be understood that although for the convenience of understanding, Figure 1 it is shown as recommending 3 delay insurance packages, it is also possible to recommend other numbers of packages without departing from the spirit scope of the present disclosure.

[0069] It should be understood that in this embodiment, the recommendation module 102 can recommend delay insurance packages based on the compensation indicators of the multiple delay insurance packages in various ways. For example, the recommendation module 102 can recommend delay insurance packages with lower compensation indicators for economic benefits, or recommend delay insurance packages with higher compensation indicators to promote user purchases. For another example, delay insurance packages higher or lower than a predetermined compensation indicator can be recommended according to the predetermined compensation indicator. For yet another example, the top M (M is a positive integer) delay insurance packages can be recommended in descending or ascending order of the compensation indicators, and so on. The ways of recommending delay insurance packages based on the compensation indicators are not limited to those exemplified above, and some ways not exemplified in this disclosure are also possible without departing from the spirit of this disclosure.

[0070] For example, based on a predetermined compensation indicator (such as 51%), the recommendation module 102 can filter out and recommend the first number of delay insurance packages (such as packages 1, 2, and 4 with a compensation indicator greater than 51%) from multiple delay insurance packages (such as packages 1, 2,..., N) in the package pool whose compensation indicators meet the predetermined compensation indicator.

[0071] The predetermined compensation indicator can be set in advance, or can be automatically set by the data processing system through machine learning based on the user's previous multiple selections. The setting of the predetermined compensation indicator can be made with a higher predetermined compensation indicator for business promotion, or a lower predetermined compensation indicator for economic benefit considerations. For business promotion, one or more delay insurance packages with a compensation indicator higher than the predetermined compensation indicator of the determined delay insurance packages can be filtered out, or for economic benefit, one or more delay insurance packages with a compensation indicator lower than the predetermined compensation indicator of the determined delay insurance packages can be filtered out.

[0072] Of course, the above examples are only used to illustrate the inventive concept of this disclosure, rather than to limit this disclosure.

[0073] Thus, the data processing system according to the embodiment of this disclosure can also recommend suitable delay insurance packages to the user terminal based on the compensation indicator of each delay insurance package, enabling the user to enjoy the real-time, fast, and automatic delay insurance package recommendation function and improving the user experience in the takeaway scenario.

[0074] Figure 2 The block diagram of a data processing system according to another embodiment of this disclosure is shown.

[0075] Reference Figure 2 , in which the delay module 101 is the same as the delay module 101 shown in Figure 1 , so the detailed description thereof will be omitted. And Figure 1In contrast, the data processing system 10 in this other embodiment may further include a purchase module 103. The purchase module 103 may further include a purchase prediction module 1031 and a second preset machine learning model 1032.

[0076] In response to receiving a delay insurance package request corresponding to a delivery order, the purchase prediction module 1031 inputs user features, merchant features, current environmental features, and delay insurance package features related to the delivery order into the second preset machine learning model to determine the probability of the user purchasing delay insurance.

[0077] The second preset machine learning model 1032 is established by inputting user features, merchant features, environmental features, delay insurance package features, and delay insurance purchase labels related to historical orders and using machine learning algorithms.

[0078] The recommendation module 102' included in the data processing system 10 may recommend at least one delay insurance package among multiple delay insurance packages based on the compensation indicators of the multiple delay insurance packages and the probability of the user purchasing each delay insurance package.

[0079] Reference Figure 9 To describe the data processing method for establishing the second preset machine learning model 1032 and the data processing method for the purchase prediction module 1031.

[0080] S901, identify the delay insurance purchase label according to whether the user in the historical order has purchased a delay insurance package.

[0081] The delay insurance purchase label indicates whether the user has purchased a delay insurance package. For example, label 0 indicates that the user did not purchase the delay insurance package in this historical order, and label 1 indicates that the user purchased the delay insurance package in this historical order.

[0082] S902, perform feature encoding on the data related to the user, the data related to the merchant, the data related to the environment, and the data related to the delay insurance package extracted from the historical order to obtain corresponding feature vectors.

[0083] The delay insurance package features are obtained through the following steps: normalizing, logarithmizing, taking the square root, and performing the first encoding on the data related to the delay insurance package to obtain the delay insurance package feature vector.

[0084] The user features, merchant features, and environmental features are obtained in the same manner as above from the data related to the user, the data related to the merchant, and the data related to the environment.

[0085] Figure 6 It is a schematic diagram showing the feature vectors used to establish the second preset machine learning model.

[0086] ReferenceFigure 6 The feature vectors used to establish the second preset machine learning model 103 are from historical takeaway orders, and its structure is divided into a label part and a feature part. Among them, for the feature vectors of the delay insurance package 1 features, the label part only contains 0 or 1. The label 0 indicates that for the delay insurance package 1 of a certain historical order, the user did not purchase the delay insurance package 1, and the label 1 indicates that for the delay insurance package 1 of this historical order, the user purchased the delay insurance package 1. Similarly, for the feature vectors of the delay insurance package 2 features, the label part only contains 0 or 1. The label 0 indicates that for the delay insurance package 2 of a certain historical order, the user did not purchase the delay insurance package 2, and the label 1 indicates that for the delay insurance package 2 of this historical order, the user purchased the delay insurance package 2. Among them, the feature part includes user features, merchant features, environmental features, and delay insurance package 2 features.

[0087] The user features, merchant features, and environmental features are the same as those above and will not be elaborated here.

[0088] The delay insurance package features include, but are not limited to, the delay insurance level, premium, insurance amount, etc. of the delay insurance package. Of course, the user features, merchant features, environmental features, and delay insurance package features are not limited to those exemplified here, and other examples that can represent these features can also be included in this disclosure.

[0089] For example, for the features of the delay insurance package 1 of a historical order, the user purchased the delay insurance package 1, and its user features, merchant features, and environmental features are user feature 1, merchant feature 1, and environmental feature 1 respectively. The label of this historical order is label 1. For example, for the features of the delay insurance package 2 of the same historical order, since the user did not purchase the delay insurance package 2, its user features, merchant features, and environmental features are user feature 1, merchant feature 1, and environmental feature 1 respectively. The label of this historical order is label 0. For other delay insurance packages, the same applies.

[0090] The above-mentioned feature vectors of user features, merchant features, environmental features, delay insurance package features, and delay insurance purchase labels are obtained for establishing the second preset machine learning model.

[0091] Return Figure 9 In S903, train the second preset machine learning model 1032 based on the feature vectors and the delay insurance purchase labels. Thus, the second preset machine learning model 1032 is established.

[0092] The machine learning algorithms used for training can include various classification algorithms, such as gradient regression tree algorithms, deep learning algorithms, etc. The second preset machine learning model 1032 can learn the involved data and algorithms offline using machine learning principles.

[0093] The second preset machine learning model 1032 of the embodiments of the present disclosure also considers the purchase behaviors of users for each delay insurance package, and thus more accurately reflects the comprehensive information of historical orders, reflecting the relationship between various features including the characteristics of the delay insurance package and the purchase behaviors of users for the delay insurance package, so as to more accurately predict the purchase probability of users of new delivery orders for the delay insurance package in the future.

[0094] S904. In response to receiving a delay insurance package request corresponding to a delivery order, the purchase estimation module 1031 extracts multiple data of the delivery order, including data related to the current user, data related to the current merchant, data related to the current environment, and data related to the current delay insurance package.

[0095] Herein, the data related to the current delay insurance package may include currently recommended delay insurance packages. For example, the delay insurance packages 1, ……, N in the package pool as Figure 2 shown. The currently recommended delay insurance packages may be pre-determined or may change in real time according to the compensation indicators or user purchase behaviors.

[0096] S905. The purchase estimation module 1031 performs feature encoding on the multiple data extracted from the delivery order to obtain corresponding feature vectors, and inputs the encoded feature vectors into the trained second preset machine learning model to obtain the probabilities of users of the delivery order purchasing each delay insurance package estimated by the second preset machine learning model.

[0097] The data related to the current delay insurance package is normalized, logarithmized, square-rooted, and first encoded to obtain the delay insurance package features.

[0098] The methods for obtaining user features, merchant features, and environment features from the data related to the current user, the data related to the current merchant, and the data related to the current environment are the same as those described above and will not be elaborated herein.

[0099] As Figure 2 shown, when receiving the feature vectors of user features, merchant features, environment features, and delay insurance package features input from the purchase estimation module 1031, the second preset machine learning model 1032 determines the purchase probabilities of users for each delay insurance package according to the classification algorithm.

[0100] For example, the second preset machine learning model 1032 determines the purchase probabilities of users for each delay insurance package as follows: the purchase probability for delay insurance package 1 is 30%, the purchase probability for delay insurance package 2 is 10%, the purchase probability for delay insurance package 3 is 45%, and the purchase probability for delay insurance package 4 is 15%.

[0101] Embodiments of the present disclosure can accurately calculate the respective purchase probabilities of users for each delay insurance package, and more accurately reflect the purchase habits of users.

[0102] The recommendation module 102' is configured to recommend at least one delay insurance package from the multiple delay insurance packages based on the compensation indicators of the multiple delay insurance packages and the probabilities of the user purchasing each delay insurance package.

[0103] Specifically, as Figure 2 shown, the recommendation module 102' has received the probabilities of multiple delay duration intervals from the delay module 101 to determine the compensation indicators of the multiple delay insurance packages based on these probabilities (as described in reference Figure 1 ), and has received the purchase probabilities of the user for each delay insurance package from the purchase module 103. The recommendation module 102' selects one or more packages in the package pool (for example, Package 1 and Package 2 that meet a certain predetermined compensation indicator and the user's purchase probability is greater than a certain predetermined threshold) and recommends them to the user based on the compensation indicators of the multiple delay insurance packages and the purchase probabilities of the user for each delay insurance package.

[0104] Thus, the recommendation module 102' recommends suitable packages to the user by considering both the compensation indicators and purchase probabilities of each package. It should be understood that although for convenience of understanding, Figure 2 it is shown as recommending 2 packages in, it is also possible to recommend other numbers of packages without departing from the spirit scope of the present disclosure.

[0105] It should be understood that in this embodiment, the recommendation module 102' can recommend delay insurance packages based on the compensation indicators of the multiple delay insurance packages and the purchase probabilities of the user for the multiple delay insurance packages in various ways. For example, the recommendation module 102' can recommend delay insurance packages with lower compensation indicators for economic benefits, or recommend delay insurance packages with higher purchase probabilities to promote user purchases. Another example is that, according to the predetermined compensation indicator, delay insurance packages higher or lower than the predetermined compensation indicator can be recommended, or according to the predetermined purchase probability, delay insurance packages higher or lower than the predetermined purchase probability can be recommended, or recommendations can be made by combining both the predetermined compensation indicator and the predetermined purchase probability. Another example is that the top M (M is a positive integer) delay insurance packages can also be recommended sorted by compensation indicator or purchase probability from high to low or from low to high, etc. The ways of recommending delay insurance packages based on determining the compensation indicator and determining the purchase probability are not limited to those exemplified above, and some ways not exemplified in the present disclosure are also possible without departing from the spirit scope of the present disclosure.

[0106] The above examples are only used to illustrate the inventive concept of the present disclosure, rather than to limit the present disclosure.

[0107] Thus, when the data processing system of another embodiment of the present disclosure recommends a suitable delay insurance package to a user terminal, it not only considers the compensation indicators of the delay insurance package but also considers the purchase probability of each delay insurance package by the user, so as to improve the user experience in the takeaway scenario and make the recommended delay insurance package more in line with the user's wishes.

[0108] Figure 3 FIG. shows a block diagram of a data processing system according to still another embodiment of the present disclosure.

[0109] Reference Figure 3 , Figure 3 the delay module 101 including the delay prediction module 1011 and the first delay machine learning model 1012 in Figure 1 and Figure 2 is the same as the delay module 101 in Figure 3 the purchase module 103 including the purchase prediction module 1031 and the second preset machine learning model 1032 in Figure 2 is the same as the purchase module 103 in , so a detailed description thereof will be omitted.

[0110] Different from Figure 1 and Figure 2 the recommendation module 102'' in the data processing system 10 under this still another embodiment is configured to filter out the first number of delay insurance packages whose compensation indicators meet the predetermined compensation indicators from multiple delay insurance packages based on the predetermined compensation indicators; and perform a descending order sorting on the filtered first number of delay insurance packages based on the probability of the user purchasing each delay insurance package, so as to recommend the second number of delay insurance packages ranked at the top, where the second number is less than the first number.

[0111] As Figure 3 shown, when the recommendation module 102'' receives the determined compensation indicators of multiple delay insurance packages from the delay module 101, it compares the predetermined compensation indicators with the determined compensation indicators of multiple delay insurance packages, and filters out one or more delay insurance packages whose determined compensation indicators of the delay insurance packages are higher or lower than the predetermined compensation indicators from the package pool including Package 1, Package 2,..., Package N, such as Package 2, 3, 4.

[0112] The predetermined compensation indicator can be preset by the system in advance, or can be automatically set by the data processing system through machine learning based on the user's previous multiple selections. The setting of the predetermined compensation indicator can be made based on business promotion to set a relatively high predetermined compensation indicator, or can be made for economic benefit considerations to set a relatively low predetermined compensation indicator. And for business promotion, one or more delay insurance packages whose determined compensation indicators of the delay insurance packages are higher than the predetermined compensation indicator can be filtered out, or for economic benefit, one or more delay insurance packages whose determined compensation indicators of the delay insurance packages are lower than the predetermined compensation indicator can be filtered out.

[0113] For example, for economic benefits, it is necessary to filter out the delay insurance packages whose compensation indicators of the determined delay insurance packages are lower than those of the predetermined delay insurance packages. For example, when the compensation indicator of the predetermined delay insurance package is 60%, the recommendation module 102” will filter out the delay insurance packages 2, 3, and 4 whose compensation indicators of the determined delay insurance packages are lower than 60%.

[0114] Then, the recommendation module 102” selects one or more packages from the filtered one or more delay insurance packages based on the purchase probabilities of each delay insurance package received from the purchase module and recommends them to the user.

[0115] For example, the recommendation module 102” recommends the delay insurance package to the user based on the purchase probabilities of the user for the filtered delay insurance packages 2, 3, and 4 determined by the purchase module (for example, recommends the package 3 with the highest purchase probability or the purchase probability greater than 40%). It should be understood that, although for ease of understanding, Figure 3 it is shown as only recommending 1 package, but without departing from the spirit scope of the present disclosure, it is also possible to recommend other quantities of packages.

[0116] The above examples are only used to illustrate the inventive concept of the present disclosure, rather than to limit the present disclosure.

[0117] Thus, according to another embodiment of the present disclosure, the data processing system preliminarily filters the delay insurance package pool according to the predetermined compensation indicator, and then recommends the delay insurance package based on the purchase probabilities of each package by the user, so as to improve the user experience in the takeaway scenario and make the recommended delay insurance package more in line with the user's wishes and the expected compensation indicator.

[0118] It should be understood that in this embodiment, the purchase module 103 can also only determine the purchase probability of each package for the delay insurance packages filtered out by the recommendation module 102” as the above-mentioned currently recommendable delay insurance packages, which can further save computing resources.

[0119] As Figure 4 shown, Figure 4 is a block diagram showing a data processing system according to another embodiment of the present disclosure. The recommendation module 102” can recommend the delay insurance package based on the purchase probabilities of the user for the filtered delay insurance packages in various ways. For example, the recommendation module 102” can recommend the delay insurance packages higher or lower than the predetermined purchase probability according to the predetermined purchase probability. Another example is that it can also recommend the top M (M is a positive integer) delay insurance packages sorted from high to low or from low to high according to the purchase probability, and so on. The ways of recommending the delay insurance package based on the predetermined compensation indicator, the determined compensation indicator, and the purchase probability are not limited to those exemplified above, and some ways not exemplified in the present disclosure are also possible without departing from the spirit scope of the present disclosure.

[0120] In another embodiment, the recommendation module 102” can also be configured to perform a descending order sorting on the filtered delay insurance packages based on the determined purchase probability of the user for at least one filtered delay insurance package, so as to recommend a predetermined number of delay insurance packages with top rankings. After the recommendation module 102” filters out one or more delay insurance packages from the package pool including Package 1, Package 2, …, Package N, whose compensation indexes of the determined delay insurance packages are higher or lower than those of the predetermined delay insurance packages, it performs a descending order sorting on each delay insurance package in the filtered package pool based on the purchase probability of the user for each delay insurance package in the filtered package pool determined by the purchase module, and recommends a predetermined number (for example, 2) of delay insurance packages with top rankings to the user. This predetermined number can be set by the user in advance, or can be automatically set by the data processing system through machine learning based on the user's previous multiple reservations.

[0121] For example, the recommendation module 102” has filtered out one or more delay insurance packages (for example, filtering out Package 2, 3, 4) whose compensation indexes of the determined delay insurance packages are 60% lower than those of the predetermined delay insurance packages. Then, the recommendation module 102” performs a descending order sorting on each of the filtered delay insurance packages based on the purchase probability of the user for each of the filtered delay insurance packages determined by the purchase module. In this example, since the purchase probabilities of Package 2, 3, 4 are 10%, 45%, 15% respectively, the filtered delay insurance packages will be sorted in descending order of purchase probability as Package 2, Package 4, and Package 3.

[0122] Thus, for example, when the predetermined number of the recommended delay insurance packages is 2 (i.e., M = 2), the recommendation module 102” will recommend the top 2 packages to the user, namely Package 2 and 4.

[0123] The above examples are only used to illustrate the inventive concept of the present disclosure, rather than to limit the present disclosure.

[0124] In this way, according to a further embodiment of the present disclosure, the data processing system preliminarily filters the delay insurance package pool according to the compensation indexes of the predetermined delay insurance packages, sorts the packages in descending order according to the purchase probability of the user for each of the filtered packages, and recommends the top predetermined number of delay insurance packages, which balances the capital compensation index while enhancing the user's purchase intention.

[0125] It should be understood that in this embodiment, the purchase module 103 may also not pre-determine the purchase probabilities of all delay insurance packages, but only determine the purchase probabilities of the delay insurance packages "filtered out" by the recommendation module 102, so as to save computing resources. The recommendation module 102" can recommend delay insurance packages based on the purchase probabilities of the filtered delay insurance packages in various ways other than those described in this disclosure. For example, the recommendation module 102" can recommend the delay insurance packages in the filtered delay insurance packages that are higher or lower than a predetermined purchase probability according to the predetermined purchase probability. Another example is that the recommendation module 102" can sort the delay insurance packages in the filtered package pool in ascending order according to the purchase probabilities of the user for each delay insurance package determined by the purchase module, and recommend a predetermined number (e.g., M) of delay insurance packages with the top rankings to the user. The methods of sorting and recommending delay insurance packages based on the predetermined compensation index, the determined compensation index, and the determined purchase probability are not limited to those exemplified above. Without departing from the spirit scope of this disclosure, some methods not exemplified in this disclosure are also possible.

[0126] Figure 11 is a block diagram showing a data processing device according to the present disclosure.

[0127] Reference Figure 11 , the data processing device 30 according to the present disclosure includes a processor 301 and a memory 302. The memory 302 is, for example, a computer storage medium. The memory 302 stores computer-executable instructions, and when the instructions are run by the processor, they execute the data processing method according to the present disclosure.

[0128] Of course, the above specific embodiments are only examples and not limitations, and those skilled in the art can combine and combine some steps and devices from the above separately described embodiments according to the concept of the present disclosure to achieve the effects of the present disclosure. Such combined and combined embodiments are also included in the present disclosure, and such combinations and combinations will not be described one by one here.

[0129] Note that the advantages, benefits, effects, etc. mentioned in this disclosure are only examples and not limitations, and it cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present disclosure. In addition, the above disclosed specific details are only for the purposes of illustration and easy understanding, rather than limitations. The above details do not limit the present disclosure to necessarily adopt the above specific details to implement.

[0130] The block diagrams of the systems, methods, and apparatuses involved in this disclosure are only exemplary examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these systems, methods, and apparatuses can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open-ended terms, meaning "including but not limited to", and can be used interchangeably with each other. The words "or" and "and" used herein refer to the word "and / or" and can be used interchangeably with it, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to" and can be used interchangeably with it.

[0131] The flowchart of steps in this disclosure and the above method descriptions are only exemplary examples and are not intended to require or imply that the steps of each embodiment must be performed in the order given. As those skilled in the art will recognize, the steps in the above embodiments can be performed in any order. Words such as "subsequently", "then", "next", etc. are not intended to limit the order of the steps; these words are only used to guide the reader through the description of these methods. In addition, any reference to a singular element using articles such as "a", "an", or "the" is not to be construed as limiting that element to the singular.

[0132] In addition, the steps and apparatuses in each of the embodiments herein are not limited to being implemented in a particular embodiment. In fact, according to the concepts of this disclosure, relevant partial steps and partial apparatuses in each of the embodiments herein can be combined to conceive new embodiments, and these new embodiments are also within the scope of this disclosure.

[0133] Each operation of the methods described above can be performed by any suitable means capable of performing the corresponding functions. Such means can include various hardware and / or software components and / or modules, including but not limited to hardware circuits, application-specific integrated circuits (ASICs), or processors.

[0134] The various illustrative logical blocks, modules, and circuits described can be implemented or performed using a general-purpose processor, a digital signal processor (DSP), an ASIC, a field-programmable gate array signal (FPGA), or other programmable logic device (PLD), discrete gate or transistor logic, discrete hardware components, or any combination thereof. The general-purpose processor can be a microprocessor, but alternatively, the processor can be any commercially available processor, controller, microcontroller, or state machine. The processor can also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors cooperating with a DSP core, or any other such configuration.

[0135] The steps of the methods or algorithms described in connection with the present disclosure may be directly embodied in hardware, in a software module executed by a processor, or in a combination of the two. The software modules may exist in any form of tangible storage medium. Some examples of storage media that may be used include random access memory (RAM), read only memory (ROM), flash memory, EPROM memory, EEPROM memory, registers, hard disks, removable disks, CD-ROM, etc. The storage medium may be coupled to the processor such that the processor can read information from, and write information to, the storage medium. In an alternative, the storage medium may be integral with the processor. The software modules may be single instructions or many instructions, and may be distributed over several different code segments, different programs, and across multiple storage media.

[0136] The methods disclosed herein include one or more acts for implementing the described methods. The methods and / or acts may be interchanged with one another without departing from the scope of the claims. In other words, unless a specific order of acts is specified, the order and / or use of specific acts may be modified without departing from the scope of the claims.

[0137] The above functions may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored as one or more instructions on a tangible computer-readable medium. The storage medium may be any available tangible medium accessible by a computer. By way of example, and not limitation, such computer-readable media may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other tangible medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. As used herein, disk and disc include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk, and Blu-ray disc, where disks usually reproduce data magnetically, while discs reproduce data optically with lasers.

[0138] Accordingly, a computer program product may perform the operations given herein. For example, such a computer program product may be a computer-readable tangible medium having tangible storage (and / or encoding) thereon of instructions executable by one or more processors to perform the operations described herein. The computer program product may include packaging materials.

[0139] Software or instructions may also be transmitted over a transmission medium. For example, the transmission medium may be used to transmit software from a website, server, or other remote source using a transmission medium such as coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, or microwave.

[0140] In addition, the modules and / or other suitable means for performing the methods and techniques described herein can be downloaded and / or otherwise obtained by a user terminal and / or a base station as appropriate. For example, such a device can be coupled to a server to facilitate the transfer of means for performing the methods described herein. Alternatively, the various methods described herein can be provided via a storage component (such as RAM, ROM, a physical storage medium such as a CD or a floppy disk) so that a user terminal and / or a base station can obtain the various methods when coupled to the device or provided with the storage component. In addition, any other suitable technique for providing the methods and techniques described herein to a device can be utilized.

[0141] Other examples and implementations are within the scope and spirit of the present disclosure and the appended claims. For example, due to the nature of software, the functions described above can be implemented using software executed by a processor, hardware, firmware, hardwiring, or any combination thereof. The features implementing the functions can also be physically located at various positions, including being distributed so that parts of the functions are implemented at different physical positions. Also, as used herein, including in the claims, the "or" used in the listing of items starting with "at least one" indicates a disjunctive listing so that, for example, the listing of "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Further, the phrase "exemplary" does not mean that the examples described are preferred or better than other examples.

[0142] Various changes, substitutions, and alterations to the techniques described herein can be made without departing from the teachings of the technology defined by the appended claims. In addition, the scope of the claims of the present disclosure is not limited to the specific aspects of the processes, machines, manufactures, compositions of events, means, methods, and acts described above. Current or later-developed processes, machines, manufactures, compositions of events, means, methods, or acts that perform substantially the same function or achieve substantially the same result as the corresponding aspects described herein can be utilized. Accordingly, the appended claims include such processes, machines, manufactures, compositions of events, means, methods, or acts within their scope.

[0143] The foregoing description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the present disclosure. Thus, the present disclosure is not intended to be limited to the aspects shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0144] The foregoing description has been presented for purposes of illustration and description. In addition, this description is not intended to limit embodiments of the present disclosure to the form disclosed herein. Although several example aspects and embodiments have been discussed above, those skilled in the art will recognize some variations, modifications, alterations, additions, and subcombinations thereof.

Claims

1. A data processing method, comprising: extracting user-related data, merchant-related data, environment-related data, and corresponding delay durations from historical orders; identifying a delay duration label corresponding to a delay duration interval identifier in one of a plurality of delay duration intervals into which the delay duration falls; performing feature encoding on the user-related data, the merchant-related data, and the environment-related data to obtain corresponding feature vectors; training a first preset machine learning model based on the feature vectors and the delay duration labels; in response to receiving a delay insurance package request corresponding to a delivery order, extracting a plurality of data of the delivery order, including user-related data, merchant-related data, and environment-related data of the current situation; performing feature encoding on the plurality of data extracted from the delivery order to obtain corresponding feature vectors, and inputting the encoded feature vectors into the trained first preset machine learning model to obtain the probabilities that the delay duration of the delivery order estimated by the first preset machine learning model falls into a plurality of delay duration intervals; determining claim indicators for a plurality of delay insurance packages based on the respective probabilities that the estimated delay duration of the delivery order falls into the plurality of delay duration intervals and the insured amounts and premiums of a plurality of delay insurance levels corresponding to the plurality of delay duration intervals, wherein each of the plurality of delay insurance packages includes a plurality of delay insurance levels corresponding to the plurality of delay duration intervals.

2. The method according to claim 1, wherein The determining claim indicators for a plurality of delay insurance packages based on the respective probabilities that the estimated delay duration of the delivery order falls into the plurality of delay duration intervals and the insured amounts and premiums of a plurality of delay insurance levels corresponding to the plurality of delay duration intervals includes: For each delay insurance package, using the respective probabilities of the plurality of delay duration intervals as weights, calculating a weighted sum of the respective insured amounts of the plurality of delay insurance levels corresponding to each delay insurance package to obtain the expected insured amount of each delay insurance package, and then dividing by the premium of each delay insurance package to calculate the claim indicator of each delay insurance package.

3. The method according to claim 1, further comprising: filtering out and recommending a first number of delay insurance packages whose claim indicators meet the predetermined claim indicators from the plurality of delay insurance packages based on the predetermined claim indicators.

4. The method according to claim 1, further comprising: identifying a delay insurance purchase label according to whether a user in a historical order has purchased a delay insurance package; performing feature encoding on the user-related data, the merchant-related data, the environment-related data, and the data related to the delay insurance package extracted from the historical order to obtain corresponding feature vectors; training a second preset machine learning model based on the feature vectors and the delay insurance purchase label; In response to receiving a delay insurance package request corresponding to a delivery order, extracting a plurality of data of the delivery order, including data related to a current user, data related to a current merchant, data related to a current environment, and data related to a current delay insurance package; Feature encoding is performed on multiple data extracted from the delivery order to obtain corresponding feature vectors, and the encoded feature vectors are input into the trained second preset machine learning model to obtain the probability of the user of the delivery order purchasing each delay insurance package estimated by the second preset machine learning model.

5. The method according to claim 4, further comprising: At least one delay insurance package among the multiple delay insurance packages is recommended based on the compensation indicators of the multiple delay insurance packages and the probability of the user purchasing each delay insurance package.

6. The method according to claim 5, wherein, The recommending at least one of the multiple delay insurance packages based on the compensation indicators of the multiple delay insurance packages and the probability of the user purchasing each delay insurance package includes: Based on the predetermined compensation index, filtering out a first number of delay insurance packages whose compensation indexes meet the predetermined compensation index from the multiple delay insurance packages; Based on the probability of the user purchasing each delay insurance package, the first number of delayed insurance packages filtered out are sorted in descending order to recommend a second number of delay insurance packages with higher rankings, wherein the second number is smaller than the first number.

7. The method according to claim 1, wherein the feature encoding step comprises: Normalizing, logarithmizing, taking square roots, and first encoding the data related to the user to obtain user features; Normalizing, logarithmizing, and taking square roots of the data related to the merchant, and performing a first encoding to obtain merchant characteristics; Normalizing, logarithmizing, and taking square roots, and first encoding the data related to the environment to obtain environmental features, wherein the data related to the environment is obtained by the following steps: extracting a location code of a destination of the order initiated by the user and a location code of a merchant from the data related to the user and the data related to the merchant, obtaining a route from the merchant to the destination based on the location code of the destination of the order initiated by the user and the location code of the merchant, and obtaining data related to the environment of the route based on the route; The data related to the delay insurance package are normalized, logarithmized, square-rooted, and first encoded to obtain the delay insurance package characteristics, Wherein, the first encoding includes: Determine the number of values a class of features can take; Construct the number of codes including the number of bits, so that each code corresponds to a value of the type of feature, and only one bit in each code is 1, and the other bits are 0, and the number of codes are mutually exclusive.

8. The method according to claim 1, wherein: The user-related data includes at least one of the geographical location of the destination of the order initiated by the user, the historical number of delays at the geographical location, the average delay duration, the number of insurance purchases and the number of claims of the user, the number of impressions of the user, the order amount, the user's delay insurance package, and the number of orders initiated by the user; The merchant-related data includes at least one of the geographical location of the merchant, the meal preparation speed of the merchant, the delivery method of the merchant, the order volume of the merchant, the number of visits to the merchant, the historical number of delays of the merchant, the average delay duration of the merchant, the number of insurance purchases for the merchant, and the number of claims for the merchant; The data related to the environment of the route includes at least one of the weather condition, traffic condition, and time information during the delivery of the order.

9. A data processing system, comprising: A first extraction device configured to extract user-related data, merchant-related data, environment-related data, and the corresponding delay duration from historical orders; An identification device configured to identify a corresponding delay duration label according to a delay duration interval in which the delay duration falls among a plurality of delay duration intervals; An encoding device configured to perform feature encoding on the user-related data, the merchant-related data, and the environment-related data to obtain corresponding feature vectors; A training device configured to train a first preset machine learning model based on the corresponding feature vectors and the corresponding delay duration labels; A second extraction device configured to, in response to receiving a delay insurance package request corresponding to a delivery order, extract a plurality of data of the delivery order, including user-related data, merchant-related data, and environment-related data related to the current user; A probability acquisition device configured to perform feature encoding on the plurality of data extracted from the delivery order to obtain corresponding feature vectors, and input the encoded feature vectors into the trained first preset machine learning model to obtain the probabilities that the delay duration of the delivery order estimated by the first preset machine learning model falls into a plurality of delay duration intervals; A determination device configured to determine the compensation indicators of a plurality of delay insurance packages based on the respective probabilities that the estimated delay duration of the delivery order falls into the plurality of delay duration intervals and the insured amounts and premiums of a plurality of delay insurance grades corresponding to the plurality of delay duration intervals, wherein each of the plurality of delay insurance packages includes a plurality of delay insurance grades corresponding to the plurality of delay duration intervals.

10. A computer storage medium storing computer-executable instructions that, when run by a processor, execute the method according to any one of claims 1-8.

Images