Recommended data sending method and device, storage medium and electronic device

Abstract

The application discloses a kind of recommended data sending method and device, storage medium and electronic device, wherein the above-mentioned method includes: determining the consumed data of consumer object from the collected consumption data;Determine the multidimensional prediction model corresponding to the object type of consumer object;Consumed data is input into multidimensional prediction model to determine the consumption preference probability of consumer object;Wherein, the training data of multidimensional prediction model at least includes: the historical consumed data of consumer object, the consumption preference probability corresponding to historical consumed data;Corresponding recommended data is obtained with consumption preference probability, and recommended data is sent to consumer object.The above technical scheme is used, and the problem of how to improve the matching degree of recommended data and user is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data analysis, in particular to a recommended data sending method and device, a storage medium and an electronic device. BACKGROUND

[0002] At present, after entering the digital economy era, data has become an important production factor. In order to improve the traffic operation ability of financial institutions, it is necessary to adapt the digital resources of users. In related technologies, the cosine theorem formula is generally used to calculate the similarity between products, and then similar product data is recommended to users. However, this method cannot accurately push the recommended data to the target user, resulting in low matching degree of recommended data and users.

[0003] For the problem of how to improve the matching degree of recommended data and users in related technologies, there is no effective solution at present.

[0004] Therefore, it is necessary to improve the related technology to overcome the defects in the related technology. SUMMARY

[0005] The embodiments of the present application provide a recommended data sending method and device, a storage medium and an electronic device to at least solve the problem of how to improve the matching degree of recommended data and users.

[0006] According to an aspect of an embodiment of the present application, a recommended data sending method is provided, comprising: determining the consumed data of a consumption object from the collected consumption data; determining a multi-dimensional prediction model corresponding to the object type of the consumption object; inputting the consumed data into the multi-dimensional prediction model to determine the consumption preference probability of the consumption object; wherein the training data of the multi-dimensional prediction model at least includes: the historical consumed data of the consumption object, and the consumption preference probability corresponding to the historical consumed data; obtaining the recommended data corresponding to the consumption preference probability, and sending the recommended data to the consumption object.

[0007] In one exemplary embodiment, the consumed data of the consumption object is determined from the collected consumption data, comprising: determining the consumption behavior data and public behavior data contained in the consumption data, wherein the public behavior data represents the behavior data corresponding to the public behavior of the consumption object in a public place, and the consumption behavior data and the public behavior data both correspond to the object number of the consumption object; determining the target consumption behavior data and the target public behavior data corresponding to the target object number; in the case that the target object number belongs to the preset object information table, determining the target consumption behavior data and the target public behavior data as the consumed data of the consumption object.

[0008] In an example embodiment, before determining the multi-dimensional prediction model corresponding to the object type of the consumption object, the method further comprises: obtaining a first consumption behavior record, a second consumption behavior record and a redemption activity record contained in the consumed data; wherein the first consumption behavior record contains historical consumption behavior and current consumption behavior, the second consumption behavior record contains only the current consumption behavior, and the redemption activity record is a record of the consumption object participating in a redemption activity in a historical time period; determining the object type of the consumption object with the first consumption behavior record and / or the redemption activity record as a registered object, wherein the registered object includes at least one of the following: a first registered object, a second registered object and a third registered object, the first consumption frequency of the first registered object is greater than or equal to a first preset value; the second consumption frequency of the second registered object is less than the first preset value and greater than or equal to a second preset value, and the third consumption frequency of the third registered object is less than the second preset value; and determining the object type of the consumption object with the second consumption behavior record as an unregistered object, wherein the registration probability of the unregistered object after a preset time period is greater than a third preset value.

[0009] In an example embodiment, after determining the multi-dimensional prediction model corresponding to the object type of the consumption object, the method further comprises: obtaining a similarity between the unregistered object and the registered object; in a case where the similarity is greater than a preset threshold, obtaining a first multi-dimensional prediction model corresponding to the registered object, and determining the first multi-dimensional prediction model as a second multi-dimensional prediction model corresponding to the unregistered object; and / or, in a case where the similarity is greater than a preset threshold, determining a target registered object from a plurality of registered objects, obtaining a third multi-dimensional prediction model corresponding to the target registered object, and determining the third multi-dimensional prediction model as a fourth multi-dimensional prediction model corresponding to the unregistered object; wherein the similarity between the target registered object and the unregistered object is greater than the similarity between the target registered object and other unregistered objects, and the other unregistered objects are registered objects in the plurality of registered objects except the target registered object.

[0010] In an example embodiment, before determining the multi-dimensional prediction model corresponding to the object type of the consumption object, the method further comprises: classifying the historical consumed data according to a preset consumption type to obtain multi-dimensional data; performing normalization processing on the multi-dimensional data to obtain a processing result; performing feature extraction on the processing result to obtain a plurality of feature vectors, wherein the plurality of feature vectors correspond to different dimensions; performing correlation analysis on the plurality of feature vectors and the consumption preference probability to obtain a correlation degree between the plurality of feature vectors and the consumption preference probability; determining a target feature vector with a correlation degree greater than a second preset threshold between the consumption preference probability using a component analysis algorithm; and training the multi-dimensional prediction model using the target feature vector as an input sample and using the consumption preference probability as an output sample.

[0011] In an example embodiment, classifying the historical consumed data according to a preset consumption type to obtain multi-dimensional data comprises: classifying the historical consumed data using a plurality of preset consumption types to obtain a plurality of first-level dimensional data, wherein the preset consumption types correspond to consumption weight values, and each first-level dimensional data corresponds to a preset consumption type; labeling each first-level dimensional data using a first-level label corresponding to the preset consumption type, wherein there is a corresponding relationship between the labeled first-level dimensional data and the first-level label; binding the labeled first-level dimensional data and the consumption weight value corresponding to the labeled first-level dimensional data to obtain second-level dimensional data; and determining the multi-dimensional data according to a plurality of second-level dimensional data.

[0012] In an example embodiment, before determining the multi-dimensional prediction model corresponding to the object type of the consumption object, the method further comprises: obtaining a historical consumption frequency from the historical consumed data, and determining the consumption preference probability according to the size of the historical consumption frequency; or obtaining a historical exchange record from the historical consumed data, and determining the consumption preference frequency according to the exchange times in the historical exchange record; or obtaining the historical consumption frequency and the historical exchange record from the historical consumed data, and determining the consumption preference probability according to a weighted sum of the historical consumption frequency and the exchange times in the historical exchange record.

[0013] In an example embodiment, the acquiring the recommendation data corresponding to the consumption preference probability comprises: acquiring a plurality of consumption preference probabilities obtained by inputting the consumed data into a plurality of multi-dimensional prediction models; wherein the plurality of multi-dimensional prediction models are obtained by using different training algorithms and have different model structures; determining a target consumption preference probability from the plurality of consumption preference probabilities, wherein the target consumption preference probability is the maximum value; and acquiring the recommendation data corresponding to the target consumption preference probability.

[0014] According to another aspect of the embodiments of the present application, a recommendation data sending device is further provided, which comprises: a data determining module configured to determine consumed data of a consumption object from collected consumption data; a model determining module configured to determine a multi-dimensional prediction model corresponding to an object type of the consumption object; a probability determining module configured to input the consumed data into the multi-dimensional prediction model to determine a consumption preference probability of the consumption object; wherein training data of the multi-dimensional prediction model at least comprises historical consumed data of the consumption object and a consumption preference probability corresponding to the historical consumed data; and a result sending module configured to acquire recommendation data corresponding to the consumption preference probability and send the recommendation data to the consumption object.

[0015] According to still another aspect of the embodiments of the present application, a computer readable storage medium is further provided, which stores a computer program, wherein the computer program is configured to execute the recommendation data sending method when running.

[0016] According to still another aspect of the embodiments of the present application, an electronic device is further provided, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the recommendation data sending method through the computer program.

[0017] Through the present application, the consumed data of a consumption object can be acquired from collected consumption data, and then the consumed data is input into a multi-dimensional prediction model corresponding to an object type of the consumption object to obtain a consumption preference probability of the consumption object; wherein the multi-dimensional prediction model is trained by using training data at least comprising historical consumed data of the consumption object and a consumption preference probability corresponding to the historical consumed data; and after acquiring recommendation data corresponding to the consumption preference probability, the recommendation data is sent to the consumption object. By using the above technical solution, the problem of how to improve the matching degree of recommendation data and users in related technologies is solved, and the effect of improving the matching degree of recommendation data and users can be achieved. BRIEF DESCRIPTION OF DRAWINGS

[0018] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:

[0019] Figure 1 is a hardware structure block diagram of a computer terminal for executing the recommended data sending method according to an embodiment of the present application;

[0020] Figure 2 is a flow chart of the recommended data sending method according to an embodiment of the present application;

[0021] Figure 3 is a schematic flow chart (I) of the recommended data sending method according to an embodiment of the present application;

[0022] Figure 4 is a schematic flow chart (II) of the recommended data sending method according to an embodiment of the present application;

[0023] Figure 5 is a schematic flow chart (III) of the recommended data sending method according to an embodiment of the present application;

[0024] Figure 6 is a schematic diagram of a consumption data classification result according to an embodiment of the present application;

[0025] Figure 7 is a schematic flow chart (IV) of the recommended data sending method according to an embodiment of the present application;

[0026] Figure 8 is a flow chart of a multi-dimension prediction model training process according to an embodiment of the present application;

[0027] Figure 9 is a schematic diagram of a multi-dimension prediction model running process according to an embodiment of the present application;

[0028] Figure 10 is a structure block diagram of a recommended data sending device according to an embodiment of the present application. DETAILED DESCRIPTION

[0029] In order to make the personnel in the art better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by the personnel in the art without creative labor should belong to the protection scope of the present application.

[0030] It should be noted that the terms and terms such as "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0031] The methods and embodiments provided in this invention can be executed on a computer terminal or similar computing device. Taking running on a computer terminal as an example, Figure 1 This is a hardware structure block diagram of a computer terminal that executes the method for sending recommended data according to embodiments of the present invention. For example... Figure 1 As shown, a computer terminal may include one or more ( Figure 1 Only one is shown in the diagram. The processor 102 (which may include, but is not limited to, a microprocessor unit (MPU) or a programmable logic device (PLD)) and a memory 104 configured to store data are also included. In one exemplary embodiment, the computer terminal may further include a transmission device 106 configured for communication and an input / output device 108. Those skilled in the art will understand that… Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the computer terminal described above. For example, the computer terminal may also include components that are more complex than those described above. Figure 1 The more or fewer components shown, or having the same Figure 1 Equivalent functions or ratios shown Figure 1 The functions shown have more different configurations.

[0032] The memory 104 can be configured to store computer programs, such as software programs of application software and modules, such as a computer program corresponding to the recommendation data sending method in the embodiments of the present application. The processor 102 can execute various functional applications and data processing, i.e., implement the above method, by running the computer programs stored in the memory 104. The memory 104 can include a high-speed random access memory, and can further include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some examples, the memory 104 can further include memories remotely arranged with respect to the processor 102, which can be connected to the computer terminal through a network. Examples of the above network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0033] The transmission device 106 is configured to receive or send data via a network. A specific example of the above network can include a wireless network provided by a communication provider of the computer terminal. In one example, the transmission device 106 includes a network adapter (NIC), which can be connected to other network devices through a base station so as to be capable of communicating with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is configured to communicate with the Internet in a wireless manner.

[0034] In the present embodiment, a recommendation data sending method is provided, Figure 2 According to the flowchart of the recommendation data sending method in the embodiments of the present application, the flow includes the following steps:

[0035] In step S202, the consumed data of the consumption object is determined from the collected consumption data.

[0036] In step S204, the multi-dimensional prediction model corresponding to the object type of the consumption object is determined.

[0037] In step S206, the consumed data is input into the multi-dimensional prediction model to determine the consumption preference probability of the consumption object; wherein the training data of the multi-dimensional prediction model at least includes the historical consumed data of the consumption object and the consumption preference probability corresponding to the historical consumed data.

[0038] In step S208, the recommendation data corresponding to the consumption preference probability is obtained, and the recommendation data is sent to the consumption object.

[0039] The consumed data of the consumption object is determined from the collected consumption data through the above steps; a multi-dimensional prediction model corresponding to the object type of the consumption object is determined; the consumed data is input into the multi-dimensional prediction model to determine the consumption preference probability of the consumption object; wherein the training data of the multi-dimensional prediction model at least includes: historical consumed data of the consumption object, and consumption preference probability corresponding to the historical consumed data; the recommendation data corresponding to the consumption preference probability is obtained, and the recommendation data is sent to the consumption object. The above embodiment solves the problem of how to improve the matching degree of recommendation data and users in the related art, thereby achieving the effect of improving the matching degree of recommendation data and users.

[0040] In one example embodiment, for the implementation process of determining the consumed data of the consumption object from the collected consumption data in step S202, it can further include: step S11, determining consumption behavior data and public behavior data contained in the consumption data, wherein the public behavior data represents behavior data corresponding to public behavior of the consumption object in a public place, and the consumption behavior data and the public behavior data both correspond to the object number of the consumption object; step S12, determining target consumption behavior data and target public behavior data corresponding to the target object number; step S13, in the case of determining that the target object number belongs to the preset object information table, determining the target consumption behavior data and the target public behavior data as the consumed data of the consumption object.

[0041] Optionally, in the above embodiment, taking the bank field as an example, the above consumption behavior data can be obtained from bank internal information, including but not limited to customer personal information (gender, age, assets, liabilities, income level, occupation, account transaction flow, consumption flow, etc.), usage data of bank-related application software such as online bank, mobile bank (digital resource exchange data, flow operation activity participation data, etc.); the above public behavior data can be obtained from third-party platforms cooperating with the bank, including but not limited to customer education, flight record, tax data, shopping information.

[0042] Optionally, in the above embodiment, a full-amount customer information wide table (not containing sensitive information such as customer name, mobile phone number, certificate number, address) with customer number (equivalent to object number) as unique identification code can be established as a preset object information table, and the above obtained consumption behavior data and public behavior data are stored in the preset object information table.

[0043] In an example embodiment, before performing the above step S204, the method further comprises the following processes: step S21, obtaining a first consumption behavior record, a second consumption behavior record and a redemption activity record contained in the consumed data; wherein the first consumption behavior record contains a historical consumption behavior and a current consumption behavior, the second consumption behavior record contains only the current consumption behavior, and the redemption activity record is a record of the consumption object participating in a redemption activity in a historical time period; step S22, determining an object type of the consumption object with the first consumption behavior record and / or the redemption activity record as a registered object, wherein the registered object at least includes one of the following: a first registered object, a second registered object and a third registered object, a first consumption frequency of the first registered object is greater than or equal to a first preset value; a second consumption frequency of the second registered object is less than the first preset value and greater than or equal to a second preset value, and a third consumption frequency of the third registered object is less than the second preset value; step S23, determining an object type of the consumption object with the second consumption behavior record as an unregistered object, wherein a registration probability of the unregistered object after a preset time period is greater than a third preset value.

[0044] Optionally, it should be noted that the first preset value, the second preset value and the third preset value are all natural numbers. The registration probability of the unregistered object after the preset time period can be understood as a probability of the unregistered object choosing to register after a period of time.

[0045] Optionally, the redemption activity includes but is not limited to a rights library redemption behavior, a redemption rights detail, a rights point change behavior and the like.

[0046] Optionally, in the above embodiment, if the current consumption object is not registered, i.e., the preset object information table does not contain the current object information, and the historical consumption behavior cannot be queried through the object number, it is considered that the current object is a cold start customer (equivalent to an unregistered object).

[0047] Optionally, in the above embodiment, for example, the first preset value is 2, the second preset value is 1, and the third preset value is 0, the registered object with a consumption frequency greater than or equal to 2 times is regarded as a regular customer (equivalent to the first registered object), the registered object with a consumption frequency greater than or equal to 1 time and less than 2 times is regarded as an ordinary customer (equivalent to the second registered object), and the registered object with a consumption frequency equal to 0 times is regarded as a new customer (equivalent to the third registered object).

[0048] In an example embodiment, the implementation process of the multi-dimensional prediction model corresponding to the object type of the consumption object determined in step S204 can specifically include: obtaining the similarity between the unregistered object and the registered object; in a case where the similarity is greater than a preset threshold, obtaining a first multi-dimensional prediction model corresponding to the registered object, and determining the first multi-dimensional prediction model as a second multi-dimensional prediction model corresponding to the unregistered object; and / or, in a case where the similarity is greater than a preset threshold, determining a target registered object from a plurality of registered objects, obtaining a third multi-dimensional prediction model corresponding to the target registered object, and determining the third multi-dimensional prediction model as a fourth multi-dimensional prediction model corresponding to the unregistered object; wherein the similarity between the target registered object and the unregistered object is greater than the similarity between the target registered object and other unregistered objects, and the other unregistered objects are registered objects in the plurality of registered objects except the target registered object.

[0049] Optionally, in the above embodiment, the similarity between the unregistered object and the registered object can be obtained by: calculating the similarity of the consumption data of the two objects in different dimensions based on a collaborative filtering algorithm; and manually determining.

[0050] In an example embodiment, before performing step S204, the method can further include the following steps: step S31, classifying the historical consumption data according to a preset consumption type to obtain multi-dimensional data; step S32, performing normalization processing on the multi-dimensional data to obtain a processing result; step S33, performing feature extraction on the processing result to obtain a plurality of feature vectors, wherein the plurality of feature vectors correspond to different dimensions; step S34, performing correlation analysis on the plurality of feature vectors and the consumption preference probability to obtain the correlation degree between the plurality of feature vectors and the consumption preference probability; step S35, using a component analysis algorithm to determine a target feature vector whose correlation degree with the consumption preference probability is greater than a second preset threshold; and step S36, using the target feature vector as an input sample and using the consumption preference probability as an output sample to train the multi-dimensional prediction model.

[0051] Optionally, the component analysis algorithm can include a principal component analysis algorithm, but is not limited thereto.

[0052] Optionally, in the above embodiment, the multi-dimensional data can be understood as including data of clothing, food, housing, transportation, shopping, education expenses, entertainment consumption, phone top-up, and car owner consumption.

[0053] Optionally, in the above embodiment, before the multi-dimensional data is normalized, the data can also be pre-processed by calculating the maximum value, minimum value, average value and median value of the multi-dimensional data, checking the rationality and deleting unreasonable data; for example, if the age is 150 years old, it is determined that the data is unreasonable and the data is deleted.

[0054] In an example embodiment, for the implementation process of classifying the historical consumed data according to the preset consumption types in the above step S31 to obtain the multi-dimensional data, it can specifically include: classifying the historical consumed data using a plurality of preset consumption types to obtain a plurality of first-level dimension data, wherein the preset consumption types correspond to consumption weight values, and each first-level dimension data corresponds to a preset consumption type; labeling each first-level dimension data using a first-level label corresponding to the preset consumption type, wherein there is a corresponding relationship between the labeled first-level dimension data and the first-level label; binding the labeled first-level dimension data and the consumption weight values corresponding to the labeled first-level dimension data to obtain second-level dimension data; and determining the multi-dimensional data according to a plurality of second-level dimension data.

[0055] In an example embodiment, before the above step S204 is performed, the method can further include: obtaining a historical consumption frequency from the historical consumed data, and determining the consumption preference probability according to the size of the historical consumption frequency; or obtaining a historical exchange record from the historical consumed data, and determining the consumption preference frequency according to the exchange times in the historical exchange record; or obtaining the historical consumption frequency and the historical exchange record from the historical consumed data, and determining the consumption preference probability according to the weighted sum of the historical consumption frequency and the exchange times in the historical exchange record.

[0056] In an example embodiment, for the implementation process of obtaining the recommended data corresponding to the consumption preference probability in the above step S204, it can specifically include the following steps: step S41, obtaining a plurality of consumption preference probabilities obtained by inputting the consumed data into a plurality of multi-dimensional prediction models; wherein the plurality of multi-dimensional prediction models are obtained by training using different training algorithms and have different model structures; step S42, determining a target consumption preference probability from the plurality of consumption preference probabilities, wherein the target consumption preference probability is the maximum value; and step S43, obtaining recommended data corresponding to the target consumption preference probability.

[0057] Optionally, in the above embodiment, the plurality of multi-dimensional prediction models can include, for example, a logistic regression model, an XGB model, a decision tree model and a collaborative filtering model.

[0058] Obviously, the above-described embodiments are only a part of the embodiments of the present application, not all the embodiments. In order to better understand the above-mentioned recommendation data sending method, the above process is described below in combination with the embodiments, but not used to limit the technical solutions of the embodiments of the present application, specifically:

[0059] In an optional embodiment, Figure 3 is a schematic flowchart of the recommendation data sending method according to an embodiment of the present application (one), and the recommendation data sending method of the present embodiment will be described below in combination with the model training steps, and the specific steps are as follows: Figure 3 The recommendation data sending method of the present embodiment is described from the perspective of business process, and the specific steps are as follows:

[0060] Step S302: Formulate an analysis framework, determine the customer range and the collection direction of consumption data.

[0061] Step S304: Divide the types and dimensions of consumption data.

[0062] Step S306: Determine the corresponding relationship between different consumption dimensions and consumption types, and construct different Y variables (equivalent to consumption preference probability).

[0063] Step S308: Process the consumption data of the customer, extreme value processing, normalization processing and other feature engineering, calculate the correlation degree of the consumption dimension and the Y variable according to the correlation analysis, and reduce the dimension to synthesize the main Y variable according to the principal component analysis method.

[0064] Step S310: Classify the customer according to the consumed data, and train multiple multi-dimensional prediction models using collaborative filtering, logistic regression, decision tree and other algorithms according to the customer type.

[0065] Step S312: Select the corresponding optimal model to calculate the customer's favorite consumption scene and digital resources.

[0066] Through the above steps, the customer activity can be promoted, the customer relationship can be deepened, the customer value creation can be improved, the comprehensive coverage of digital ecological service can be realized in the way of customer stratified management, the value mining of the inventory customer can be increased, and the platform management capability can be improved.

[0067] In an optional embodiment, Figure 4 is a schematic flowchart of the recommendation data sending method according to an embodiment of the present application (two), and the recommendation data sending method of the present embodiment will be described below in combination with the model training steps, and the specific steps are as follows:

[0068] Step S402: Formulate an analysis framework, determine the customer range and the collection direction of consumption data.

[0069] Step S404: Collect the consumption behavior data and public behavior data of the customers using the big data intelligent platform and store them in the database.

[0070] Step S406: Process the collected data to obtain multi-dimensional data, and perform extreme value processing and normalization processing on the multi-dimensional data to obtain an independent variable index library (equivalent to multiple feature vectors).

[0071] Step S408: Divide the processed independent variable index library into a training set, a test set, and a validation set, use the training set and the test set for multi-dimensional prediction model training, train according to customer types using collaborative filtering, logistic regression, decision tree, and other algorithms to obtain multiple multi-dimensional prediction models, and use the validation set to verify the trained multi-dimensional prediction models.

[0072] The above-mentioned collaborative filtering (Collaborative Filtering) algorithm is a widely used recommendation algorithm in the field of recommendation systems, which mainly recommends based on the similarity between users. This algorithm first calculates the similarity between users, and usually uses cosine similarity or Pearson correlation coefficient to measure the similarity between users. Then, find some users most similar to the current user, and recommend items to the current user according to the behavior data of these users.

[0073] Step S410: Package the trained multiple multi-dimensional prediction models, and select the optimal model corresponding to the customer type for prediction.

[0074] Step S412: Send the customer's favorite consumption scenarios and digital resources predicted by the optimal model to the customer.

[0075] Through the above steps, the digital operation concept solves the technical problem of traffic operation and digital resource adaptation of financial institutions, promotes traffic conversion, and establishes an application scenario with traffic sedimentation effect.

[0076] In an optional embodiment, Figure 5 is a schematic flowchart of a method for sending recommendation data according to an embodiment of the application (three), which will be described below in combination with Figure 5 The process of collecting consumption data and dividing consumption dimensions in this embodiment is described as follows:

[0077] Step S502: Determine the customer range, divide the customer range into overall customers, customer groups, and individual customers according to granularity, and obtain the customer consumption data collection time period, such as collecting the traffic operation activity participation and digital resource redemption of the customers in the past year.

[0078] Optionally, the process of determining the customer range can set a low-frequency threshold by examining the frequency distribution of customer consumption behavior and various consumption scenarios.

[0079] Optionally, the process of dividing the customer range can be divided by examining the relevance of the customer, whether the customer has a certain tendency, and the customer life cycle.

[0080] Step S504: Divide the consumption data dimension, and correspond the collected consumption data to the consumption dimension.

[0081] Step S506: Determine the Y variable from the consumption behavior according to the correlation analysis, for example, by constructing the Y variable in the following way: determining the customer consumption preference probability according to the consumption frequency or the number of exchanges of the customer historical consumption data; weighting the consumption frequency and the number of exchanges to obtain the customer consumption preference probability.

[0082] Step S508: Determine the consumption dimension range, for example, classify around the clothing, food, housing, transportation, shopping, education, and entertainment consumption dimensions, and the dimensions are as follows: clothing (clothing, accessories, luggage), food (dining), housing (furniture, building materials, decoration, real estate rental and sale), travel (travel, refueling, transportation), shopping (online shopping, shopping malls), education (schools, training and examination, online reading, academic papers, overseas study), and entertainment (online video, online music, games, amusement parks, etc.).

[0083] Optionally, in combination with Figure 6 The consumption dimension range in step S504 is described, Figure 6 is a schematic diagram of a consumption data classification result according to an embodiment of the application, as Figure 6 shown, the above-mentioned consumption dimension range is, for example, 9 dimensions of clothing, food, housing, transportation, shopping, education expenditure, entertainment consumption, phone charge, and car owner consumption, each first-level label is subdivided by consumption data, for example, some groups, Hema, and Hungry are classified as second-level labels (equivalent to first-level labels) of takeout ordering.

[0084] Step S510: Divide the customer range into active customers, ordinary customers, and no-record cold-start customers according to the activity level, and standardize the data according to the consumption preference.

[0085] Optionally, the data standardization process includes determining the necessity according to the phone charge interval time, the number of charges, and the amount, and regarding some irregular consumption behavior as a preference when the amount and the number of charges are both higher than the average or median of the same group.

[0086] Step S512: Train multiple multi-dimensional prediction models using collaborative filtering, logistic regression, decision tree, and other algorithms according to the customer type, and encapsulate the models.

[0087] In an optional embodiment, the processing of the consumption data is combined with the processing of the transaction data. Figure 7 The processing of the consumption data is described below, Figure 7 is a schematic flowchart of a method for sending recommendation data according to an embodiment of the present application (four), as shown in the figure, the specific steps are as follows: Figure 7

[0088] Step S702: The collected consumption data is classified to obtain multi-dimensional data.

[0089] Step S704: The multi-dimensional data is investigated to view data distribution, maximum value, minimum value, mean value, median, etc., and the data rationality is checked.

[0090] Step S706: The multi-dimensional data is processed to delete unreasonable data (e.g., age 150 years old) and supplement null data.

[0091] Step S708: The multi-dimensional data is normalized to obtain a processing result, and the processing result is feature extracted to obtain a plurality of feature vectors.

[0092] Step S710: The plurality of feature vectors are analyzed for correlation by using a Pearson correlation coefficient, and a principal component analysis method is used to determine a feature vector with a high degree of correlation with a consumption preference probability.

[0093] Step S712: The feature vector with a high degree of correlation with the consumption preference probability is added to a model training data set.

[0094] Through the above steps, different marketing scenarios, consumption transaction frequencies, and differences in digital marketing activities can be considered in the data quantization process, and various activity factors such as customer levels, consumption dimensions, activity information, and benefit redemption information can be integrated to improve the accuracy of traffic management and digital resource adaptation. According to the activity effect data and feedback information, iterative updating is performed in a timely manner to obtain different digital resource probabilities, and comparison between consumption preferences is realized in the form of scores.

[0095] In an optional embodiment, Figure 8 is a flowchart of a multi-dimensional prediction model training process according to an embodiment of the present application, and the multi-dimensional prediction model training process is described below, Figure 8 the specific steps are as follows:

[0096] Step S802: The collected consumption data is used as sample data.

[0097] Step S804: The processed sample data is divided into a training set, a test set, and a validation set in a ratio of 65%, 25%, and 10% by using a PYTHON program. ​

[0098] Step S806: Classify the customers, divide the customers containing multiple historical consumption behaviors into regular customers (equivalent to the first registered object), divide the customers containing one historical consumption behavior into ordinary customers (equivalent to the second registered object), divide the customers containing only historical exchange records into new customers (equivalent to the third registered object), and divide the customers containing no historical consumption behaviors and historical exchange records into cold-start customers (equivalent to the unregistered object).

[0099] Step S808: Use the above data set to train the multi-dimensional prediction model.

[0100] Step S810: Optimize the multi-dimensional prediction model based on algorithms such as logistic regression, XGB, decision tree, and collaborative filtering to obtain multiple multi-dimensional prediction models.

[0101] Step S812: Package the optimized multi-dimensional prediction model and store it in the model library.

[0102] Optionally, the above model packaging process implementation includes: combining the variable feature library and the optimal model, and using PYTHON function to package into a ".py" file under the my_package directory, or using IMPORT to package into a pickle file; when digital resources need to be configured, the packaged model is used for prediction; when new data sets need to be adapted, use PYTHON to call the packaging function for feature engineering and model training.

[0103] Through the above steps, models can be established for various consumption dimensions such as clothing, food, housing, transportation, shopping, education, and entertainment based on different training sets and test sets, and a digital resource intelligent matching comprehensive model can be constructed through model fusion, and the constructed model can be packaged for easy calling at any time.

[0104] In an optional embodiment, Figure 9 is a schematic diagram of a multi-dimensional prediction model running process according to an embodiment of the application. The multi-dimensional prediction model running process of the present embodiment will be described below with reference to Figure 8 The specific process is as follows:

[0105] As shown in Figure 9 , the multi-dimensional prediction model first classifies the customers, divides the customers into registered objects (such as Figure 9 traffic entity customers and seed entity customers) and unregistered objects (such as Figure 9For cold-start customers, the consumption preferences of registered users are predicted using a multi-algorithm fusion model combining logistic regression and decision trees to obtain multiple consumption preference probabilities. The label corresponding to the highest consumption preference probability is then determined. For unregistered customers, information on registered customers with similar user profiles is obtained. The consumption preference probabilities of similar users are determined using a collaborative filtering model, and then the labels corresponding to the consumption preference probabilities are determined. Recommended content corresponding to the labels is then sent to the customers.

[0106] Optionally, for example, if the cold start user is A, first find a customer group B that is similar to the cold start user A in dimensions such as clothing, food, housing, transportation, shopping, education expenditure, entertainment consumption, mobile phone top-up, and car owner consumption. If a large number of customers in customer group B like to watch Tmall live broadcasts, the probability of customer group B's consumption preference for live broadcasts is obtained to the maximum value through a multi-dimensional prediction model. Then, obtain recommendation data related to live broadcasts, such as live broadcast channels of a certain platform, and recommend the live broadcast channels of that platform to the cold start user A.

[0107] Through the above process, we can not only analyze multi-dimensional customer data, but also consider factors such as new and old customers and cold start customers, and use algorithms such as collaborative filtering, decision trees, and logistic regression to build predictive models; then we can label different types of customers to improve the accuracy of digital resource matching.

[0108] Through the above-mentioned multiple embodiments, product information can be recommended to users based on user consumption data in multi-dimensional consumption scenarios. First, it can intuitively identify the business profile of customers, which helps business departments to take more targeted marketing measures and conduct personalized "advertising marketing" on various online channels. This enables intelligent matching of consumers and resources and intensive use of data resources, greatly improving marketing efficiency. Second, it establishes an application scenario with traffic retention effect, leveraging the "funnel effect" of new finance. It identifies customer needs for different customer types, guides customer behavior through traffic management activities, and converts traffic into "retained customers".

[0109] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0110] A sending device of recommendation data is also provided in the embodiment, which is used to implement the above-mentioned embodiments and preferred embodiments, and will not be described herein. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, hardware, or a combination of software and hardware can also be implemented and conceived.

[0111] Figure 10 is a structural block diagram of a sending device of recommendation data according to an embodiment of the application, which comprises:

[0112] A data determining module 1002 is configured to determine consumed data of a consumption object from collected consumption data.

[0113] A model determining module 1004 is configured to determine a multi-dimensional prediction model corresponding to an object type of the consumption object.

[0114] A probability determining module 1006 is configured to input the consumed data into the multi-dimensional prediction model to determine a consumption preference probability of the consumption object, wherein training data of the multi-dimensional prediction model at least comprises historical consumed data of the consumption object and a consumption preference probability corresponding to the historical consumed data.

[0115] A result sending module 1008 is configured to obtain recommendation data corresponding to the consumption preference probability and send the recommendation data to the consumption object.

[0116] The sending device of recommendation data according to the embodiment of the application can obtain consumed data of a consumption object from collected consumption data, and then input the consumed data into a multi-dimensional prediction model corresponding to an object type of the consumption object to obtain a consumption preference probability of the consumption object, wherein the multi-dimensional prediction model can be trained by training data at least comprising historical consumed data of the consumption object and a consumption preference probability corresponding to the historical consumed data, and after obtaining recommendation data corresponding to the consumption preference probability, the recommendation data is sent to the consumption object. The problem of how to improve the matching degree of recommendation data and users in the related art is solved, and the effect of improving the matching degree of recommendation data and users can be achieved.

[0117] The embodiment of the application further provides a storage medium comprising a stored program, wherein the program performs any of the above-mentioned embodiments when running.

[0118] Optionally, in the embodiment, the storage medium can be configured to store a computer program for performing the following steps:

[0119] S1, determining consumed data of a consumption object from collected consumption data.

[0120] S2, determining a multi-dimensional prediction model corresponding to the object type of the consumption object;

[0121] S3, inputting the consumed data into the multi-dimensional prediction model to determine a consumption preference probability of the consumption object; wherein training data of the multi-dimensional prediction model at least includes historical consumed data of the consumption object and a consumption preference probability corresponding to the historical consumed data;

[0122] S4, obtaining recommendation data corresponding to the consumption preference probability and sending the recommendation data to the consumption object.

[0123] In an example embodiment, the computer readable storage medium described above can include, but is not limited to, a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store computer programs.

[0124] The specific examples in the embodiment can refer to the examples described in the above embodiments and example implementations, and the embodiment will not be described here.

[0125] The embodiment of the application also provides an electronic device including a memory and a processor, the memory stores a computer program, and the processor is configured to execute the computer program to perform the steps in any of the above method embodiments.

[0126] Optionally, in the embodiment, the processor can be configured to execute the following steps through the computer program:

[0127] S1, determining consumed data of a consumption object from collected consumption data;

[0128] S2, determining a multi-dimensional prediction model corresponding to the object type of the consumption object;

[0129] S3, inputting the consumed data into the multi-dimensional prediction model to determine a consumption preference probability of the consumption object; wherein training data of the multi-dimensional prediction model at least includes historical consumed data of the consumption object and a consumption preference probability corresponding to the historical consumed data;

[0130] S4, obtaining recommendation data corresponding to the consumption preference probability and sending the recommendation data to the consumption object.

[0131] In one example embodiment, the electronic device described above can further include a transmission device connected to the processor and an input / output device connected to the processor.

[0132] The specific examples in the embodiments can refer to the examples described in the above embodiments and exemplary implementation, which will not be repeated here.

[0133] Obviously, those skilled in the art should understand that the modules or steps of the present application described above can be realized by general computing devices, which can be concentrated on a single computing device or distributed on a network composed of multiple computing devices, which can be realized by program codes executable by the computing devices, so that they can be stored in storage devices and executed by the computing devices, and in some cases, the steps shown or described can be executed in different order, or they can be made into individual integrated circuit modules, or multiple modules or steps can be made into a single integrated circuit module. Thus, the present application is not limited to any specific combination of hardware and software.

[0134] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Those skilled in the art can make various modifications and changes to the present application. Any modification, equivalent replacement, improvement, etc. within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for sending recommendation data, characterized in that, include: From the collected consumption data, determine the consumption data of the consumer; Determine a multi-dimensional prediction model corresponding to the object type of the consumer object; The consumed data is input into a multi-dimensional prediction model to determine the consumption preference probability of the consumer; wherein, the training data of the multi-dimensional prediction model includes at least: the historical consumed data of the consumer and the consumption preference probability corresponding to the historical consumed data; Obtain recommendation data corresponding to the consumption preference probability, and send the recommendation data to the consumer. The consumed data of the consumer is determined from the collected consumption data, including: The consumption data includes consumption behavior data and public behavior data, wherein the public behavior data represents the behavior data corresponding to the public behavior of the consumer in public places, and both the consumption behavior data and the public behavior data correspond to the object number of the consumer. Identify target consumer behavior data and target public behavior data, all corresponding to the target object number; If it is determined that the target object number belongs to the preset object information table, the target consumption behavior data and the target public behavior data are determined as the consumed data of the consumer object; The consumed data includes a first consumption behavior record, a second consumption behavior record, and a redemption activity record; wherein, the first consumption behavior record includes historical consumption behavior and current consumption behavior, the second consumption behavior record only includes the current consumption behavior, and the redemption activity record is a record of the consumer's participation in redemption activities within a historical time period; The object type of the consumer object with the first consumption behavior record and / or the redemption activity record is determined as a registered object, wherein the registered object includes at least one of the following: a first registered object, a second registered object, and a third registered object, wherein the first registered object's first consumption frequency is greater than or equal to a first preset value; the second registered object's second consumption frequency is less than the first preset value, and the second consumption frequency is greater than or equal to a second preset value; and the third registered object's third consumption frequency is less than the second preset value. The object type of the consumer object with the second consumption behavior record is determined as an unregistered object, wherein the registration probability of the unregistered object after a preset time period is greater than a third preset value; Determine a multi-dimensional prediction model corresponding to the object type of the consumer object, including: Obtain the similarity between the unregistered object and the registered object; If the similarity is determined to be greater than a preset threshold, a first multi-dimensional prediction model corresponding to the registered object is obtained, and the first multi-dimensional prediction model is determined as the second multi-dimensional prediction model corresponding to the unregistered object. And / or, if the similarity is determined to be greater than a preset threshold, a target registered object is determined from multiple registered objects, a third multi-dimensional prediction model corresponding to the target registered object is obtained, and the third multi-dimensional prediction model is determined as the fourth multi-dimensional prediction model corresponding to the unregistered object; Wherein, the similarity between the target registered object and the unregistered object is greater than the similarity between the target registered object and other unregistered objects, and the other unregistered objects are the registered objects other than the target registered object among the plurality of registered objects.

2. The method for sending recommendation data according to claim 1, characterized in that, Before determining the multi-dimensional prediction model corresponding to the object type of the consumer object, the method further includes: The historical consumed data is classified according to preset consumption types to obtain multi-dimensional data; The multi-dimensional data is normalized to obtain the processing result; Feature extraction is performed on the processing result to obtain multiple feature vectors, wherein the multiple feature vectors correspond to different dimensions; A correlation analysis is performed on the plurality of feature vectors and the consumption preference probability to obtain the degree of correlation between the plurality of feature vectors and the consumption preference probability; The component analysis algorithm is used to determine the target feature vector that has a correlation greater than a second preset threshold with the probability of the consumption preference. The multi-dimensional prediction model is trained using the target feature vector as the input sample and the consumption preference probability as the output sample.

3. The method for sending recommendation data according to claim 2, characterized in that, The historical consumed data is categorized according to preset consumption types to obtain multi-dimensional data, including: The historical consumed data is classified using multiple preset consumption types to obtain multiple first-level dimension data, wherein each preset consumption type has a corresponding consumption weight value, and each first-level dimension data corresponds to a preset consumption type. Each first-level dimension data is labeled using a first-level label corresponding to the preset consumption type, wherein there is a corresponding relationship between the labeled first-level dimension data and the first-level label; The first-level dimension data after being tagged is bound to the consumption weight value corresponding to the first-level dimension data after being tagged to obtain the second-level dimension data; The multi-dimensional data is determined based on multiple second-level dimension data.

4. The method for sending recommendation data according to any one of claims 1 to 3, characterized in that, Before determining the multi-dimensional prediction model corresponding to the object type of the consumer object, the method further includes: Historical consumption frequency is obtained from the historical consumption data, and the consumption preference probability is determined based on the magnitude of the historical consumption frequency. Alternatively, historical redemption records can be obtained from the historical consumption data, and the consumption preference probability can be determined based on the number of redemptions in the historical redemption records; Alternatively, the historical consumption frequency and the historical redemption records can be obtained from the historical consumption data, and the consumption preference probability can be determined based on the weighted sum of the number of redemptions in the historical consumption frequency and the historical redemption records.

5. The method for sending recommendation data according to claim 1, characterized in that, Obtaining recommendation data corresponding to the consumption preference probabilities includes: The consumption preference probabilities are obtained by inputting the consumed data into multiple multi-dimensional prediction models; wherein the multiple multi-dimensional prediction models are trained using different training algorithms and have different model structures. The target consumption preference probability is determined from the plurality of consumption preference probabilities, wherein the target consumption preference probability is the maximum value; Obtain the recommendation data corresponding to the probability of the target consumption preference.

6. A device for transmitting recommendation data, characterized in that, Used to implement the method described in any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein the program, when executed, performs the method according to any one of claims 1 to 5.

8. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the method described in any one of claims 1 to 5 through the computer program.

Images