Method and device for pushing bill discount business information

Abstract

The present invention provides a method and device for pushing bill discounting business information, particularly relating to the field of artificial intelligence. The method comprises: classifying the vector elements of a historical customer feature vector according to their corresponding element values ​​to obtain a classification value corresponding to each vector element of the historical customer feature vector; obtaining weights of the element types of the vector elements based on a decision tree; determining multiple target clusters based on the weights and the historical customer feature vectors; determining whether the current customer corresponding to the current customer feature vector has a willingness to discount based on the decision tree; and if so, determining the customer type corresponding to the current customer feature vector based on the target clusters; and pushing bill discounting business information to the current customer based on the customer type. The present invention can improve the speed and accuracy of bill discounting business information push, thereby improving the efficiency of bill discounting business information push, and thus helping to increase bank revenue.

Description

Technical Field

[0001] The present invention relates to the field of information push technology, in particular to the field of artificial intelligence, and more particularly to a method and device for pushing bill discount business information. Background Art

[0002] Bill discounting is a mainstream banking business, contributing significantly to bank revenue. Therefore, identifying customers who participate in bill discounting is crucial for increasing bank revenue. This identification primarily involves pushing bill discounting information to relevant users, enabling them to participate based on this information. However, existing methods for pushing bill discounting information typically rely on account managers, based on their experience, selecting users who may be interested in discounting. This push process requires account managers to spend considerable time compiling customer profiles and conducting analysis to determine the appropriate push strategy, resulting in a slow push process. Furthermore, since user selection is often based on manual experience, the selected users may not be interested in discounting, resulting in low accuracy in the push process. In summary, existing technologies suffer from slow and inaccurate push speeds for bill discounting information, resulting in inefficient push efficiency and hindering bank revenue growth. Summary of the Invention

[0003] One object of the present invention is to provide a method for pushing bill discounting business information to address the problem in the prior art of slow and inaccurate bill discounting business information push, which results in low efficiency and hinders bank revenue growth. Another object of the present invention is to provide a device for pushing bill discounting business information. Another object of the present invention is to provide a computer device. Yet another object of the present invention is to provide a readable medium.

[0004] In order to achieve the above objectives, one aspect of the present invention discloses a method for pushing bill discount business information, the method comprising:

[0005] Classify the vector elements of the preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the historical customer feature vector; classify the vector elements of the current customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the current customer feature vector;

[0006] Obtaining weights of element types of vector elements based on a preset decision tree; wherein the decision tree is associated with a plurality of the historical customer feature vectors, the classification values ​​of the historical customer feature vectors, and the corresponding historical discount labels;

[0007] Based on the weights and historical customer feature vectors, multiple target clusters are determined; based on the decision tree, it is determined whether the current customer corresponding to the current customer feature vector has a willingness to discount; if so, the customer type corresponding to the current customer feature vector is determined based on the target cluster; and based on the customer type, bill discount business information is pushed to the current customer.

[0008] Optionally, further including:

[0009] Before classifying the vector elements of the preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element,

[0010] Based on multiple preset customer authentication codes, corresponding initial historical customer information is obtained respectively;

[0011] Performing data cleaning, data extraction, and data standardization on the initial historical customer information to obtain intermediate historical customer information;

[0012] The intermediate historical customer information is subjected to feature vector quantization processing to obtain the historical customer feature vector.

[0013] Optionally, further including:

[0014] Before classifying the vector elements of the current customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the current customer feature vector,

[0015] The preset current customer information is subjected to feature vectorization processing to obtain the current customer feature vector.

[0016] Optionally, the classifying the vector elements of the preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the historical customer feature vector includes:

[0017] Based on the element value, the element type of the corresponding vector element, and a plurality of different element value intervals corresponding to the preset element type, determining, from the element value interval, a belonging interval corresponding to the element value of each vector element of the historical customer feature vector;

[0018] Based on the attribution interval, a classification value of a corresponding vector element in the historical customer feature vector is obtained.

[0019] Optionally, further including:

[0020] Before obtaining the weight of the element type of the vector element based on the preset decision tree,

[0021] A decision tree is constructed based on the plurality of historical customer feature vectors, the classification values ​​of the historical customer feature vectors, and the corresponding historical discount labels.

[0022] Optionally, constructing a decision tree based on the plurality of historical customer feature vectors, the classification values ​​of the historical customer feature vectors, and the corresponding historical discount labels includes:

[0023] Obtaining complete information entropy based on all the historical customer feature vectors and corresponding historical discount labels; wherein the value of the historical discount label is a label with discount experience or a label without discount experience;

[0024] Obtaining the root conditional entropy of each element type according to the historical discount labels corresponding to all the historical customer feature vectors and the classification values ​​of the vector elements;

[0025] According to the complete information entropy and the root conditional entropy, the root information gain entropy corresponding to the element type is obtained, and the element type with the largest root information gain entropy is used as the root node attribute to establish the root node of the decision tree; based on each classification value that the root node attribute can take, a child node corresponding to each classification value is established respectively;

[0026] Repeat the step of establishing child nodes until no child nodes can be established to complete the construction of the decision tree, wherein the step of establishing child nodes includes:

[0027] According to the sub-classification value corresponding to each of the sub-nodes, determining a plurality of historical customer feature vectors having vector elements corresponding to the sub-classification value as sub-historical customer feature vectors of the sub-node;

[0028] Determining whether the historical discount labels corresponding to the sub-historical customer feature vectors of each of the sub-nodes are all the same, and if so, treating the sub-nodes as leaf nodes; and determining, based on the leaf classification value corresponding to each leaf node, a plurality of historical customer feature vectors having vector elements corresponding to the leaf classification value as leaf historical customer feature vectors of the leaf node;

[0029] If not, obtaining the sub-conditional entropy of each element type in the sub-node according to the historical discount labels corresponding to all the sub-historical customer feature vectors and the classification values ​​of the vector elements;

[0030] According to the complete information entropy and the sub-conditional entropy, the sub-information gain entropy corresponding to the element type is obtained, and the element type with the largest sub-information gain entropy is used as the child node attribute; and the child node of the next layer of the child node is established based on each classification value that the child node attribute can take.

[0031] Optionally, obtaining complete information entropy based on all the historical customer feature vectors and corresponding historical discount labels includes:

[0032] Obtaining, based on all the historical customer feature vectors and the corresponding historical discount labels, a first number of discounted historical customer feature vectors whose historical discount labels are values ​​of discount experience labels, and a second number of undiscounted historical customer feature vectors whose historical discount labels are values ​​of no discount experience labels;

[0033] Obtaining a discount rate based on the first quantity and the total quantity of all historical customer feature vectors;

[0034] obtaining an undiscounted rate based on the second quantity and the total quantity of all historical customer feature vectors;

[0035] Based on the discounted rate and the undiscounted rate, a complete information entropy is obtained.

[0036] Optionally, obtaining the root conditional entropy of each element type according to the historical discount labels and classification values ​​of the vector elements corresponding to all the historical customer feature vectors includes:

[0037] According to the classification value of the vector element, respectively obtain the number of divisions of the vector elements with different classification values ​​in the element type;

[0038] Obtaining a division rate according to the number of divisions and the total number of all historical customer feature vectors;

[0039] Using historical customer feature vectors corresponding to different classification values ​​in the element type as corresponding partitioning vectors, and obtaining partitioning information entropies corresponding to different classification values ​​based on the partitioning vectors corresponding to different classification values ​​in the element type and historical discount labels corresponding to the partitioning vectors;

[0040] Based on the partitioning rates and partitioning information entropies corresponding to different classification values ​​that the element type can take, the root conditional entropy of the element type is obtained.

[0041] Optionally, obtaining the sub-conditional entropy of each element type in the child node according to the historical discount labels corresponding to all the sub-historical customer feature vectors and the classification values ​​of the vector elements includes:

[0042] According to the classification values ​​of the vector elements of the sub-historical customer feature vector, respectively obtaining the number of sub-divisions of the vector elements with different classification values ​​in the element type of the sub-historical customer feature vector;

[0043] Obtaining a sub-division rate according to the sub-division quantity and the sub-vector quantity of the sub-historical customer feature vector;

[0044] The sub-historical customer feature vectors corresponding to different classification values ​​in the element type are respectively used as corresponding sub-partition vectors, and based on the sub-partition vectors corresponding to different classification values ​​in the element type and the historical discount labels corresponding to the sub-partition vectors, the sub-partition information entropy corresponding to the different classification values ​​is obtained;

[0045] Based on the sub-division rates and sub-division information entropies corresponding to different classification values ​​that the element type can take, the sub-conditional entropy of the element type is obtained.

[0046] Optionally, obtaining the weight of the element type of the vector element based on a preset decision tree includes:

[0047] Based on the element type, determining the total number of nodes in the decision tree corresponding to the child node attributes and the root node attributes that are the same as the element type;

[0048] The total number of nodes is used as the weight of the element type.

[0049] Optionally, obtaining the weight of the element type of the vector element based on a preset decision tree includes:

[0050] Based on the element type, determining the depth of a child node corresponding to a child node attribute of the same type as the element in the decision tree;

[0051] Determine whether the root node attribute in the decision tree is the same as the element type; if not, obtain the weight of the element type according to the depth;

[0052] If so, a sub-weight of the element type is obtained according to the depth, and the sub-weight is added to a preset root weight to obtain the weight of the element type.

[0053] Optionally, determining multiple target clusters based on the weights and historical customer feature vectors includes:

[0054] Determine whether the historical discount label of the historical customer feature vector is a discount experience label, and if so, use the historical customer feature vector as a candidate vector;

[0055] Based on the weights, determining a plurality of element types with the largest weights and a preset number of reserved elements as standard types;

[0056] A plurality of cluster center vectors are selected from the candidate vectors, and a plurality of target clusters are determined based on vector elements of standard types of the candidate vectors and the cluster center vectors.

[0057] Optionally, determining a plurality of target clusters according to vector elements of a standard type of the candidate vector and the cluster center vector includes:

[0058] Using vector elements of standard types of the candidate vectors and the cluster center vector as standard elements, and using other vectors in the candidate vectors except the cluster center vector as vectors to be classified;

[0059] Determining a first Euclidean distance between each of the vectors to be classified and the cluster center vector based on the standard element;

[0060] Determining, based on the first Euclidean distance, a nearest cluster center vector that is closest to the vector to be classified;

[0061] Clustering the multiple vectors to be classified that have the same nearest cluster center vector to obtain multiple initial target clusters;

[0062] Repeat the clustering iteration step until there is a vector to be classified other than the cluster center vector whose corresponding second Euclidean distance is less than a preset distance threshold, wherein the clustering iteration step includes: based on multiple initial target clusters, obtaining a corresponding intermediate cluster center vector, and using other vectors in the candidate vectors except the intermediate cluster center vector as the vector to be classified; based on the standard element, determining the second Euclidean distance between each of the vector to be classified and the intermediate cluster center vector; based on the second Euclidean distance, determining the nearest intermediate cluster center vector closest to the vector to be classified; clustering multiple vectors to be classified having the same nearest intermediate cluster center vector to obtain an intermediate target cluster; and using the intermediate target cluster as the initial target cluster;

[0063] A plurality of the intermediate target clusters are used as the target clusters.

[0064] Optionally, the clustering based on multiple initial targets to obtain corresponding intermediate cluster center vectors includes:

[0065] Obtaining, according to all the eigenvectors included in the initial target cluster, a mean vector of all the eigenvectors corresponding to the initial target cluster;

[0066] The mean vector is used as the intermediate cluster center vector.

[0067] Optionally, judging whether the current customer corresponding to the current customer feature vector has a discount willingness according to the decision tree, and if so, determining the customer type corresponding to the current customer feature vector based on the clustering includes:

[0068] Determining a corresponding path in a decision tree according to the classification value of the vector element of the current customer feature vector;

[0069] According to the path, a corresponding leaf node is obtained;

[0070] It is determined whether the historical discount label corresponding to the leaf node is a label with discount experience. If so, the customer type corresponding to the current customer feature vector is determined based on the target cluster.

[0071] Optionally, determining the customer type corresponding to the current customer feature vector based on the target cluster includes:

[0072] Based on the target clusters, obtaining all final history vectors included in each of the target clusters;

[0073] Based on the final history vector, a corresponding individual discount degree value is obtained, and an average value of the individual discount degree values ​​is used as the overall discount degree value of the target cluster;

[0074] sorting the target clusters according to the overall discount degree values, so as to determine the historical discount type corresponding to the target clusters based on the sorting order;

[0075] Taking the average value of all final historical vectors of each target cluster as the average vector corresponding to the target cluster, and determining the final Euclidean distance between each average vector and the current customer feature vector;

[0076] The historical discount type of the target cluster corresponding to the average vector with the shortest final Euclidean distance is used as the customer type corresponding to the current customer feature vector.

[0077] In order to achieve the above objectives, another aspect of the present invention discloses a device for pushing information on bill discounting services, the device comprising:

[0078] a classification value determination module, configured to classify the vector elements of a preset historical customer feature vector according to the corresponding element values, and obtain the classification value corresponding to each vector element in the historical customer feature vector; and classify the vector elements of a current customer feature vector according to the corresponding element values, and obtain the classification value corresponding to each vector element in the current customer feature vector;

[0079] a decision tree processing module, configured to obtain weights of element types of vector elements based on a preset decision tree; wherein the decision tree is associated with a plurality of historical customer feature vectors, classification values ​​of the historical customer feature vectors, and corresponding historical discount labels;

[0080] A push module is used to determine multiple target clusters based on the weights and historical customer feature vectors; determine whether the current customer corresponding to the current customer feature vector has a willingness to discount according to the decision tree; if so, determine the customer type corresponding to the current customer feature vector based on the target cluster; and push bill discount business information to the current customer according to the customer type.

[0081] The present invention also discloses a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the method described above is implemented when the processor executes the program.

[0082] The present invention also discloses a computer-readable medium on which a computer program is stored. When the program is executed by a processor, the method described above is implemented.

[0083] The method and device for pushing information on bill discounting business provided by the present invention classify the vector elements of a preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the historical customer feature vector; classify the vector elements of a current customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the current customer feature vector, and can map some vector elements with an infinite value range in the historical customer feature vector and the current customer feature vector to classification values ​​with a finite value range, and the classification values ​​can take only a finite number of values, so that subsequent processing related to the decision tree, determination of clustering and corresponding customer types can be implemented based on the classification values ​​in a finite range, reducing the amount of data that needs to be traversed and operated in the process of data processing and classification, thereby significantly improving the speed of pushing information on the overall bill discounting business; obtaining the weight of the element type of the vector element based on the preset decision tree; wherein, the decision tree is related to a plurality of the historical customer feature vectors, ... The classification value of the customer feature vector is associated with the corresponding historical discount label, which can quickly and accurately determine the decisive role and influence of different vector element types in the feature vector on the feature based on the decision tree, thereby facilitating that in subsequent steps, only the vector elements corresponding to the element types with a greater degree of influence on the feature are retained as the basis for determining the customer type, further reducing the amount of data that needs to be traversed and operated, thereby indirectly improving the speed and accuracy of determining the customer type in subsequent steps, thereby indirectly improving the speed and accuracy of pushing information on the overall bill discount business; by determining multiple target clusters based on the weights and historical customer feature vectors, the vector elements corresponding to the element types with a greater degree of influence on the feature in the historical customer feature vector can be determined as the basis for clustering based on the weights of the element types, and clusters with different overall characteristics can be quickly and accurately formed; according to the decision tree, it is judged whether the current customer corresponding to the current customer feature vector has a willingness to discount, and if so, the customer type corresponding to the current customer feature vector is determined based on the target cluster;Pushing bill discounting business information to the current customer based on the customer type can first use a decision tree for preliminary screening, filtering out feature vectors of current customers without a willingness to discount, and retaining only feature vectors of current customers with a willingness to discount as objects for determining the customer type. This effectively reduces the amount of customer information that needs to be processed and analyzed during subsequent pushes, improving the overall push speed. Furthermore, based on the target cluster, the customer type corresponding to the current customer feature vector is determined. Based on the type and overall characteristics corresponding to the target cluster, the characteristics of the current customer feature vector are analyzed to determine which target cluster's overall characteristics are most similar, thereby selecting the type of the closest target cluster as the customer type corresponding to the current customer feature vector. Since the target clusters are generated based on actual historical customer feature vectors, the above steps can not only quickly determine the current customer's customer type in a simple manner, but also significantly improve the accuracy of the determined customer type, thereby effectively improving the speed and accuracy of pushing bill discounting business information to the corresponding current customer based on the customer type. In summary, the bill discounting business information push method and device provided by the present invention can improve the speed and accuracy of bill discounting business information push, thereby improving the efficiency of bill discounting business information push, and thus facilitating increased bank revenue. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0085] Figure 1 A schematic diagram showing a flow chart of a method for pushing bill discount business information according to an embodiment of the present invention;

[0086] Figure 2 A schematic diagram showing an optional step of obtaining a classification value corresponding to each vector element in a historical customer feature vector according to an embodiment of the present invention is shown;

[0087] Figure 3 A schematic diagram showing an exemplary decision tree according to an embodiment of the present invention is shown;

[0088] Figure 4 A schematic diagram showing an optional step of determining multiple target clusters according to an embodiment of the present invention is shown;

[0089] Figure 5 A schematic diagram showing an optional step of determining the customer type corresponding to the current customer feature vector according to an embodiment of the present invention is shown;

[0090] Figure 6 A schematic diagram of a module of a bill discount business information push device according to an embodiment of the present invention is shown;

[0091] Figure 7 A schematic diagram showing the structure of a computer device suitable for implementing an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0092] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0093] The terms “first,” “second,” etc. used herein do not particularly refer to an order or sequence, nor are they used to limit the present invention. They are only used to distinguish elements or operations described with the same technical terms.

[0094] The words “include,” “including,” “have,” “contain,” etc. used in this document are open-ended terms, meaning including but not limited to.

[0095] As used herein, "and / or" includes any and all combinations of the items mentioned.

[0096] It should be noted that the acquisition, storage, use, and processing of data in the technical solution of the present invention are in compliance with the relevant provisions of national laws and regulations.

[0097] The embodiment of the present invention discloses a method for pushing bill discount business information, such as Figure 1 As shown, the method specifically includes the following steps:

[0098] S101: Classify the vector elements of the preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the historical customer feature vector; classify the vector elements of the current customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the current customer feature vector.

[0099] S102: Obtaining weights of element types of vector elements based on a preset decision tree; wherein the decision tree is associated with a plurality of historical customer feature vectors, classification values ​​of the historical customer feature vectors, and corresponding historical discount labels.

[0100] S103: Based on the weights and historical customer feature vectors, multiple target clusters are determined; based on the decision tree, whether the current customer corresponding to the current customer feature vector has a willingness to discount is determined; if so, the customer type corresponding to the current customer feature vector is determined based on the target cluster; and bill discount business information is pushed to the current customer based on the customer type.

[0101] For example, the specific implementation method for pushing bill discounting service information to the current customer based on the customer type can be determined by those skilled in the art based on actual circumstances, and the embodiments of the present invention do not impose any limitation thereto. For example, for a current customer with a "high discount probability" customer type, bill discounting service information is pushed to the customer via face-to-face communication. For a current customer with a "high discount probability" customer type, bill discounting service information is pushed to the customer via telephone communication. For a current customer with a "medium discount probability" customer type, bill discounting service information is pushed to the customer via text message. For a current customer with a "low discount probability" customer type, bill discounting service information is pushed to the customer via a pop-up advertisement in the customer's relevant application. For a current customer with a "very low discount probability" customer type, bill discounting service information is not pushed to the customer. The specific method for pushing bill discounting service information to the current customer based on the customer type may include, but is not limited to, determining the current customer's bill discounting intention strength based on the customer type, and then pushing bill discounting service information to the current customer based on the bill discounting intention strength.

[0102] Exemplarily, the element type of the vector element corresponds to the relevant attributes of the customer, including but not limited to gender (if the customer is an individual, it is the individual's gender; if it is a company, it is the chairman's gender or CEO's gender, etc.), bill holding bank, bill holding time, bill maturity date, bill holding date, issuer information, acceptor information, bill number, customer social unified credit code, customer number, bill holding behavior category, total number of bills held, historical discount probability, historical endorsement probability, loan amount, trade financing amount, historical discount amount, upstream and downstream enterprise capital transaction amount, enterprise assets and operating outflow amount, etc. Among them, the element type of the vector element of the historical customer feature vector needs to be consistent with the element type of the vector element of the current customer feature vector. It should be noted that the element type of the vector element can be determined by those skilled in the art according to actual conditions. The above description is only an example and does not constitute a limitation.

[0103] The method and device for pushing information on bill discounting business provided by the present invention classify the vector elements of a preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the historical customer feature vector; classify the vector elements of a current customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the current customer feature vector, and can map some vector elements with an infinite value range in the historical customer feature vector and the current customer feature vector to classification values ​​with a finite value range, and the classification values ​​can take only a finite number of values, so that subsequent processing related to the decision tree, determination of clustering and corresponding customer types can be implemented based on the classification values ​​in a finite range, reducing the amount of data that needs to be traversed and operated in the process of data processing and classification, thereby significantly improving the speed of pushing information on the overall bill discounting business; obtaining the weight of the element type of the vector element based on the preset decision tree; wherein, the decision tree is related to a plurality of the historical customer feature vectors, ... The classification value of the customer feature vector is associated with the corresponding historical discount label, which can quickly and accurately determine the decisive role and influence of different vector element types in the feature vector on the feature based on the decision tree, thereby facilitating that in subsequent steps, only the vector elements corresponding to the element types with a greater degree of influence on the feature are retained as the basis for determining the customer type, further reducing the amount of data that needs to be traversed and operated, thereby indirectly improving the speed and accuracy of determining the customer type in subsequent steps, thereby indirectly improving the speed and accuracy of pushing information on the overall bill discount business; by determining multiple target clusters based on the weights and historical customer feature vectors, the vector elements corresponding to the element types with a greater degree of influence on the feature in the historical customer feature vector can be determined as the basis for clustering based on the weights of the element types, and clusters with different overall characteristics can be quickly and accurately formed; according to the decision tree, it is judged whether the current customer corresponding to the current customer feature vector has a willingness to discount, and if so, the customer type corresponding to the current customer feature vector is determined based on the target cluster;Pushing bill discounting business information to the current customer based on the customer type can first use a decision tree for preliminary screening, filtering out feature vectors of current customers without a willingness to discount, and retaining only feature vectors of current customers with a willingness to discount as objects for determining the customer type. This effectively reduces the amount of customer information that needs to be processed and analyzed during subsequent pushes, improving the overall push speed. Furthermore, based on the target cluster, the customer type corresponding to the current customer feature vector is determined. Based on the type and overall characteristics corresponding to the target cluster, the characteristics of the current customer feature vector are analyzed to determine which target cluster's overall characteristics are most similar, thereby selecting the type of the closest target cluster as the customer type corresponding to the current customer feature vector. Since the target clusters are generated based on actual historical customer feature vectors, the above steps can not only quickly determine the current customer's customer type in a simple manner, but also significantly improve the accuracy of the determined customer type, thereby effectively improving the speed and accuracy of pushing bill discounting business information to the corresponding current customer based on the customer type. In summary, the bill discounting business information push method and device provided by the present invention can improve the speed and accuracy of bill discounting business information push, thereby improving the efficiency of bill discounting business information push, and thus facilitating increased bank revenue.

[0104] In an optional embodiment, further comprising:

[0105] Before classifying the vector elements of the preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element,

[0106] Based on multiple preset customer authentication codes, corresponding initial historical customer information is obtained respectively;

[0107] Performing data cleaning, data extraction, and data standardization on the initial historical customer information to obtain intermediate historical customer information;

[0108] The intermediate historical customer information is subjected to feature vector quantization processing to obtain the historical customer feature vector.

[0109] Exemplarily, the customer authentication code may be, but is not limited to, the customer's unified social credit code or organizational structure code.

[0110] Exemplarily, the data cleaning of the initial historical customer information may include, but is not limited to, replacing or deleting abnormal data in the historical customer information using cleaning methods such as spline interpolation and linear regression. The data extraction may include, but is not limited to, performing a dimensionality reduction operation on highly correlated variables. For example, if the initial historical customer information contains two variables, the total amount of historical discounts and the total amount of historical discounts using silver cards, since the total amount of historical discounts using silver cards is not of concern in this embodiment of the present invention, and the total amount of historical discounts and the total amount of historical discounts using silver cards are essentially the same and highly correlated, the variable of the total amount of historical discounts using silver cards is deleted (so that subsequent related attributes and element types do not include the total amount of historical discounts using silver cards) to complete the dimensionality reduction operation. The data standardization processing may include, but is not limited to, converting the relevant data into various appropriate formats. For example, for the relevant amount, its format is converted to a format with a granularity of two decimal places, such as converting the amount of 10,000 yuan into the amount of 10,000.00 yuan. It should be noted that the specific implementation method for performing data cleaning, data extraction and data standardization on the initial historical customer information to obtain intermediate historical customer information can be determined by those skilled in the art based on actual conditions. The above description is only an example and does not constitute a limitation to this.

[0111] Exemplarily, the feature vectorization processing of the intermediate historical customer information to obtain the historical customer feature vector may be, but is not limited to, extracting corresponding features from each attribute information in the intermediate historical customer information to obtain vector elements of the corresponding attribute information, and then concatenating and integrating the respective vector elements to obtain the corresponding historical customer feature vector. Specifically, for attribute information whose attribute values ​​are numbers, when determining the vector elements, the numerical attribute values ​​may be directly determined as the corresponding vector elements, or the attribute values ​​may be normalized to obtain the corresponding vector elements; and for attribute information whose attribute values ​​are not numbers (e.g., text or characters), when determining the vector elements, the attribute information may be digitized to obtain the corresponding vectorized elements, or the digitized products may be normalized to obtain the corresponding vectorized elements. The digitization processing may be implemented by, but is not limited to, using a digital encoding (e.g., one-hot encoding) algorithm or processing against an ASCII code table. For attribute information whose attribute value is a category (e.g., ticket holding behavior category, gender, etc.), all possible categories of the attribute information can be encoded to obtain numbers corresponding to the category, and the attribute value in the attribute information can be replaced with the corresponding number. For example, if the attribute information is gender information and the possible categories are male and female, the number 0 can be used to represent male and the number 1 can be used to represent female. It should be noted that the specific implementation method for performing feature vector quantization processing on the intermediate historical customer information to obtain the historical customer feature vector can be determined by those skilled in the art based on actual circumstances. The above description is only an example and does not constitute a limitation.

[0112] For example, the acquisition and processing of historical customer information may be implemented through, but not limited to, a corresponding big data platform, such as, but not limited to, a Hadoop big data platform.

[0113] Through the above steps, on the basis of error correction and simplification of historical customer information, historical customer information can be quickly and accurately converted into a vector form that is easy to participate in calculations and processing, thereby making the operations of related calculations and processing in subsequent steps more concise, and effectively improving the speed of pushing the overall bill discounting business information.

[0114] In an optional embodiment, further comprising:

[0115] Before classifying the vector elements of the current customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the current customer feature vector,

[0116] The preset current customer information is subjected to feature vectorization processing to obtain the current customer feature vector.

[0117] For example, for the specific implementation method of performing feature vectorization processing on the preset current customer information to obtain the current customer feature vector, reference can be made to the description of the step of performing feature vectorization processing on the intermediate historical customer information to obtain the historical customer feature vector in the embodiment of the present invention, which will not be repeated here.

[0118] Through the above steps, the current customer information can be quickly and accurately converted into a vector form that is easy to participate in calculations and processing, thereby making the operations in subsequent steps more concise and effectively improving the speed of pushing the overall bill discounting business information.

[0119] In a preferred embodiment, before the preset current customer information is subjected to feature vector quantization processing, the current customer information is first subjected to data cleaning, data extraction and data standardization processing to correct errors and simplify the current customer information, so as to improve the accuracy of the generated current customer feature vector and make it more convenient to participate in calculation and processing.

[0120] In an optional embodiment, if Figure 2 As shown, the process of classifying the vector elements of the preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the historical customer feature vector includes the following steps:

[0121] S201: Based on the element value, the element type of the corresponding vector element, and a plurality of preset element value intervals corresponding to the element type, determine a belonging interval corresponding to the element value of each vector element of the historical customer feature vector from the element value interval.

[0122] S202: Based on the attribution interval, obtain a classification value of a corresponding vector element in the historical customer feature vector.

[0123] Exemplarily, the multiple different element value intervals corresponding to the element type can be determined by those skilled in the art according to actual conditions, and the embodiments of the present invention do not limit this. For example, for the numerical element type of historical discount amount, the element value interval can be set to, but not limited to: [0, 100000), [100000, 500000), [500000, 1000000) and [1000000, +∞), etc. For example, for the categorical element type, the element value interval can be set to the element value itself. For example, for the categorical element type of gender, the two values ​​of "male" and "female" can be taken (which can correspond to 0 and 1 or other corresponding numbers after encoding), and the corresponding element value interval can be set to "male" and "female" (corresponding to the encoded value, it can also be 0 and 1 or other corresponding numbers).

[0124] Exemplarily, the determination of the belonging interval corresponding to the element value of each vector element of the historical customer feature vector from the element value interval can be, but is not limited to, first obtaining multiple element value intervals of the element type to which the vector element belongs, and then determining the element value interval in which the element value of the vector element lies as the corresponding belonging interval. For example, an element in a historical customer feature vector (assuming the corresponding element type is the historical discounted amount) has an element value of 300,000, and it is known that the multiple element value intervals corresponding to the historical discounted amount are: [0, 100,000), [100,000, 500,000), [500,000, 1000,000), and [1000,000, +∞). It can be seen that the interval corresponding to the element value 300,000 is [100,000, 500,000). Therefore, the belonging interval of the vector element of the historical customer feature vector is determined to be [100,000, 500,000). It should be noted that the specific implementation of step S201 can be determined by those skilled in the art according to actual conditions, and the above description is only an example and does not constitute a limitation thereto.

[0125] Exemplarily, the classification value of the corresponding vector element in the historical customer feature vector obtained based on the attribution interval can be determined by, but not limited to, querying the corresponding relationship between a preset element value interval and the classification value, or taking the average of the upper limit and the lower limit of the attribution interval or either of the two as the classification value. For example, for a vector element of a historical discount amount type in a historical customer feature vector, its attribution interval is [100000, 500000). Then, by querying the corresponding relationship between its element value interval and the classification value, it can be obtained that: the classification value corresponding to [0, 100000) is 0, the classification value corresponding to [100000, 500000) is 1, the classification value corresponding to [500000, 1000000) is 2, and the classification value corresponding to [1000000, +∞) is 3. Therefore, it can be seen that the classification value corresponding to this vector element in the historical customer feature vector is 1. For another example, for a vector element of a certain gender type in a historical customer feature vector, if its belonging interval is 0 (corresponding to male gender), then its belonging interval can be directly used as the classification value, that is, the classification value is also 0. The classification value is set mainly for element types such as historical discount amounts and total number of votes held, whose value range is an infinite set, so that the values ​​of the corresponding elements of the vector can be mapped to classification values ​​of a finite set during the relevant processing, and the classification values ​​are used instead of the original real values ​​for subsequent classification processing, so that the relevant processing processes related to decision trees and clustering will not cause processing troubles and long processing times due to the large number of element values ​​of the corresponding elements (for example, if there are 1,000 historical customer feature vectors, then there may be 1,000 different corresponding element values ​​for the element type of historical discount amounts, and if mapped to classification values ​​in the manner of the above example, there will be at most 4 different corresponding classification values).

[0126] Through the above steps, the specific basis for determining the classification value can be refined, so that there will be no contradictions between the classification values ​​corresponding to all vector elements in multiple historical customer feature vectors, and there will be no situation where the accuracy is low due to a large coverage range, thereby improving the accuracy of the processing steps of the decision tree based on the classification value in subsequent steps, and mapping the element value to a simple classification value to participate in the processing of the decision tree can make the decision tree have a faster decision-making speed, thereby improving the speed of initial screening of current customers with discount willingness based on the decision tree, and thereby improving the overall push speed.

[0127] In an optional embodiment, further comprising:

[0128] Before obtaining the weight of the element type of the vector element based on the preset decision tree,

[0129] A decision tree is constructed based on the plurality of historical customer feature vectors, the classification values ​​of the historical customer feature vectors, and the corresponding historical discount labels.

[0130] Exemplarily, the corresponding historical discount label may be, but is not limited to, a label indicating whether the historical customer corresponding to the historical customer feature vector has any discount experience.

[0131] Exemplarily, the classification value of the historical customer feature vector includes a classification value corresponding to each vector element of each historical customer feature vector.

[0132] Through the above steps, a decision tree can be constructed with multiple feature vector samples and corresponding decision category labels required for constructing the decision tree as input, so that the constructed decision tree is complete. When the decision tree is used for relevant processing, there will be no processing stagnation due to the lack of tree structure, which is more conducive to the smooth execution of the subsequent steps of initially screening the feature vectors of current customers with discount willingness.

[0133] In an optional embodiment, constructing a decision tree based on the plurality of historical customer feature vectors, the classification values ​​of the historical customer feature vectors, and the corresponding historical discount labels includes:

[0134] Obtaining complete information entropy based on all the historical customer feature vectors and corresponding historical discount labels; wherein the value of the historical discount label is a label with discount experience or a label without discount experience;

[0135] Obtaining the root conditional entropy of each element type according to the historical discount labels corresponding to all the historical customer feature vectors and the classification values ​​of the vector elements;

[0136] According to the complete information entropy and the root conditional entropy, the root information gain entropy corresponding to the element type is obtained, and the element type with the largest root information gain entropy is used as the root node attribute to establish the root node of the decision tree; based on each classification value that the root node attribute can take, a child node corresponding to each classification value is established respectively;

[0137] Repeat the step of establishing child nodes until no child nodes can be established to complete the construction of the decision tree, wherein the step of establishing child nodes includes:

[0138] According to the sub-classification value corresponding to each of the sub-nodes, determining a plurality of historical customer feature vectors having vector elements corresponding to the sub-classification value as sub-historical customer feature vectors of the sub-node;

[0139] Determining whether the historical discount labels corresponding to the sub-historical customer feature vectors of each of the sub-nodes are all the same, and if so, treating the sub-nodes as leaf nodes; and determining, based on the leaf classification value corresponding to each leaf node, a plurality of historical customer feature vectors having vector elements corresponding to the leaf classification value as leaf historical customer feature vectors of the leaf node;

[0140] If not, obtaining the sub-conditional entropy of each element type in the sub-node according to the historical discount labels corresponding to all the sub-historical customer feature vectors and the classification values ​​of the vector elements;

[0141] According to the complete information entropy and the sub-conditional entropy, the sub-information gain entropy corresponding to the element type is obtained, and the element type with the largest sub-information gain entropy is used as the child node attribute; and the child node of the next layer of the child node is established based on each classification value that the child node attribute can take.

[0142] Exemplarily, the complete information entropy obtained based on all the historical customer feature vectors and the corresponding historical discount labels can be expressed as the following formula:

[0143]

[0144] Where H(p) represents the complete information entropy, p i When i=1, it indicates the probability of discounted experience corresponding to the historical customer feature vector, and p i When i=2, it represents the probability of no discounting experience corresponding to the historical customer feature vector. The above-mentioned probabilities of discounting experience and no discounting experience are determined by the historical discount label corresponding to each historical customer feature vector. For example, there are 10,000 historical customer feature vectors, of which 300 historical customer feature vectors correspond to historical discount labels with discounting experience labels. Then it can be confirmed that the probability of discounting experience is 300 / 10,000=3%. Similarly, the probability of no discounting experience is 97%. Among them, the specific value of the historical discount label of each historical customer feature vector can be manually set by relevant staff according to relevant historical circumstances. It should be noted that the specific implementation method for obtaining complete information entropy based on all the historical customer feature vectors and the corresponding historical discount labels can be determined by those skilled in the art according to actual conditions. The above description is only an example and does not constitute a limitation.

[0145] Exemplarily, the root conditional entropy of each element type is obtained based on the historical discount labels corresponding to all the historical customer feature vectors and the classification values ​​of the vector elements, which can be expressed as the following formula:

[0146]

[0147] Among them, H(Y|X) represents the root conditional entropy corresponding to the element type, p iH(Y|X=X) represents the probability (division rate) of the element type taking the i-th classification value (for example, among 10,000 historical customer feature vectors, for the element type historical discount amount, there are 1,000 historical customer feature vectors with a classification value of 1, then the probability of the element type taking the classification value 1 is 1,000 / 10,000=10%, and the classification value 1 can be set as the i-th classification value). m represents the number of classification values ​​that the element type can take. For example, for the element type historical discount amount, the possible classification values ​​are 0, 1, 2, and 3, then the value of m is 4. i ) represents the partitioned information entropy corresponding to the element type X when taking a certain classification value Xi (the partitioned information entropy has the same properties as the complete information entropy, both are information entropy, but the range of the historical customer feature vectors based on which the information entropy is determined is different). Among them, one element type corresponds to one root conditional entropy. It should be noted that the specific implementation method for obtaining the root conditional entropy of each element type based on the historical discount labels and classification values ​​of the vector elements corresponding to all the historical customer feature vectors can be determined by those skilled in the art according to actual circumstances. The above description is only an example and does not constitute a limitation thereto.

[0148] Exemplarily, obtaining the root information gain entropy corresponding to the element type according to the complete information entropy and the root conditional entropy is performed by subtracting the root conditional entropy from the complete information entropy to obtain the root information gain entropy.

[0149] For example, establishing the root node of a decision tree based on the root node attributes is a conventional technical means in this field and will not be described in detail here. Figure 3 As shown in the figure, if the element type with the largest root information entropy is the total number of votes, the total number of votes is used as the root node attribute to establish the root node of the decision tree - node 1.

[0150] Exemplarily, based on each possible classification value of the root node attribute, a child node corresponding to each classification value is established, as follows:

[0151] like Figure 3As shown, node 1 is the root node, and its root node attribute is the total number of votes held. At this time, it is known that the element of the total number of votes held can take three classification values ​​of 10, 11 and 12 (corresponding to different element value intervals, for example, they can correspond to but are not limited to [0, 1500), [1500, 10000) and [10000, +∞)). Then, three paths are extended from the root node, and each path is provided with a child node at one end away from the root node. The classification value (sub-classification value) corresponding to each child node is different. For example, the sub-classification value corresponding to node 2 as a child node is 10, the sub-classification value corresponding to node 3 is 11, and the sub-classification value corresponding to node 4 is 12. It should be noted that the specific implementation method of establishing the child node corresponding to each classification value based on each classification value that can be taken by the root node attribute can be determined by those skilled in the art according to actual conditions. The above description is only an example and does not constitute a limitation to this.

[0152] Exemplarily, according to the sub-classification value corresponding to each of the child nodes, determining multiple historical customer feature vectors having vector elements corresponding to the sub-classification value as the sub-historical customer feature vectors of the child node, there are the following examples:

[0153] For node 2, its corresponding sub-classification value is 10, and among all historical customer feature vectors, the historical customer feature vectors whose classification value of the vector element corresponding to the element type "total number of votes held" is 10 (that is, the historical customer feature vectors whose classification value of the element representing "total number of votes held" in the vector is 10) are A, B and C, then A, B and C are determined as the sub-historical customer feature vectors of node 2.

[0154] For node 3, its corresponding sub-classification value is 11. Among all historical customer feature vectors, the historical customer feature vectors whose classification value is 11 for the vector element corresponding to the element type "total number of votes held" are Ding, Wu and Ji. Therefore, Ding, Wu and Ji are determined as the sub-historical customer feature vectors of node 3.

[0155] For node 4, its corresponding sub-classification value is 12. Among all historical customer feature vectors, the historical customer feature vectors with classification values ​​of 12 corresponding to the vector elements of element type "total number of votes held" are Geng, Xin, Ren and Gui. Geng, Xin, Ren and Gui are determined as the sub-historical customer feature vectors of node 4.

[0156] It should be noted that the specific implementation method of determining multiple historical customer feature vectors having vector elements corresponding to the sub-classification values ​​corresponding to each of the sub-nodes as the sub-historical customer feature vectors of the sub-node can be determined by those skilled in the art based on actual conditions. The above description is only an example and does not constitute a limitation to this.

[0157] Exemplarily, the process of determining whether the historical discount labels corresponding to the sub-historical customer feature vectors of each of the sub-nodes are the same, and if so, treating the sub-nodes as leaf nodes; and determining, based on the leaf classification value corresponding to each leaf node, multiple historical customer feature vectors having vector elements corresponding to the leaf classification value as leaf historical customer feature vectors of the leaf node, may include the following examples:

[0158] The sub-historical customer feature vectors of node 2 are A, B and C, among which the historical discount labels corresponding to A, B and C are all discount experience labels. It can be seen that the historical discount labels of the sub-historical customer feature vectors of node 2 are all the same. Then node 2 is taken as a leaf node, and its corresponding leaf classification value is 10. The corresponding leaf historical customer feature vectors are A, B and C.

[0159] It should be noted that, for respectively judging whether the historical discount labels corresponding to the sub-historical customer feature vectors of each of the sub-nodes are the same, if so, the sub-nodes are regarded as leaf nodes; according to the leaf classification value corresponding to each leaf node, determining that multiple historical customer feature vectors having vector elements corresponding to the leaf classification value are the leaf historical customer feature vectors of the leaf node. The specific implementation method can be determined by those skilled in the art according to actual conditions. The above description is only an example and does not constitute a limitation to this.

[0160] Exemplarily, the sub-conditional entropy for each element type in the child node is obtained based on the historical discount labels and classification values ​​of the vector elements corresponding to all the sub-historical customer feature vectors. This means that the conditional entropy is calculated for all the sub-historical customer feature vectors corresponding to the current node (for example, for node 3, all its sub-historical customer feature vectors are D, E, and F), rather than for all historical customer feature vectors. Therefore, the sub-conditional entropy is not equivalent to the root conditional entropy. However, for the specific principles of calculating the sub-conditional entropy, please refer to the description of calculating the root conditional entropy in the embodiments of the present invention and will not be repeated here.

[0161] Exemplarily, the sub-information gain entropy corresponding to the element type is obtained according to the complete information entropy and the sub-conditional entropy, and the element type with the largest sub-information gain entropy is used as the child node attribute; the specific implementation method of establishing the child node of the next layer of the child node based on each classification value that the child node attribute can take can refer to the embodiment of the present invention for obtaining the root information gain entropy corresponding to the element type according to the complete information entropy and the root conditional entropy, and establishing the root node of the decision tree with the element type with the largest root information gain entropy as the root node attribute; based on each classification value that the root node attribute can take, the description of the steps of establishing the child node corresponding to each classification value respectively, which will not be repeated here.

[0162] For example, Figure 3As shown in the example, if the child node attribute of node 3 is ticket holding time, and its corresponding possible classification values ​​are 20 and 21, then nodes 5 and 6 are created with the classification value 20 and 21, respectively. Similarly, if the child node attribute of node 4 is ticket number, and its corresponding possible classification values ​​are 30 and 31, then nodes 7 and 8 are created with the classification value 30 and 31, respectively. If the classification values ​​corresponding to the historical customer feature vectors contained in nodes 5, 6, 7, and 8 are consistent, then nodes 5, 6, 7, and 8 are determined to be leaf nodes.

[0163] It should be noted that the specific implementation of each step of constructing the decision tree can be determined by those skilled in the art according to actual conditions. The above description is only an example and does not constitute a limitation thereto.

[0164] Through the above steps, a decision tree can be constructed using the standard method for establishing an ID3 decision tree. Since the structure and parameters of ID3 decision trees are not complex, the processing and calculation speed is faster when using the ID3 decision tree to initially screen the feature vectors of current customers with discount willingness, thereby improving the overall push speed. Moreover, the steps required to construct an ID3 decision tree are also not complex, so the decision tree construction speed is also relatively fast, which can indirectly improve the overall push speed.

[0165] In an optional embodiment, obtaining the complete information entropy based on all the historical customer feature vectors and the corresponding historical discount labels includes:

[0166] Obtaining, based on all the historical customer feature vectors and the corresponding historical discount labels, a first number of discounted historical customer feature vectors whose historical discount labels are values ​​of discount experience labels, and a second number of undiscounted historical customer feature vectors whose historical discount labels are values ​​of non-discount experience labels;

[0167] Obtaining a discount rate based on the first quantity and the total quantity of all historical customer feature vectors;

[0168] obtaining an undiscounted rate based on the second quantity and the total quantity of all historical customer feature vectors;

[0169] Based on the discounted rate and the undiscounted rate, a complete information entropy is obtained.

[0170] Exemplarily, obtaining the discount rate based on the first quantity and the total quantity of all historical customer feature vectors is performed by dividing the first quantity by the total quantity to obtain the discount rate.

[0171] Exemplarily, the obtaining of the undiscounted rate based on the second number and the total number of all historical customer feature vectors is performed by dividing the second number by the total number to obtain the undiscounted rate.

[0172] Exemplarily, the complete information entropy is obtained based on the discount rate and the undiscounted rate, and reference may be made to the formula in the embodiment of the present invention:

[0173]

[0174] The description of p is not repeated here. i When i=1, it indicates the probability of discounted experience (discount rate) corresponding to the historical customer feature vector, and p i When i=2, it represents the undiscounted experience probability (undiscounted rate) corresponding to the historical customer feature vector.

[0175] Through the above steps, the complete information entropy can be determined using the standard information entropy method, thereby improving the accuracy of the obtained complete information entropy, and then improving the accuracy of the decision tree constructed in the subsequent steps and the related processing based on the decision tree, thereby improving the accuracy of the overall push.

[0176] In an optional embodiment, obtaining the root conditional entropy of each element type based on the historical discount labels corresponding to all the historical customer feature vectors and the classification values ​​of the vector elements includes:

[0177] According to the classification value of the vector element, respectively obtain the number of divisions of the vector elements with different classification values ​​in the element type;

[0178] Obtaining a division rate according to the number of divisions and the total number of all historical customer feature vectors;

[0179] Using historical customer feature vectors corresponding to different classification values ​​in the element type as corresponding partitioning vectors, and obtaining partitioning information entropies corresponding to different classification values ​​based on the partitioning vectors corresponding to different classification values ​​in the element type and historical discount labels corresponding to the partitioning vectors;

[0180] Based on the partitioning rates and partitioning information entropies corresponding to different classification values ​​that the element type can take, the root conditional entropy of the element type is obtained.

[0181] Exemplarily, the number of partitions of vector elements with different classification values ​​in the element type is the number of historical customer feature vectors with different classification values ​​for a certain element type. For example, if the element type is historical discounted amount, there are 1,000 historical customer feature vectors with a classification value of 1 for the historical discounted amount element, 3,000 historical customer feature vectors with a classification value of 0 for the historical discounted amount element, 5,000 historical customer feature vectors with a classification value of 2, and 1,000 historical customer feature vectors with a classification value of 3. It can be seen that for the element type of historical discounted amount, the number of partitions with a classification value of 1 is 1,000, the number of partitions with a classification value of 2 is 5,000, the number of partitions with a classification value of 0 is 3,000, and the number of partitions with a classification value of 3 is 1,000.

[0182] Exemplarily, the division rate is obtained based on the number of divisions and the total number of all historical customer feature vectors. Specifically, the division rate is obtained by dividing the number of divisions by the total number. For example, for the element type of historical discount amount, the number of divisions with a classification value of 1 is 1000, and the total number is 10000, then the division rate is 1000 / 10000=10%.

[0183] Exemplarily, based on the partitioning vectors corresponding to different classification values ​​in the element type and the historical discount labels corresponding to the partitioning vectors, the specific principle of obtaining the partitioning information entropy corresponding to different classification values ​​can refer to the description of the steps for calculating the complete information entropy in the embodiment of the present invention, which will not be repeated here. The only difference is that when calculating the partitioning information entropy, the range is all partitioning vectors corresponding to a certain classification value, while when calculating the complete information entropy, the range is all historical customer feature vectors.

[0184] For example, based on the partitioning rate and partitioning information entropy corresponding to different classification values ​​that the element type can take, a specific implementation method of obtaining the root conditional entropy of the element type can refer to the formula in the embodiment of the present invention:

[0185]

[0186] The relevant description will not be repeated here.

[0187] Through the above steps, the root conditional entropy of each element type can be obtained by the standard method of calculating conditional entropy, so that the obtained root conditional entropy has a higher accuracy, thereby improving the accuracy of the decision tree constructed in the subsequent steps and the related processing based on the decision tree, and improving the accuracy of the overall push.

[0188] In an optional embodiment, obtaining the sub-conditional entropy of each element type in the child node according to the historical discount labels corresponding to all the sub-historical customer feature vectors and the classification values ​​of the vector elements includes:

[0189] According to the classification values ​​of the vector elements of the sub-historical customer feature vector, respectively obtaining the number of sub-divisions of the vector elements with different classification values ​​in the element type of the sub-historical customer feature vector;

[0190] Obtaining a sub-division rate according to the sub-division quantity and the sub-vector quantity of the sub-historical customer feature vector;

[0191] The sub-historical customer feature vectors corresponding to different classification values ​​in the element type are respectively used as corresponding sub-partition vectors, and based on the sub-partition vectors corresponding to different classification values ​​in the element type and the historical discount labels corresponding to the sub-partition vectors, the sub-partition information entropy corresponding to the different classification values ​​is obtained;

[0192] Based on the sub-division rates and sub-division information entropies corresponding to different classification values ​​that the element type can take, the sub-conditional entropy of the element type is obtained.

[0193] Exemplarily, the specific principle of obtaining the number of sub-divisions of the vector elements with different classification values ​​in the element type of the sub-historical customer feature vector according to the classification values ​​of the vector elements of the sub-historical customer feature vector can be referred to the description of the steps of obtaining the number of divisions of the vector elements with different classification values ​​in the element type according to the classification values ​​of the vector elements in the embodiment of the present invention, which will not be repeated here.

[0194] Exemplarily, the specific principle of obtaining the sub-division rate based on the number of sub-divisions and the number of sub-vectors of the sub-historical customer feature vectors can be referred to the description of the step of obtaining the division rate based on the number of divisions and the total number of all historical customer feature vectors in the embodiment of the present invention, and will not be repeated here.

[0195] Exemplarily, the specific principle of using the sub-historical customer feature vectors corresponding to different classification values ​​in the element type as corresponding sub-partitioning vectors, and obtaining the sub-partitioning information entropy corresponding to different classification values ​​based on the sub-partitioning vectors corresponding to different classification values ​​in the element type and the historical discount labels corresponding to the sub-partitioning vectors, can refer to the description of the steps of using the historical customer feature vectors corresponding to different classification values ​​in the element type as corresponding partitioning vectors, and obtaining the partitioning information entropy corresponding to different classification values ​​based on the partitioning vectors corresponding to different classification values ​​in the element type and the historical discount labels corresponding to the partitioning vectors in the embodiments of the present invention, which will not be repeated here.

[0196] Exemplarily, the specific principle of obtaining the sub-conditional entropy of the element type based on the sub-division rate and sub-division information entropy corresponding to different classification values ​​that the element type can take can refer to the description of the steps of obtaining the root conditional entropy of the element type based on the division rate and division information entropy corresponding to different classification values ​​that the element type can take in the embodiment of the present invention, which will not be repeated here.

[0197] Through the above steps, the sub-conditional entropy of each element type in a specific child node can be obtained by using the standard method of calculating conditional entropy, so that the obtained sub-conditional entropy has a higher accuracy, thereby improving the accuracy of the decision tree constructed in the subsequent steps and the related processing based on the decision tree, and improving the accuracy of the overall push.

[0198] In an optional embodiment, obtaining the weight of the element type of the vector element based on a preset decision tree includes:

[0199] Based on the element type, determining the total number of nodes in the decision tree corresponding to the child node attributes and the root node attributes that are the same as the element type;

[0200] The total number of nodes is used as the weight of the element type.

[0201] Exemplarily, the method of determining the total number of nodes in the decision tree corresponding to the child node attributes and root node attributes of the same element type based on the element type, and using the total number of nodes as the weight of the element type, can include, but is not limited to, traversing the root node and all child nodes in the decision tree and obtaining the node attributes corresponding to each node, determining the number of nodes belonging to the node attributes of the same element type as the total number of nodes, and using the total number of nodes as the weight of the element type. For example, for a certain element type "amount of funds transferred between upstream and downstream enterprises," in a corresponding decision tree, the number of child nodes with the node attribute "amount of funds transferred between upstream and downstream enterprises" is 2, and the node attribute of the root node is also "amount of funds transferred between upstream and downstream enterprises." Therefore, the total number of nodes is 3, and the weight of the element type "amount of funds transferred between upstream and downstream enterprises" is also 3. It should be noted that the specific implementation of determining the total number of nodes in the decision tree corresponding to the child node attributes and root node attributes of the same element type based on the element type and using the total number of nodes as the weight of the element type can be determined by those skilled in the art based on actual circumstances. The above description is merely an example and does not constitute a limitation.

[0202] Since the number of nodes to which the node attributes corresponding to the element type belong in the decision tree is one of the parameters that mainly reflects the degree of influence of the element type on the overall vector characteristics, the above steps can make the determined weight consistent with and positively correlated with the degree of influence of the corresponding element type on the overall vector characteristics, thereby improving the accuracy of the determined weight, and further improving the accuracy of weight-based clustering and determination of the user type of the current user feature vector in subsequent steps.

[0203] In an optional embodiment, obtaining the weight of the element type of the vector element based on a preset decision tree includes:

[0204] Based on the element type, determining the depth of a child node corresponding to a child node attribute of the same type as the element in the decision tree;

[0205] Determine whether the root node attribute in the decision tree is the same as the element type; if not, obtain the weight of the element type according to the depth;

[0206] If so, a sub-weight of the element type is obtained according to the depth, and the sub-weight is added to a preset root weight to obtain the weight of the element type.

[0207] Exemplarily, the depth of the child node corresponding to the child node attribute of the same element type in the decision tree based on the element type can be determined, but not limited to, the path length from each child node of the child node attribute of the same element type in the decision tree to the root node is determined as the depth of the corresponding child node, wherein determining the path length between the child node and the root node is a conventional technical means in the art and will not be repeated here. For example, Figure 3 As shown, the path length between node 3 and root node 1 is 1, and the path length between node 6 and root node 1 is 2.

[0208] Exemplarily, the weight of the element type obtained according to the depth can be obtained by, but is not limited to, subtracting the height of the decision tree (the path length from the node with the greatest depth in the decision tree to the root node) from the depth of each child node corresponding to the child node attribute with the same element type in the decision tree to obtain the relative sub-weight of each child node, and superimposing or averaging the relative sub-weight of each child node to obtain the weight of the element type. It should be noted that the specific implementation method for obtaining the weight of the element type according to the depth can be determined by those skilled in the art according to actual conditions, and the above description is only an example and does not constitute a limitation thereto.

[0209] Exemplarily, the specific implementation method of obtaining the sub-weight of the element type according to the depth can refer to the description of the step of obtaining the weight of the element type according to the depth in the embodiment of the present invention, which will not be repeated here.

[0210] For example, the specific value of the root weight can be determined by those skilled in the art according to actual conditions, and the embodiment of the present invention does not limit this. For example, the root weight can be set to but not limited to 5, 6 or 10, etc.

[0211] Since the depth of the node to which the node attribute corresponding to the element type belongs is one of the parameters in the decision tree that mainly reflects the degree of influence of the element type on the overall vector feature, the above steps can make the determined weight consistent with the degree of influence of the corresponding element type on the overall vector feature and positively correlated, thereby improving the accuracy of the determined weight, and then improving the accuracy of the subsequent steps of clustering and determining the user type of the current user feature vector based on the weight. Moreover, when the root node attribute in the decision tree is the same as the element type, it means that the degree of influence of the element type on the overall vector feature has additional additional strength. Therefore, when it is determined that the root node attribute in the decision tree is the same as the element type, the sub-weight of the element type is obtained according to the depth. It is reasonable to add the sub-weight to the preset root weight to obtain the weight of the element type, which can further improve the accuracy of the obtained weight.

[0212] In an optional embodiment, if Figure 4 As shown, the step of determining multiple target clusters based on the weights and historical customer feature vectors includes the following steps:

[0213] S401: Determine whether the historical discount label of the historical customer feature vector is a discount experience label. If so, use the historical customer feature vector as a candidate vector.

[0214] S402: Based on the weights, determine a plurality of element types with a preset reserved number and the largest weights as standard types.

[0215] S403: Select multiple cluster center vectors from the candidate vectors, and determine multiple target clusters based on the candidate vectors and vector elements of standard types of the cluster center vectors.

[0216] Exemplarily, the preset retention number can be determined by those skilled in the art according to actual conditions, and the embodiments of the present invention do not limit this. For example, the preset retention number can be but is not limited to 2, 3, 4, 5 or 6, preferably 6.

[0217] Exemplarily, the method of determining, based on the weights, a preset number of element types with the largest weights as standard types includes the following examples:

[0218] Assume that the element types of the vector elements include historical discount amount, historical discount probability, issuer information, bill holding behavior category, customer number and gender, and the weight of the historical discount amount has been determined to be 6, the weight of the historical discount probability is 8, the weight of the issuer information is 1, the weight of the bill holding behavior category is 5, the weight of the customer number is 2, and the weight of the gender is 2. At this time, the preset retention quantity has also been determined to be 2. It can be seen that the multiple element types of the preset retention quantity with the largest weight are historical discount probability and historical discount amount, that is, the standard types are historical discount probability and historical discount amount.

[0219] It should be noted that the specific implementation method for determining multiple element types with the largest preset retention number of weights as standard types based on the weights can be determined by those skilled in the art according to actual conditions. The above description is only an example and does not constitute a limitation to this.

[0220] Exemplarily, the selection of multiple cluster center vectors from the candidate vectors may include, but is not limited to, randomly selecting multiple candidate vectors as the cluster center vectors, where the number of cluster center vectors depends on the number of historical discount types desired to be obtained by classifying the candidate vectors. It should be noted that the specific implementation of selecting multiple cluster center vectors from the candidate vectors can be determined by those skilled in the art based on actual circumstances, and the above description is merely illustrative and does not constitute a limitation.

[0221] Through the above steps, the element type that has a greater impact on the feature can be determined more accurately and quickly as the standard type based on the size of the weight, and clustering and determination of customer types can be performed based on the standard type in subsequent steps, thereby further reducing the amount of data that needs to be traversed and operated, and indirectly improving the speed and accuracy of determining customer types in subsequent steps, and indirectly improving the speed and accuracy of pushing information on the overall bill discounting business.

[0222] In an optional embodiment, determining a plurality of target clusters based on the candidate vectors and the vector elements of the standard type of the cluster center vectors includes:

[0223] Using standard-type vector elements of the candidate vector and the cluster center vector as standard elements, and using other vectors in the candidate vector except the cluster center vector as vectors to be classified;

[0224] Determining a first Euclidean distance between each of the vectors to be classified and the cluster center vector based on the standard element;

[0225] Determining, based on the first Euclidean distance, a nearest cluster center vector that is closest to the vector to be classified;

[0226] Clustering the plurality of vectors to be classified having the same nearest cluster center vector respectively to obtain a plurality of initial target clusters;

[0227] Repeat the clustering iteration step until there is a vector to be classified other than the cluster center vector whose corresponding second Euclidean distance is less than a preset distance threshold, wherein the clustering iteration step includes: based on multiple initial target clusters, obtaining a corresponding intermediate cluster center vector, and using other vectors in the candidate vectors except the intermediate cluster center vector as the vector to be classified; based on the standard element, determining the second Euclidean distance between each of the vector to be classified and the intermediate cluster center vector; based on the second Euclidean distance, determining the nearest intermediate cluster center vector closest to the vector to be classified; clustering multiple vectors to be classified having the same nearest intermediate cluster center vector to obtain an intermediate target cluster; and using the intermediate target cluster as the initial target cluster;

[0228] A plurality of the intermediate target clusters are used as the target clusters.

[0229] Exemplarily, the vector elements of the standard type of the candidate vector and the cluster center vector are used as standard elements, and the following examples are provided:

[0230] There is an alternative vector A (900000, 0.5, 123456, 1, 54321, 0), where 900000 represents the vector element in vector A whose element type is historical discount amount (and the element value of the vector element is 900000), 0.5 represents the vector element in vector A whose element type is historical discount probability (and the element value of the vector element is 0.5), 123456 represents the vector element in vector A whose element type is issuer information (and the element value of the vector element is 123456), 1 represents the vector element in vector A whose element type is ticket holding behavior category (and the element value of the vector element is 1), 54321 represents the vector element in vector A whose element type is customer number (and the element value of the vector element is 54321), and 0 represents the vector element in vector A whose element type is gender (and the element value of the vector element is 0).

[0231] Since the standard type is known to be the historical discount probability and the historical discount amount, we know that the standard type vector elements of candidate vector A are (0.5) and (900000), respectively. The same applies to using the standard type vector elements of the cluster center vector as the standard elements.

[0232] It should be noted that the specific implementation method of using the standard type vector elements of the candidate vector and the cluster center vector as standard elements can be determined by those skilled in the art according to actual conditions. The above description is only an example and does not constitute a limitation to this.

[0233] Exemplarily, determining the first Euclidean distance between each of the vectors to be classified and the cluster center vector based on the standard element may be, but is not limited to, first truncating the vectors to be classified and the cluster center vectors by vector elements of the standard type to obtain a computational vector to be classified and a computational cluster center vector, and then calculating the Euclidean distance (also known as the Euclidean distance) between the computational vector to be classified and the computational cluster center vector as the first Euclidean distance. Examples of truncating the vector elements of the standard type to obtain the computational vector include:

[0234] It is known that there is an alternative vector A (900000, 0.5, 123456, 1, 54321, 0), where the vector elements of the standard type are (900000) and (0.5), then the alternative vector A corresponding to the operation of the alternative vector A is (900000, 0.5).

[0235] The Euclidean distance between the vector to be classified and the cluster center vector is obtained as the first Euclidean distance, which is a conventional technical means in this field and can be expressed as the following formula:

[0236]

[0237] Among them, dist ed (x1, x2) represents the Euclidean distance between the vector to be classified and the cluster center vector, while x1 represents the vector to be classified, x2 represents the cluster center vector, and x 11 The element value of the vector element of the first standard type of the vector to be classified for operation, x 21 The element value of the vector element of the first standard type representing the cluster center vector used for calculation, x 21 The element value of the vector element of the second standard type representing the vector to be classified for operation, x 22 The element value of the vector element of the second standard type representing the cluster center vector used for calculation (in this example, the number of standard types is 2, and the first standard type is set to the historical discount probability, and the second standard type is set to the historical discount amount).

[0238] It should be noted that the specific implementation method of determining the first Euclidean distance between each of the vectors to be classified and the cluster center vector based on the standard elements can be determined by those skilled in the art according to actual conditions. The above description is only an example and does not constitute a limitation to this.

[0239] Exemplarily, the determining of the nearest cluster center vector closest to the vector to be classified based on the first Euclidean distance can be that, for a certain vector to be classified, the cluster center vector having the smallest first Euclidean distance to the vector to be classified is determined as the nearest cluster center vector of the vector to be classified.

[0240] For example, clustering multiple vectors is a conventional technical means in this field and will not be described in detail here.

[0241] Exemplarily, the principle of determining the second Euclidean distance between each of the vectors to be classified and the intermediate cluster center vector based on the standard element can be referred to the description of the step of determining the first Euclidean distance between each of the vectors to be classified and the cluster center vector based on the standard element in the embodiment of the present invention, which will not be repeated here.

[0242] Exemplarily, the principle of determining the nearest intermediate cluster center vector closest to the vector to be classified based on the second Euclidean distance can be referred to the description of the steps of determining the nearest cluster center vector closest to the vector to be classified based on the first Euclidean distance in the embodiment of the present invention, and will not be repeated here.

[0243] Exemplarily, the existence of a vector to be classified other than the cluster center vector whose corresponding second Euclidean distance is less than the preset distance threshold specifically refers to the existence of a Euclidean distance between a certain vector to be classified and a certain cluster center vector that is less than the preset distance threshold, wherein the preset distance threshold can be determined by those skilled in the art based on actual conditions, and the embodiments of the present invention do not impose any restrictions on this. For example, the preset distance threshold can be but is not limited to 10, 20 or 30, etc.

[0244] Through the above steps, vectors can be clustered around cluster centers based on Euclidean distance. Since Euclidean distance accurately reflects the proximity of features between vectors, the vector elements in a cluster determined based on Euclidean distance are closer to the features of the cluster center, resulting in a higher cluster accuracy. Furthermore, based on the current cluster, the cluster center is continuously adjusted and the cluster is continuously redefined based on the Euclidean distance between its corresponding center of gravity and other vectors. Furthermore, the spatial distribution relationship between multiple vectors can be further considered to modify the clustering. Therefore, the accuracy of the determined clusters can be improved through continuous modification. When there are vectors to be classified other than the cluster center vector whose corresponding second Euclidean distance is less than the preset distance threshold, it means that the iterative process of re-determining the cluster has become stable, and the differences between the clusters corrected in multiple cycles are no longer significant. Therefore, the clustering situation can be frozen, and the multiple clusters determined at this time are used as the final target clusters. Since the iteration has become stable and each cycle will improve the accuracy of the determined clusters, the final target clusters determined through the above steps can be made extremely accurate. The vector elements within each target cluster are extremely close in characteristics. This also improves the accuracy of determining the customer type of the current customer based on the target cluster in the subsequent steps, thereby improving the accuracy of the overall bill discount business information push.

[0245] In an optional embodiment, the step of clustering multiple initial targets to obtain corresponding intermediate cluster center vectors includes:

[0246] Obtaining, according to all the eigenvectors included in the initial target cluster, a mean vector of all the eigenvectors corresponding to the initial target cluster;

[0247] The mean vector is used as the intermediate cluster center vector.

[0248] Exemplarily, the method obtains the mean vector of all eigenvectors corresponding to the initial target cluster based on all eigenvectors included in the initial target cluster, specifically by superimposing all eigenvectors included in the initial target cluster to obtain a sum vector, and dividing the sum vector by the number of initial target cluster elements of the eigenvectors included in the initial target cluster to obtain the mean vector.

[0249] Through the above steps, the center of gravity of the spatial features of the initial target cluster can be accurately and quickly used as the new cluster center, thereby improving the accuracy of the subsequent redetermination of the new intermediate target cluster around the new cluster center, and then the final target cluster can be determined more quickly with fewer iterations, thereby indirectly improving the accuracy and speed of the overall push.

[0250] In an optional embodiment, judging whether the current customer corresponding to the current customer feature vector has a discount willingness according to the decision tree, and if so, determining the customer type corresponding to the current customer feature vector based on the clustering includes:

[0251] Determining a corresponding path in a decision tree according to the classification value of the vector element of the current customer feature vector;

[0252] According to the path, a corresponding leaf node is obtained;

[0253] It is determined whether the historical discount label corresponding to the leaf node is a label with discount experience. If so, the customer type corresponding to the current customer feature vector is determined based on the target cluster.

[0254] Exemplarily, according to the classification value of the vector element of the current customer feature vector, a corresponding path is determined in the decision tree. There are the following examples:

[0255] If the current customer feature vector B is (800000, 0.8, 111111, 2, 52013, 1), 800000 represents the vector element of the historical discount amount in vector B (and the element value of the vector element is 800000), 0.8 represents the vector element of the historical discount probability in vector B (and the element value of the vector element is 0.8), 111111 represents the vector element of the issuer information in vector B (and the element value of the vector element is 111111), 2 represents the vector element of the ticket holding behavior category in vector B (and the element value of the vector element is 111111). The element value of element 2, 52013 represents the vector element in vector B whose element type is customer number (and whose element value is 52013), and 1 represents the vector element in vector B whose element type is gender (and whose element value is 1). Furthermore, it is known that the classification values ​​corresponding to the current customer feature vector can be expressed as (2, 100, 31, 2, 200, 1) (classification values ​​at corresponding positions correspond to vector elements at corresponding positions in feature vector B. For example, the first classification value 2 corresponds to the historical discount amount vector element 800,000, and the second classification value 100 corresponds to the historical discount probability vector element 0.8).

[0256] The classification values ​​corresponding to each element of the current customer feature vector B are input into the decision tree. If the root node attribute of the decision tree is the historical discount amount, then the path where the classification value of the root node attribute is 2 is followed to reach the first child node; if the sub-node attribute of the first child node is the historical discount probability, then the path where the classification value of the sub-node attribute is 100 is followed to reach the second child node; if the sub-node attribute of the second child node is the ticket holding behavior category, then the path where the classification value of the sub-node attribute is 2 is followed to reach the third child node; if the sub-node attribute of the third child node is the historical discount probability, then the path where the classification value of the sub-node attribute is 2 is followed to reach the fourth child node; if the sub-node attribute of the fourth child node is gender, then the path where the classification value of the sub-node attribute is 1 is followed to reach the fifth child node. If it is found that the fifth child node is already a leaf node, then the corresponding path can be determined to be: root node -> first child node -> second child node -> third child node -> fourth child node -> a leaf node.

[0257] It should be noted that determining the corresponding path in the decision tree based on the classification value of the vector element of the current customer feature vector is a basic implementation method of existing classification using decision trees. The specific content can be determined by those skilled in the art based on actual conditions. The above description is only an example and does not constitute a limitation to this.

[0258] Exemplarily, to determine whether the historical discount label corresponding to the leaf node is a label with discount experience, it can be, but is not limited to, first determining the corresponding leaf historical customer feature vector based on the leaf node, and obtaining the corresponding historical discount label based on the leaf historical customer feature vector (the historical discount label values ​​corresponding to each leaf historical customer feature vector in a leaf node are consistent) as the historical discount label corresponding to the leaf node, and then making a judgment.

[0259] Through the above steps, a decision tree can be used in a standard way to predict whether the current customer corresponding to the current customer feature vector will be discounted in the future. The prediction speed is fast and the accuracy is high. Therefore, the speed and accuracy of the step of preliminarily screening out the current customer feature vectors corresponding to the current customers who are willing to discount can be improved, thereby improving the speed and accuracy of the overall push.

[0260] In an optional embodiment, if Figure 5 As shown, the determining the customer type corresponding to the current customer feature vector based on the target cluster includes the following steps:

[0261] S501: Based on the target clusters, obtain all final history vectors included in each of the target clusters.

[0262] S502: Based on the final historical vector, a corresponding individual discount degree value is obtained, and an average value of the individual discount degree values ​​is used as the overall discount degree value of the target cluster.

[0263] S503: Sort the target clusters according to the overall discount degree values, so as to determine the historical discount type corresponding to the target cluster based on the sorting order.

[0264] S504: Taking the average value of all final historical vectors of each target cluster as the average vector corresponding to the target cluster, and determining the final Euclidean distance between each average vector and the current customer feature vector.

[0265] S505: The historical discount type of the target cluster corresponding to the average vector with the shortest final Euclidean distance is used as the customer type corresponding to the current customer feature vector.

[0266] Exemplarily, obtaining the corresponding individual discount level value based on the final historical vector may be, but is not limited to, comprehensively processing the element values ​​of one or more vector elements related to the discount level in the final historical vector to obtain the individual discount level value corresponding to the final historical vector. For example, the element values ​​of the vector elements corresponding to the historical discount probability, the element values ​​of the vector elements corresponding to the historical discount amount, and the element values ​​of the vector elements corresponding to the total number of tickets held in the final historical vector may be superimposed or averaged to obtain the individual discount level value corresponding to the final historical vector. It should be noted that the specific implementation method for obtaining the corresponding individual discount level value based on the final historical vector can be determined by those skilled in the art based on actual circumstances, and the above description is merely an example and does not constitute a limitation thereto.

[0267] Exemplarily, the target clusters are sorted by the overall discount value, which may be, but is not limited to, sorting from largest to smallest or from smallest to largest. For example, if there are five target clusters, namely, target cluster A, target cluster B, target cluster C, target cluster D, and target cluster E, where the overall discount value corresponding to target cluster A is 10,000, the overall discount value corresponding to target cluster B is 5,000, the overall discount value corresponding to target cluster C is 15,000, the overall discount value corresponding to target cluster D is 12,500, and the overall discount value corresponding to target cluster E is 20,000, then if sorting by overall discount value from largest to smallest, the resulting order is: target cluster E, target cluster C, target cluster D, target cluster A, target cluster B.

[0268] Exemplarily, the determination of the historical discount type corresponding to the target cluster based on the sorting order may include, but is not limited to, first presetting different historical discount types (the number of which needs to be consistent with the number of target clusters), then sorting the different historical discount types according to their degree of discount in the same manner as the target clusters described above (both need to be sorted from large to small or from small to large), and then associating the target clusters and historical discount types at the same position in the two sorted sequences. For example, the pre-set historical discount types include "very high discount probability," "high discount probability," "medium discount probability," "low discount probability," and "very low discount probability." From the example above, it can be seen that the sorting order is: target cluster E, target cluster C, target cluster D, target cluster A, target cluster B. Thus, target cluster E corresponds to a historical discount type of "high discount probability," target cluster C corresponds to a historical discount type of "high discount probability," target cluster D corresponds to a historical discount type of "medium discount probability," target cluster A corresponds to a historical discount type of "low discount probability," and target cluster B corresponds to a historical discount type of "very low discount probability." It should be noted that the specific implementation of determining the historical discount type corresponding to the target clusters based on the sorting order can be determined by those skilled in the art based on actual circumstances, and the above description is merely an example and does not constitute a limitation.

[0269] Exemplarily, determining the average value (average vector) of multiple vectors is a conventional technical means in this field and will not be described in detail here.

[0270] Exemplarily, determining the final Euclidean distance between each of the average vectors and the current customer feature vector (the nature of which is to calculate the Euclidean distance between different vectors) is a conventional technical means in this field and will not be described in detail here.

[0271] Exemplarily, the historical discount type of the target cluster corresponding to the average vector with the closest final Euclidean distance is used as the customer type corresponding to the current customer feature vector. This can be accomplished by, for the current customer feature vector, using the historical discount type of the target cluster corresponding to the average vector closest to it (with the shortest final Euclidean distance between them) as the customer type corresponding to the current customer feature vector. For example, if the historical discount type of the target cluster corresponding to the average vector closest to the current customer feature vector is "high discount probability," then the customer type corresponding to the current customer feature vector is determined to be "high discount probability."

[0272] Through the above steps, it is possible to determine the corresponding discount type for different target clusters based on the characteristics of the discount degree in the historical customer feature vectors contained therein (which can reflect the actual historical discount degree), thereby making the discount type corresponding to the target cluster clear and accurate. The degree of feature proximity between the current customer feature vector and different target clusters can be determined based on the Euclidean distance between the center of gravity of the target cluster and the current customer feature vector, thereby improving the accuracy of determining the degree of feature proximity between the current customer feature vector and different target clusters, thereby improving the accuracy of subsequently determining the type corresponding to the target cluster with the highest feature proximity as the customer type of the current customer feature vector. Therefore, the above steps directly improve the accuracy and efficiency of pushing bill discount business information to corresponding customers based on customer type.

[0273] The push method in the embodiment of the present invention can be automatically implemented through programs, software, functions or applications, thereby greatly reducing the degree of staff intervention required, thereby greatly improving the speed and efficiency of pushing bill discount business information, and having strong accuracy.

[0274] Based on the same principle, the embodiment of the present invention discloses a bill discount business information push device 600, such as Figure 6 As shown, the bill discount business information push device 600 includes:

[0275] The classification value determination module 601 is configured to classify the vector elements of the preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the historical customer feature vector; and classify the vector elements of the current customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the current customer feature vector;

[0276] A decision tree processing module 602 is configured to obtain weights of element types of vector elements based on a preset decision tree, wherein the decision tree is associated with a plurality of historical customer feature vectors, classification values ​​of the historical customer feature vectors, and corresponding historical discount labels;

[0277] Push module 603 is used to determine multiple target clusters based on the weights and historical customer feature vectors; determine whether the current customer corresponding to the current customer feature vector has a willingness to discount according to the decision tree; if so, determine the customer type corresponding to the current customer feature vector based on the target cluster; and push bill discount business information to the current customer according to the customer type.

[0278] In an optional embodiment, the system further includes a historical customer information vectorization module for:

[0279] Before classifying the vector elements of the preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element,

[0280] Based on multiple preset customer authentication codes, corresponding initial historical customer information is obtained respectively;

[0281] Performing data cleaning, data extraction, and data standardization on the initial historical customer information to obtain intermediate historical customer information;

[0282] The intermediate historical customer information is subjected to feature vector quantization processing to obtain the historical customer feature vector.

[0283] In an optional embodiment, the system further includes a current customer information vectorization module for:

[0284] Before classifying the vector elements of the current customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the current customer feature vector,

[0285] The preset current customer information is subjected to feature vectorization processing to obtain the current customer feature vector.

[0286] In an optional embodiment, the classification value determination module 601 is configured to:

[0287] Based on the element value, the element type of the corresponding vector element, and a plurality of different element value intervals corresponding to the preset element type, determining, from the element value interval, a belonging interval corresponding to the element value of each vector element of the historical customer feature vector;

[0288] Based on the attribution interval, a classification value of a corresponding vector element in the historical customer feature vector is obtained.

[0289] In an optional embodiment, the method further includes a decision tree construction module for:

[0290] Before obtaining the weight of the element type of the vector element based on the preset decision tree,

[0291] A decision tree is constructed based on the plurality of historical customer feature vectors, the classification values ​​of the historical customer feature vectors, and the corresponding historical discount labels.

[0292] In an optional embodiment, the decision tree construction module is used to:

[0293] Obtaining complete information entropy based on all the historical customer feature vectors and corresponding historical discount labels; wherein the value of the historical discount label is a label with discount experience or a label without discount experience;

[0294] Obtaining the root conditional entropy of each element type according to the historical discount labels corresponding to all the historical customer feature vectors and the classification values ​​of the vector elements;

[0295] According to the complete information entropy and the root conditional entropy, the root information gain entropy corresponding to the element type is obtained, and the element type with the largest root information gain entropy is used as the root node attribute to establish the root node of the decision tree; based on each classification value that the root node attribute can take, a child node corresponding to each classification value is established respectively;

[0296] Repeat the step of establishing child nodes until no child nodes can be established to complete the construction of the decision tree, wherein the step of establishing child nodes includes:

[0297] According to the sub-classification value corresponding to each of the sub-nodes, determining a plurality of historical customer feature vectors having vector elements corresponding to the sub-classification value as sub-historical customer feature vectors of the sub-node;

[0298] Determining whether the historical discount labels corresponding to the sub-historical customer feature vectors of each of the sub-nodes are all the same, and if so, treating the sub-nodes as leaf nodes; and determining, based on the leaf classification value corresponding to each leaf node, a plurality of historical customer feature vectors having vector elements corresponding to the leaf classification value as leaf historical customer feature vectors of the leaf node;

[0299] If not, obtaining the sub-conditional entropy of each element type in the sub-node according to the historical discount labels corresponding to all the sub-historical customer feature vectors and the classification values ​​of the vector elements;

[0300] According to the complete information entropy and the sub-conditional entropy, the sub-information gain entropy corresponding to the element type is obtained, and the element type with the largest sub-information gain entropy is used as the child node attribute; and the child node of the next layer of the child node is established based on each classification value that the child node attribute can take.

[0301] In an optional embodiment, the decision tree construction module is used to:

[0302] Obtaining, based on all the historical customer feature vectors and the corresponding historical discount labels, a first number of discounted historical customer feature vectors whose historical discount labels are values ​​of discount experience labels, and a second number of undiscounted historical customer feature vectors whose historical discount labels are values ​​of non-discount experience labels;

[0303] Obtaining a discount rate based on the first quantity and the total quantity of all historical customer feature vectors;

[0304] obtaining an undiscounted rate based on the second quantity and the total quantity of all historical customer feature vectors;

[0305] Based on the discounted rate and the undiscounted rate, a complete information entropy is obtained.

[0306] In an optional embodiment, the decision tree construction module is used to:

[0307] According to the classification value of the vector element, respectively obtain the number of divisions of the vector elements with different classification values ​​in the element type;

[0308] Obtaining a division rate according to the number of divisions and the total number of all historical customer feature vectors;

[0309] Using historical customer feature vectors corresponding to different classification values ​​in the element type as corresponding partitioning vectors, and obtaining partitioning information entropies corresponding to different classification values ​​based on the partitioning vectors corresponding to different classification values ​​in the element type and historical discount labels corresponding to the partitioning vectors;

[0310] Based on the partitioning rates and partitioning information entropies corresponding to different classification values ​​that the element type can take, the root conditional entropy of the element type is obtained.

[0311] In an optional embodiment, the decision tree construction module is used to:

[0312] According to the classification values ​​of the vector elements of the sub-historical customer feature vector, respectively obtaining the number of sub-divisions of the vector elements with different classification values ​​in the element type of the sub-historical customer feature vector;

[0313] Obtaining a sub-division rate according to the sub-division quantity and the sub-vector quantity of the sub-historical customer feature vector;

[0314] The sub-historical customer feature vectors corresponding to different classification values ​​in the element type are respectively used as corresponding sub-partition vectors, and based on the sub-partition vectors corresponding to different classification values ​​in the element type and the historical discount labels corresponding to the sub-partition vectors, the sub-partition information entropy corresponding to the different classification values ​​is obtained;

[0315] Based on the sub-division rates and sub-division information entropies corresponding to different classification values ​​that the element type can take, the sub-conditional entropy of the element type is obtained.

[0316] In an optional embodiment, the decision tree processing module 602 is configured to:

[0317] Based on the element type, determining the total number of nodes in the decision tree corresponding to the child node attributes and the root node attributes that are the same as the element type;

[0318] The total number of nodes is used as the weight of the element type.

[0319] In an optional embodiment, the decision tree processing module 602 is configured to:

[0320] Based on the element type, determining the depth of a child node corresponding to a child node attribute of the same type as the element in the decision tree;

[0321] Determine whether the root node attribute in the decision tree is the same as the element type; if not, obtain the weight of the element type according to the depth;

[0322] If so, a sub-weight of the element type is obtained according to the depth, and the sub-weight is added to a preset root weight to obtain the weight of the element type.

[0323] In an optional implementation, the push module 603 is configured to:

[0324] Determine whether the historical discount label of the historical customer feature vector is a discount experience label, and if so, use the historical customer feature vector as a candidate vector;

[0325] Based on the weights, determining a plurality of element types with the largest weights and a preset number of reserved elements as standard types;

[0326] A plurality of cluster center vectors are selected from the candidate vectors, and a plurality of target clusters are determined based on vector elements of standard types of the candidate vectors and the cluster center vectors.

[0327] In an optional implementation, the push module 603 is configured to:

[0328] Using standard-type vector elements of the candidate vector and the cluster center vector as standard elements, and using other vectors in the candidate vector except the cluster center vector as vectors to be classified;

[0329] Determining a first Euclidean distance between each of the vectors to be classified and the cluster center vector based on the standard element;

[0330] Determining, based on the first Euclidean distance, a nearest cluster center vector that is closest to the vector to be classified;

[0331] Clustering the plurality of vectors to be classified having the same nearest cluster center vector respectively to obtain a plurality of initial target clusters;

[0332] Repeat the clustering iteration step until there is a vector to be classified other than the cluster center vector whose corresponding second Euclidean distance is less than a preset distance threshold, wherein the clustering iteration step includes: based on multiple initial target clusters, obtaining a corresponding intermediate cluster center vector, and using other vectors in the candidate vectors except the intermediate cluster center vector as the vector to be classified; based on the standard element, determining the second Euclidean distance between each of the vector to be classified and the intermediate cluster center vector; based on the second Euclidean distance, determining the nearest intermediate cluster center vector closest to the vector to be classified; clustering multiple vectors to be classified having the same nearest intermediate cluster center vector to obtain an intermediate target cluster; and using the intermediate target cluster as the initial target cluster;

[0333] A plurality of the intermediate target clusters are used as the target clusters.

[0334] In an optional implementation, the push module 603 is configured to:

[0335] Obtaining, according to all the eigenvectors included in the initial target cluster, a mean vector of all the eigenvectors corresponding to the initial target cluster;

[0336] The mean vector is used as the intermediate cluster center vector.

[0337] In an optional implementation, the push module 603 is configured to:

[0338] Determining a corresponding path in a decision tree according to the classification value of the vector element of the current customer feature vector;

[0339] According to the path, a corresponding leaf node is obtained;

[0340] It is determined whether the historical discount label corresponding to the leaf node is a label with discount experience. If so, the customer type corresponding to the current customer feature vector is determined based on the target cluster.

[0341] In an optional implementation, the push module 603 is configured to:

[0342] Based on the target clusters, obtaining all final history vectors included in each of the target clusters;

[0343] Based on the final history vector, a corresponding individual discount degree value is obtained, and an average value of the individual discount degree values ​​is used as the overall discount degree value of the target cluster;

[0344] sorting the target clusters according to the overall discount degree values, so as to determine the historical discount type corresponding to the target clusters based on the sorting order;

[0345] Taking the average value of all final historical vectors of each target cluster as the average vector corresponding to the target cluster, and determining the final Euclidean distance between each average vector and the current customer feature vector;

[0346] The historical discount type of the target cluster corresponding to the average vector with the shortest final Euclidean distance is used as the customer type corresponding to the current customer feature vector.

[0347] Since the principle of solving the problem by the bill discounting business information pushing device 600 is similar to that of the above method, the implementation of the bill discounting business information pushing device 600 can refer to the implementation of the above method and will not be repeated here.

[0348] The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer device. Specifically, the computer device may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0349] In a typical example, a computer device specifically includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method described above is implemented.

[0350] Reference below Figure 7 , which shows a schematic structural diagram of a computer device 700 suitable for implementing an embodiment of the present application.

[0351] like Figure 7 As shown, computer device 700 includes a central processing unit (CPU) 701, which can perform various appropriate tasks and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage portion 708 into a random access memory (RAM) 703. Various programs and data required for the operation of system 700 are also stored in RAM 703. CPU 701, ROM 702, and RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to bus 704.

[0352] The following components are connected to the I / O interface 705: an input section 706 including a keyboard, a mouse, and the like; an output section 707 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 708 including devices such as a hard disk; and a communication section 709 including a network interface card such as a LAN card or a modem. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the I / O interface 705 as needed. Removable media 711, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 710 as needed, so that computer programs read therefrom can be installed in the storage section 708 as needed.

[0353] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program tangibly embodied on a machine-readable medium, the computer program including program code for executing the methods illustrated in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 709 and / or installed from removable media 711.

[0354] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0355] For the convenience of description, the above devices are described as being divided into various units according to their functions. Of course, when implementing this application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0356] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0357] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0358] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0359] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0360] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0361] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communications network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0362] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

[0363] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A method for pushing bill discount business information, characterized in that: include: Classify the vector elements of the preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the historical customer feature vector; classify the vector elements of the current customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the current customer feature vector; Based on a plurality of the historical customer feature vectors, the classification values ​​of the historical customer feature vectors and the corresponding historical discount labels, a decision tree is constructed, including: obtaining complete information entropy according to all the historical customer feature vectors and the corresponding historical discount labels; wherein the value of the historical discount label is a label with discount experience or a label without discount experience; obtaining the root conditional entropy of each element type according to the historical discount labels and the classification values ​​of the vector elements corresponding to all the historical customer feature vectors; obtaining the root information gain entropy corresponding to the element type according to the complete information entropy and the root conditional entropy, and establishing the root node of the decision tree with the element type having the largest root information gain entropy as the root node attribute; establishing child nodes corresponding to each classification value based on each possible classification value of the root node attribute; repeatedly executing the step of establishing child nodes until no child nodes can be established, so as to complete the construction of the decision tree, wherein the step of establishing child nodes includes: obtaining the root information gain entropy corresponding to each element type according to the complete information entropy and the root conditional entropy, and establishing the root node of the decision tree with the element type having the largest root information gain entropy as the root node attribute; establishing child nodes corresponding to each classification value respectively based on each classification value of the root node attribute; repeatedly executing the step of establishing child nodes until no child nodes can be established, so as to complete the construction of the decision tree, and the step of establishing child nodes includes: obtaining the root information gain entropy corresponding to each element type according to the complete information entropy and the root conditional entropy, and establishing the root node of the decision tree with the element type having the largest root information gain entropy as the root node attribute; According to the sub-classification value corresponding to the node, multiple historical customer feature vectors with vector elements corresponding to the sub-classification value are determined as the sub-historical customer feature vectors of the child node; whether the historical discount labels corresponding to the sub-historical customer feature vectors of each child node are the same, if so, the child node is used as a leaf node; according to the leaf classification value corresponding to each leaf node, multiple historical customer feature vectors with vector elements corresponding to the leaf classification value are determined as the leaf historical customer feature vector of the leaf node; if not, according to the historical discount labels corresponding to all the sub-historical customer feature vectors and the classification value of the vector element, the sub-conditional entropy of each element type in the child node is obtained; according to the complete information entropy and the sub-conditional entropy, the sub-information gain entropy corresponding to the element type is obtained, and the element type with the largest sub-information gain entropy is used as the child node attribute; based on each classification value that the child node attribute can take, a child node of the next layer of the child node is established; Obtaining weights of element types of vector elements based on a preset decision tree; wherein the decision tree is associated with a plurality of the historical customer feature vectors, the classification values ​​of the historical customer feature vectors, and the corresponding historical discount labels; Based on the weights and historical customer feature vectors, multiple target clusters are determined; based on the decision tree, it is determined whether the current customer corresponding to the current customer feature vector has a willingness to discount; if so, the customer type corresponding to the current customer feature vector is determined based on the target cluster; and based on the customer type, bill discount business information is pushed to the current customer.

2. The method according to claim 1, characterized in that Further including: Before classifying the vector elements of the preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the historical customer feature vector, Based on multiple preset customer authentication codes, corresponding initial historical customer information is obtained respectively; Performing data cleaning, data extraction, and data standardization on the initial historical customer information to obtain intermediate historical customer information; The intermediate historical customer information is subjected to feature vector quantization processing to obtain the historical customer feature vector.

3. The method according to claim 1, characterized in that Further including: Before classifying the vector elements of the current customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the current customer feature vector, The preset current customer information is subjected to feature vectorization processing to obtain the current customer feature vector.

4. The method according to claim 1, wherein The process of classifying the vector elements of the preset historical customer feature vector according to the corresponding element values ​​to obtain the classification value corresponding to each vector element in the historical customer feature vector includes: Based on the element value, the element type of the corresponding vector element, and a plurality of different element value intervals corresponding to the preset element type, determining, from the element value interval, a belonging interval corresponding to the element value of each vector element of the historical customer feature vector; Based on the attribution interval, a classification value of a corresponding vector element in the historical customer feature vector is obtained.

5. The method according to claim 1, wherein The complete information entropy is obtained based on all the historical customer feature vectors and the corresponding historical discount labels, including: Obtaining, based on all the historical customer feature vectors and the corresponding historical discount labels, a first number of discounted historical customer feature vectors whose historical discount labels are values ​​of discount experience labels, and a second number of undiscounted historical customer feature vectors whose historical discount labels are values ​​of no discount experience labels; Obtaining a discount rate based on the first quantity and the total quantity of all historical customer feature vectors; obtaining an undiscounted rate based on the second quantity and the total quantity of all historical customer feature vectors; Based on the discounted rate and the undiscounted rate, a complete information entropy is obtained.

6. The method according to claim 1, characterized in that Obtaining the root conditional entropy of each element type based on the historical discount labels and classification values ​​of the vector elements corresponding to all the historical customer feature vectors includes: According to the classification value of the vector element, respectively obtain the number of divisions of the vector elements with different classification values ​​in the element type; Obtaining a division rate according to the number of divisions and the total number of all historical customer feature vectors; Using historical customer feature vectors corresponding to different classification values ​​in the element type as corresponding partitioning vectors, and obtaining partitioning information entropies corresponding to different classification values ​​based on the partitioning vectors corresponding to different classification values ​​in the element type and historical discount labels corresponding to the partitioning vectors; Based on the partitioning rates and partitioning information entropies corresponding to different classification values ​​that the element type can take, the root conditional entropy of the element type is obtained.

7. The method according to claim 1, characterized in that Obtaining the sub-conditional entropy of each element type in the child node according to the historical discount labels corresponding to all the sub-historical customer feature vectors and the classification values ​​of the vector elements includes: According to the classification values ​​of the vector elements of the sub-historical customer feature vector, respectively obtaining the number of sub-divisions of the vector elements with different classification values ​​in the element type of the sub-historical customer feature vector; Obtaining a sub-division rate according to the sub-division quantity and the sub-vector quantity of the sub-historical customer feature vector; The sub-historical customer feature vectors corresponding to different classification values ​​in the element type are respectively used as corresponding sub-partition vectors, and based on the sub-partition vectors corresponding to different classification values ​​in the element type and the historical discount labels corresponding to the sub-partition vectors, the sub-partition information entropy corresponding to the different classification values ​​is obtained; Based on the sub-division rates and sub-division information entropies corresponding to different classification values ​​that the element type can take, the sub-conditional entropy of the element type is obtained.

8. The method according to claim 1, characterized in that The step of obtaining the weight of the element type of the vector element based on the preset decision tree includes: Based on the element type, determining the total number of nodes in the decision tree corresponding to the child node attributes and the root node attributes that are the same as the element type; The total number of nodes is used as the weight of the element type.

9. The method according to claim 1, characterized in that The step of obtaining the weight of the element type of the vector element based on the preset decision tree includes: Based on the element type, determining the depth of a child node corresponding to a child node attribute of the same type as the element in the decision tree; Determine whether the root node attribute in the decision tree is the same as the element type; if not, obtain the weight of the element type according to the depth; If so, a sub-weight of the element type is obtained according to the depth, and the sub-weight is added to a preset root weight to obtain the weight of the element type.

10. The method according to claim 1, characterized in that The determining of multiple target clusters based on the weights and historical customer feature vectors includes: Determine whether the historical discount label of the historical customer feature vector is a discount experience label, and if so, use the historical customer feature vector as a candidate vector; Based on the weights, determining a plurality of element types with the largest weights and a preset number of reserved elements as standard types; A plurality of cluster center vectors are selected from the candidate vectors, and a plurality of target clusters are determined based on vector elements of standard types of the candidate vectors and the cluster center vectors.

11. The method according to claim 10, characterized in that The determining of a plurality of target clusters according to the vector elements of the standard type of the candidate vector and the cluster center vector comprises: Using vector elements of standard types of the candidate vectors and the cluster center vector as standard elements, and using other vectors in the candidate vectors except the cluster center vector as vectors to be classified; Determining a first Euclidean distance between each of the vectors to be classified and the cluster center vector based on the standard element; Determining, based on the first Euclidean distance, a nearest cluster center vector that is closest to the vector to be classified; Clustering the plurality of vectors to be classified having the same nearest cluster center vector respectively to obtain a plurality of initial target clusters; Repeat the clustering iteration step until there is a vector to be classified other than the cluster center vector whose corresponding second Euclidean distance is less than a preset distance threshold, wherein the clustering iteration step includes: based on multiple initial target clusters, obtaining a corresponding intermediate cluster center vector, and using other vectors in the candidate vectors except the intermediate cluster center vector as the vector to be classified; based on the standard element, determining the second Euclidean distance between each of the vector to be classified and the intermediate cluster center vector; based on the second Euclidean distance, determining the nearest intermediate cluster center vector closest to the vector to be classified; clustering multiple vectors to be classified having the same nearest intermediate cluster center vector to obtain an intermediate target cluster; and using the intermediate target cluster as the initial target cluster; A plurality of the intermediate target clusters are used as the target clusters.

12. The method according to claim 11, characterized in that The method of clustering multiple initial targets to obtain corresponding intermediate cluster center vectors includes: Obtaining, according to all the eigenvectors included in the initial target cluster, a mean vector of all the eigenvectors corresponding to the initial target cluster; The mean vector is used as the intermediate cluster center vector.

13. The method according to claim 1, wherein The determining, according to the decision tree, whether the current customer corresponding to the current customer feature vector has a discount willingness, and if so, determining the customer type corresponding to the current customer feature vector based on the target cluster, includes: Determining a corresponding path in a decision tree according to the classification value of the vector element of the current customer feature vector; According to the path, a corresponding leaf node is obtained; It is determined whether the historical discount label corresponding to the leaf node is a label with discount experience. If so, the customer type corresponding to the current customer feature vector is determined based on the target cluster.

14. The method according to claim 1, wherein The determining the customer type corresponding to the current customer feature vector based on the target cluster includes: Based on the target clusters, obtaining all final history vectors included in each of the target clusters; Based on the final history vector, a corresponding individual discount degree value is obtained, and an average value of the individual discount degree values ​​is used as the overall discount degree value of the target cluster; sorting the target clusters according to the overall discount degree values, so as to determine the historical discount type corresponding to the target clusters based on the sorting order; Taking the average value of all final historical vectors of each target cluster as the average vector corresponding to the target cluster, and determining the final Euclidean distance between each average vector and the current customer feature vector; The historical discount type of the target cluster corresponding to the average vector with the shortest final Euclidean distance is used as the customer type corresponding to the current customer feature vector.

15. A bill discount business information push device, characterized in that: include: a classification value determination module, configured to classify the vector elements of a preset historical customer feature vector according to the corresponding element values, and obtain the classification value corresponding to each vector element in the historical customer feature vector; and classify the vector elements of a current customer feature vector according to the corresponding element values, and obtain the classification value corresponding to each vector element in the current customer feature vector; The decision tree construction module is used to: obtain complete information entropy based on all the historical customer feature vectors and the corresponding historical discount labels; wherein the value of the historical discount label is a discount experience label or a non-discount experience label; obtain the root conditional entropy of each element type based on the historical discount labels corresponding to all the historical customer feature vectors and the classification value of the vector elements; obtain the root information gain entropy corresponding to the element type based on the complete information entropy and the root conditional entropy, and use the element type with the largest root information gain entropy as the root node attribute to establish the root node of the decision tree; based on each classification value that the root node attribute can take, respectively establish a child node corresponding to each classification value; repeatedly execute the step of establishing child nodes until no child nodes can be established to complete the construction of the decision tree, and the step of establishing child nodes includes: determining the element type with the corresponding child classification value according to the child classification value corresponding to each child node The multiple historical customer feature vectors of the vector elements of the classification value are the sub-historical customer feature vectors of the child node; whether the historical discount labels corresponding to the sub-historical customer feature vectors of each of the child nodes are the same is determined respectively, and if so, the child node is taken as a leaf node; according to the leaf classification value corresponding to each leaf node, the multiple historical customer feature vectors with the vector elements corresponding to the leaf classification value are determined to be the leaf historical customer feature vector of the leaf node; if not, according to the historical discount labels corresponding to all the sub-historical customer feature vectors and the classification value of the vector elements, the sub-conditional entropy of each element type in the child node is obtained; according to the complete information entropy and the sub-conditional entropy, the sub-information gain entropy corresponding to the element type is obtained, and the element type with the largest sub-information gain entropy is taken as the child node attribute; based on each classification value that the child node attribute can take, a child node of the next layer of the child node is established; a decision tree processing module, configured to obtain weights of element types of vector elements based on a preset decision tree; wherein the decision tree is associated with a plurality of historical customer feature vectors, classification values ​​of the historical customer feature vectors, and corresponding historical discount labels; A push module is used to determine multiple target clusters based on the weights and historical customer feature vectors; determine whether the current customer corresponding to the current customer feature vector has a willingness to discount according to the decision tree; if so, determine the customer type corresponding to the current customer feature vector based on the target cluster; and push bill discount business information to the current customer according to the customer type.

16. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 14 is implemented.

17. A computer-readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 14 is implemented.

Images