Insurance business push method and system based on big data analysis

Abstract

The insurance business push method and system based on big data analysis provided by the present disclosure extracts the conversion coefficient of each positive insurance business in the positive insurance business vector representation set from the pushee vector representation containing the insurance business evaluation behavior of the pushee. Then, the push analysis algorithm can establish a targeted business list for the pushee through the pushee vector representation and the positive insurance business vector representation set. The targeted business list includes the conversion coefficient of each positive insurance business of the pushee and the ratio of the total conversion coefficient of the pushee to multiple positive insurance businesses. The targeted business list can characterize the distribution of the conversion coefficient of each positive insurance business in the positive insurance business vector representation set. Then, the insurance business push algorithm selects insurance business for the pushee according to the targeted business list of the pushee. The determined insurance business is more consistent with the distribution of the conversion coefficient of the pushee, which can improve the conversion rate of the insurance business.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing, and in particular, to a method and system for pushing insurance business based on big data analysis. Background Art

[0002] With the development of the insurance industry, various types of insurance have emerged in an endless stream. How companies can control the insurance purchasing needs of target users and push appropriate insurance services is the research direction of major insurance companies. When insurance companies promote insurance services on major promotion platforms, they can collect information about target users, such as basic user information and behavioral information. With the help of big data processing technology, insurance companies can identify the insurance purchasing needs of target users. At present, most companies often use the method of obtaining users' historical purchased, browsed or consulted insurance services, and call other similar insurance services for indiscriminate push. For the recipients, the experience is not good, and the pushed insurance services may not be suitable for the actual needs of users. Therefore, the conversion rate of most pushes is low. Based on this, a method that can provide targeted insurance business push is needed. Summary of the invention

[0003] The purpose of the present invention is to provide an insurance business push method and system based on big data analysis to improve the above-mentioned problems.

[0004] The technical solution of the embodiment of the present disclosure is implemented as follows:

[0005] In a first aspect, an embodiment of the present disclosure provides an insurance business push method based on big data analysis, which is applied to a data analysis system. The method includes:

[0006] Obtaining insurance business vector representations of several candidate insurance businesses and a recommended person vector representation of a recommended person; wherein the recommended person vector representation of the recommended person is obtained through recommended description information of the recommended person and insurance business evaluation behavior of the recommended person;

[0007] The insurance business vector representation set of the recommended person is obtained by obtaining the insurance business vector representations of the multiple positive insurance businesses that the recommended person is concerned about, and forming the insurance business vector representation set of the multiple positive insurance businesses;

[0008] Inputting the pushee vector representation and the positive insurance business vector representation set of the pushee into the push parsing algorithm to obtain a targeted business list of the pushee, wherein the targeted business list of the pushee includes a ratio of the conversion coefficient of the pushee to each of the positive insurance businesses to the total conversion coefficient of the pushee to the multiple positive insurance businesses;

[0009] The insurance business vector representations of the several alternative insurance businesses and the targeted business list of the recommended person are input into the insurance business push algorithm, and a first inference result of the insurance business push algorithm based on the successful conversion of the several alternative insurance businesses to the recommended person is obtained, and the insurance business to be pushed for the recommended person is determined from the several alternative insurance businesses.

[0010] In an optional implementation manner, the inputting the pushee vector representation and the positive insurance business vector representation set of the pushee into the push parsing algorithm to obtain the targeted business list of the pushee includes:

[0011] Inputting the pushee vector representation and the positive insurance business vector representation set of the pushee into the push analysis algorithm to obtain the ratio of the conversion coefficient of the pushee to each of the positive insurance businesses and the total conversion coefficient of the pushee to the multiple positive insurance businesses;

[0012] Obtaining a list adjustment sequence of the positive insurance business vector representation set of the recommended person through the ratios corresponding to the plurality of positive insurance businesses respectively;

[0013] A targeted service list for the referee is established through the list adjustment sequence and the positive insurance service vector representation set of the referee.

[0014] In an optional implementation manner, the method further includes an optimization step of the push parsing algorithm, including:

[0015] Acquire a plurality of first recommendee samples' recommendee vector representation samples and positive insurance business vector representation set samples respectively corresponding to the plurality of first recommendee samples;

[0016] Inputting the recommended person vector representation samples and the positive insurance business vector representation set samples of the plurality of first recommended person samples into the original push parsing algorithm to obtain the targeted business list samples corresponding to the plurality of first recommended person samples respectively;

[0017] Obtaining an insurance business vector representation of a first insurance business sample that each first referee sample is interested in and a comment reflecting whether the first insurance business sample has been successfully converted to the first referee sample;

[0018] Inputting the insurance business vector representation of the first insurance business sample concerned by each first referee sample and the targeted business list sample of the first referee sample into the insurance business push algorithm to obtain a first estimated inference result of whether the first insurance business sample is successfully converted to the first referee sample;

[0019] The original push parsing algorithm is optimized through the first estimated inference results and annotations of each of the first insurance business samples to obtain the push parsing algorithm.

[0020] In an optional implementation manner, inputting the insurance service vector representations of the plurality of candidate insurance services and the targeted service list of the recommended recipient into an insurance service push algorithm to obtain the pushed insurance service of the recommended recipient determined from the plurality of candidate insurance services comprises:

[0021] Inputting the insurance business vector representations of the plurality of candidate insurance businesses and the targeted business list of the recommended person into the insurance business push algorithm to obtain a first inference result that each of the plurality of candidate insurance businesses has successfully converted the recommended person;

[0022] Determining first selected intermediate values ​​corresponding to the plurality of candidate insurance businesses respectively through first inference results corresponding to the plurality of candidate insurance businesses respectively;

[0023] The pushed insurance service of the pushed recipient is determined from among the several candidate insurance services by using the first selection medians respectively corresponding to the several candidate insurance services.

[0024] In an optional embodiment, after the insurance business vector representation set of the recommended person is obtained by obtaining the insurance business vector representations of the plurality of positive insurance businesses concerned by the recommended person, the method further includes:

[0025] Inputting the recommended person vector representation and the positive insurance business vector representation set of the recommended person into a significance determination algorithm to obtain a significance weighted list of the recommended person;

[0026] The step of inputting the insurance service vector representations of the plurality of candidate insurance services and the targeted service list of the recommended recipient into an insurance service push algorithm to obtain the pushed insurance service of the recommended recipient determined from among the plurality of candidate insurance services comprises:

[0027] The insurance service vector representations of the several candidate insurance services, the targeted service list and the significance weighted list of the recommended recipient are input into the insurance service push algorithm to obtain the pushed insurance service of the recommended recipient determined among the several candidate insurance services.

[0028] In an optional implementation manner, the inputting the recommendee vector representation and the positive insurance business vector representation set of the recommendee into a significance determination algorithm to obtain a significance weighted list of the recommendee includes:

[0029] Inputting the pushee vector representation and the positive insurance business vector representation set of the pushee into the significance determination algorithm, transforming the pushee vector representation of the pushee into a transformation vector of the same dimension as the insurance business vector representation of the positive insurance business;

[0030] Obtaining the spatial proximity between the insurance business vector representation of each positive insurance business in the positive insurance business vector representation set of the recommended person and the transformation vector, and obtaining the significance value of the positive insurance business vector representation set of the recommended person;

[0031] A significance weighted list of the recommended persons is established through the significance values ​​and the positive insurance business vector representation set of the recommended persons.

[0032] In an optional implementation, the method further includes an optimization step of the significance determination algorithm, including:

[0033] Acquire a plurality of recommended person vector representation samples of second recommended person samples and positive insurance business vector representation set samples respectively corresponding to the plurality of second recommended person samples;

[0034] Inputting the recommended person vector representation samples and the positive insurance business vector representation set samples of the plurality of second recommended person samples into the original significance determination algorithm to obtain significance weighted list samples corresponding to the plurality of second recommended person samples respectively;

[0035] Obtaining the insurance business vector representation of the second insurance business sample that each second referee sample is concerned about and a comment reflecting whether the second insurance business sample has been successfully converted to the second referee sample;

[0036] Inputting the insurance business vector representation of the second insurance business sample concerned by each second recommendee sample and the significance weighted list sample of the second recommendee sample into the insurance business push algorithm to obtain a second estimated inference result of whether the second insurance business sample is successfully converted to the second recommendee sample;

[0037] The original significance determination algorithm is optimized through the second estimated inference results and annotations of each of the second insurance business samples to obtain the significance determination algorithm.

[0038] In an optional implementation manner, the step of inputting the insurance business vector representation of the second insurance business sample that each second referee sample is concerned about and the significance weighted list sample of the second referee sample into the insurance business push algorithm to obtain a second estimated inference result of whether the second insurance business sample is successfully converted into the second referee sample includes:

[0039] Obtaining a targeted business list sample for each of the second recommended person samples;

[0040] The insurance business vector representation of the second insurance business sample that each second recommended person sample pays attention to, the significance weighted list sample and the targeted business list sample of the second recommended person sample are input into the insurance business push algorithm to obtain a second estimated inference result of whether the second insurance business sample is successfully converted to the second recommended person sample.

[0041] In an optional implementation manner, inputting the insurance business vector representations of the plurality of candidate insurance businesses, the targeted business list and the significance weighted list of the recommended recipient into the insurance business push algorithm to obtain the pushed insurance business of the recommended recipient determined from the plurality of candidate insurance businesses comprises:

[0042] Inputting the insurance business vector representations of the plurality of candidate insurance businesses, the targeted business list and the significance weighted list of the recommended person into the insurance business push algorithm to obtain a second inference result that the plurality of candidate insurance businesses have successfully converted the recommended person;

[0043] Determining second selected neutrals corresponding to the plurality of candidate insurance businesses respectively through second inference results corresponding to the plurality of candidate insurance businesses respectively;

[0044] The insurance service pushed to the recipient is determined from among the several candidate insurance services by using the second selection median corresponding to the several candidate insurance services respectively.

[0045] In a second aspect, an embodiment of the present disclosure provides an insurance business push system, including a client and a server that are communicatively connected to each other, the server including a processor and a memory, the memory storing computer instructions, and when the processor executes the computer instructions, the above method is implemented.

[0046] The present disclosure has at least the following beneficial effects: the insurance business push method and system based on big data analysis provided by the present disclosure extracts the conversion coefficient of each positive insurance business in the positive insurance business vector representation set from the pushee vector representation containing the insurance business evaluation behavior of the pushee, and the push analysis algorithm can establish a targeted business list for the pushee through the pushee vector representation and the positive insurance business vector representation set of the pushee, and the pushee's targeted business list includes the pushee's conversion coefficient for each positive insurance business and the ratio of the pushee's total conversion coefficient for multiple positive insurance businesses, and the pushee's targeted business list can characterize the distribution of the pushee's conversion coefficient for each positive insurance business in the positive insurance business vector representation set. Then the insurance business push algorithm selects insurance business for the pushee based on the pushee's targeted business list, and determines that the distribution of the insurance business and the pushee's conversion coefficient is more consistent, which can improve the conversion rate of the insurance business.

[0047] Other features will be described in part in the following description. Those skilled in the art will discover these features in part upon inspection of the following content and drawings, or may learn these features through production or use. The features of the present application may be implemented and obtained by practicing or using various aspects of the methods, tools, and combinations listed in the detailed examples described below. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used to illustrate the technical solutions of the present disclosure together with the specification.

[0049] Figure 1 It is a schematic diagram of an application scenario of the insurance business push method based on big data analysis provided in an embodiment of the present disclosure.

[0050] Figure 2 It is a flowchart of a method for pushing insurance business based on big data analysis provided in an embodiment of the present disclosure.

[0051] Figure 3 It is a schematic diagram of the functional module architecture of the insurance business push device provided in an embodiment of the present disclosure.

[0052] Figure 4 It is a schematic diagram of the composition of a server provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0053] In order to make the objectives, technical solutions and advantages of the present disclosure clearer, the present disclosure will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present disclosure. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0054] In the following description, references to “some embodiments”, “as an implementation / scheme”, and “in one implementation” describe a subset of all possible embodiments, but it can be understood that “some embodiments”, “as an implementation / scheme”, and “in one implementation” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0055] In the following description, the terms "first\second\third" and the like are only used to distinguish similar objects and do not represent a specific order for the objects. It is understandable that "first\second\third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present disclosure described here can be implemented in an order other than that illustrated or described here. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of the present disclosure. The terms used herein are only for the purpose of describing the embodiments of the present disclosure and are not intended to limit the present disclosure.

[0056] The insurance business push method based on big data analysis provided in the embodiments of the present disclosure can be executed by the server, which can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), as well as big data and artificial intelligence platforms.

[0057] Figure 1 Schematic diagram of the application scenario of the insurance business push method based on big data analysis provided by the embodiment of the present disclosure. The insurance business push system 10 provided by the embodiment of the present disclosure includes multiple clients 100, a network 200 and a server 300, and the multiple clients 100 and the server 300 are connected to each other through the network 200. The server 300 is used to execute the method provided by the embodiment of the present disclosure. Specifically, the embodiment of the present disclosure provides an insurance business push method based on big data analysis, and the method is applied to the server 300, such as Figure 2 As shown, the method includes:

[0058] Step 110: Obtain insurance business vector representations of several candidate insurance businesses and referee vector representations of referees.

[0059] Step 120: Obtain insurance business vector representations of multiple positive insurance businesses that the recommended person is interested in, and construct a positive insurance business vector representation set of the recommended person.

[0060] Step 130, input the pushee vector representation and the positive insurance business vector representation set of the pushee into the push analysis algorithm to obtain a targeted business list of the pushee, wherein the targeted business list of the pushee includes the ratio of the conversion coefficient of the pushee for each positive insurance business to the total conversion coefficient of the pushee for multiple positive insurance businesses.

[0061] Step 140, input the insurance business vector representations of the several candidate insurance businesses and the targeted business list of the recommended recipient into the insurance business push algorithm, obtain the first inference result of the insurance business push algorithm based on the successful conversion of the several candidate insurance businesses to the recommended recipient, and determine the recommended insurance business for the recommended recipient from the several candidate insurance businesses.

[0062] In the disclosed embodiment, the insurance business can be any type of insurance, such as property insurance and life insurance. Before the data is executed by the insurance business push algorithm, the insurance business, the pushee, etc. are vectorized to obtain the insurance business vector representation of the insurance business and the pushee vector representation of the pushee. The insurance business vector representation can be obtained based on various characteristics of the insurance business. It is a vector (vector). The number of elements of the vector, that is, the dimension, is related to the characteristics it has. For example, the insurance business vector representation includes characteristics such as the category, insured amount, loss ratio, and cycle of the insurance business. The alternative insurance business is the insurance business in the insurance business pool. The multiple positive insurance businesses that the pushee pays attention to are the positive insurance businesses that the pushee currently or historically pays attention to. Positive means that the pushee is interested in it or not disgusted with it, and it may be successfully converted (that is, purchased by the pushee). Paying attention means having a preset behavior, such as purchasing, consulting or browsing. The pushee's positive insurance business vector representation set is a plurality of positive insurance business vectors that the pushee pays attention to. A positive insurance business vector representation set includes the behavior of each insurance business being positively responded to by the pushee. For example, for 20 positive insurance businesses, a purchase list, a consultation list, and a browsing list are established. The consultation list has insurance business vector representations corresponding to the 20 positive insurance businesses, and the insurance business vector representations of the insurance businesses consulted among the 20 positive insurance businesses are marked as "Y", and the insurance business vector representations of the insurance businesses not consulted are marked as "N". According to a similar principle, each positive insurance business is marked in the purchase list and the browsing list.

[0063] The referee vector representation is obtained through the referee's referee description information and the referee's insurance business evaluation behavior, etc., which is a vector. The referee description information can be the user portrait of the referee, such as income, age, property allocation data, etc. The referee's insurance business evaluation behavior is the referee's browsing, consulting, and correlation comparison of insurance information on the insurance platform.

[0064] The pushee vector representation includes the pushee's information. Then, the pushee vector representation can be processed to obtain the ratio of the pushee's conversion coefficient for each positive insurance business in the positive insurance business vector representation set to the total conversion coefficient of the pushee for multiple positive insurance businesses in the positive insurance business vector representation set. It is difficult to extract the conversion coefficient (indicating the probability of successfully converting the positive insurance business) for each positive insurance business in the positive insurance business vector representation set from the pushee vector representation. The pushee's conversion coefficient for a positive insurance business in the positive insurance business vector representation set is the ratio of the conversion coefficient for the positive insurance business to the total conversion coefficient of all positive insurance businesses in the positive insurance business vector representation set. At different times or in different positive insurance business vector representation sets, the pushee's conversion coefficient for a certain insurance business may be different. For example, the pushee has a rebellious psychology towards the insurance business that appears repeatedly. At this time, when the pushee has seen insurance business of the same type as insurance business Q, the conversion coefficient of the pushee to insurance business Q is small, and when the pushee has not seen insurance business of the same type as insurance business Q, the conversion coefficient of insurance business Q is large. Then, the positive insurance business in the positive insurance business vector representation set needs to be arranged in order according to a certain rationality, and the conversion coefficient of each positive insurance business extracted through the referee vector representation also has a sequence, and the corresponding conversion coefficient can accurately show the distribution of the conversion coefficient of the referee.

[0065] The conversion coefficient of the pushee to the insurance business can be evaluated by the pushee's reaction to the insurance business. For example, the conversion coefficient of the pushee to the browsed insurance business is generally greater than that to the unbrowsed insurance business, and the conversion coefficient to the consulted insurance business is greater than that to the browsed insurance business. The pushee's positive insurance business vector representation set is weighted by comparing the conversion coefficient of each positive insurance business consulted by the pushee with the total conversion coefficient of multiple insurance businesses in the positive insurance business vector representation set to obtain the pushee's targeted business list. The pushee's targeted business list reflects the distribution of the pushee's conversion coefficient for each positive insurance business in the positive insurance business vector representation set. The larger the pushee's conversion coefficient, the greater the insurance business weight is assigned. When transforming the pushee vector representation into a list adjustment sequence, the embodiment of the present disclosure adopts an optimized push parsing algorithm, and extracts the ratio of the pushee's conversion coefficient for each positive insurance business to the total conversion coefficient of the pushee for multiple positive insurance businesses in the pushee vector representation through the push parsing algorithm. The list adjustment sequence is a sequence used to indicate the conversion coefficient of a pushee for each positive insurance business in the pushee's positive insurance business vector representation set. Each element in the list adjustment sequence corresponds to a sequence of precedence. The position of each element in the list adjustment sequence is consistent with the sequence of each positive insurance business in the positive insurance business vector representation set. If the pushee has one positive insurance business vector representation set, the list adjustment sequence corresponding to the positive insurance business vector representation set is extracted from the pushee's vector representation. The pushee's positive insurance business vector representation set is processed through the list adjustment sequence to obtain a targeted business list of the pushee (a list of insurance businesses with a sequence of precedence). Then, targeted insurance business is pushed to the pushee through the targeted business list. If the pushee has multiple positive insurance business vector representation sets, the list adjustment sequences extracted from the pushee's vector representation are the same number as the positive insurance business vector representation sets. For example, for the consultation list and browsing list containing the same positive insurance business, the list adjustment sequence of the consultation list and the list adjustment sequence of the browsing list are extracted from the vector representation of the recommended person, and then the elements in the list are weighted by the corresponding list adjustment sequences to obtain the consultation targeted business list and the browsing targeted business list respectively. After obtaining the above two targeted business lists, the consultation targeted business list and the browsing targeted business list are merged to obtain the recommended person's targeted business list, and insurance business is pushed through the targeted business list, or a selected degree is obtained through each targeted business list, and then insurance business is pushed through each selected degree. The selected degree is used to measure the recommendation degree of pushing the corresponding insurance business, which can be a numerical score.

[0066] In the process of pushing insurance business based on two targeted business lists, both the consultation targeted business list and the browsing targeted business list can be used as targeted business lists for insurance business push, and the insurance business vector representations of the consultation targeted business list, the browsing targeted business list and several alternative insurance businesses are input into the insurance business push algorithm. The consultation degree of each alternative insurance business is obtained according to the consultation targeted business list through the insurance business push algorithm, and the browsing degree of each alternative insurance business is obtained through the browsing targeted business list. The first selection center of each alternative insurance business is obtained based on the consultation degree and browsing degree of each alternative insurance business, and the pushed insurance business of the recipient is determined through the first selection center.

[0067] The original push parsing algorithm is optimized (i.e., the training or debugging process) through the pushee vector representation samples and the positive insurance business vector representation set samples of the first pushee sample, and an optimized push parsing algorithm is obtained. The push parsing algorithm learns how to extract the list adjustment sequence of the positive insurance business vector representation set samples from the pushee vector representation samples, and the list adjustment sequence is extracted from the pushee vector representation by the push parsing algorithm, and then the positive insurance business vector representation set of the pushee is weighted by the list adjustment sequence to obtain the targeted business list of the pushee. In the targeted business list of the pushee, the weights of the insurance business vector representations of each positive insurance business are different, so the targeted business list of the pushee can accurately reflect the distribution of the conversion coefficient of the pushee. When the insurance business push algorithm pushes insurance business to the pushee through the targeted business list of the pushee, the selected insurance business is more in line with the conversion coefficient of the pushee. The optimization method of the insurance business push algorithm is not limited, and can be executed based on a general neural network supervised training process, such as obtaining an estimated result, calculating the cost of the estimated result and the labeled result, performing gradient optimization based on the cost, and adjusting the parameters of the network until the network converges.

[0068] In the disclosed embodiment, the pushee vector representation corresponds to the insurance business evaluation behavior of the pushee, and the conversion coefficient of the pushee to each positive insurance business in the positive insurance business vector representation set can be extracted from the pushee vector representation. The push analysis algorithm can establish a targeted business list for the pushee through the pushee vector representation and the positive insurance business vector representation set of the pushee. The targeted business list of the pushee includes the ratio of the pushee's conversion coefficient for each positive insurance business to the pushee's total conversion coefficient for multiple positive insurance businesses. The targeted business list of the pushee can characterize the distribution of the pushee's conversion coefficient for each positive insurance business in the positive insurance business vector representation set. Then, the insurance business push algorithm pushes insurance business to the pushee based on the pushee's targeted business list, and the selected insurance business is more consistent with the pushee's conversion coefficient distribution.

[0069] In the following content, the training method of the push parsing algorithm will be described, which optionally includes: obtaining a plurality of first pushee samples of the pushee vector representation samples and a plurality of first pushee samples of the positive insurance business vector representation set samples; inputting the plurality of first pushee samples of the pushee vector representation samples and the positive insurance business vector representation set samples into the original push parsing algorithm to obtain a plurality of first pushee samples of the targeted business list samples. The original push parsing algorithm is the push parsing algorithm to be optimized, and the algorithm architecture is the same as the optimized push parsing algorithm. The method of obtaining the targeted business list samples of the first pushee samples through the original push parsing algorithm is the same as the method of obtaining the pushee's targeted business list through the push parsing algorithm.

[0070] In the disclosed embodiment, the original push parsing algorithm obtains a targeted business list sample of the first pushee sample, including: inputting the pushee vector representation sample and the positive insurance business vector representation set sample of the first pushee sample into the original push parsing algorithm, obtaining the ratio of the conversion coefficient of the first pushee sample to each positive insurance business in the positive insurance business vector representation set sample and the total conversion coefficient of the first pushee sample to each positive insurance business in the positive insurance business vector representation set sample, and obtaining the list adjustment sequence of the positive insurance business vector representation set sample of the first pushee sample through the ratios corresponding to multiple positive insurance businesses.

[0071] The disclosed embodiment can adjust the ratio by 2 times Sigmoid to make each ratio in the same numerical range. Because it is 2 times, it is limited between 0 and 2, thereby obtaining a list adjustment sequence of the positive insurance business vector characterization set samples of the first referee sample. Through the list adjustment sequence and the positive insurance business vector characterization set samples of the first referee sample, a targeted business list sample of the first referee sample is established. The number of elements in the list adjustment sequence is the same as the insurance business vector characterization in the positive insurance business vector characterization set samples. Each insurance business vector characterization is weighted by the elements in the list adjustment sequence corresponding to each insurance business vector characterization to obtain a targeted business list sample of the first referee sample.

[0072] The push parsing algorithm obtains the targeted business list of the pushee in the following manner: inputting the pushee vector representation and the positive insurance business vector representation set of the pushee into the push parsing algorithm, obtaining the ratio of the conversion coefficient of the pushee to each positive insurance business in the positive insurance business vector representation set and the total conversion coefficient of the multiple positive insurance businesses of the pushee, obtaining the list adjustment sequence of the positive insurance business vector representation set of the pushee through the ratios corresponding to the multiple positive insurance businesses, and establishing the targeted business list of the pushee through the list adjustment sequence and the positive insurance business vector representation set of the pushee. Further, adjusting the ratio through 2 times Sigmoid to obtain the list adjustment sequence of the positive insurance business vector representation set of the pushee. The obtained targeted business list of the pushee reflects the conversion coefficient of the pushee to each positive insurance business in the positive insurance business vector representation set, and further adjusting through the ratios corresponding to the multiple positive insurance businesses to normalize the obtained list adjustment sequence to help the subsequent task calculation.

[0073] When the push parsing algorithm is optimized based on the insurance business push algorithm, the push parsing algorithm is determined as a pre-link of the insurance business push algorithm. The insurance business push algorithm is used to estimate whether the insurance business is positively concerned by the recipients, and then the push parsing algorithm is optimized based on the actual attention paid by the recipients of the insurance business to obtain an optimized push parsing algorithm.

[0074] The targeted business list obtained by the push parsing algorithm cannot be optimized directly, but the insurance business push algorithm can be used to help the push parsing algorithm to be optimized. The insurance business push algorithm pushes based on the targeted business list output by the push parsing algorithm. The insurance business push algorithm can estimate whether the insurance business is positively concerned by the recipients, and optimize the push parsing algorithm based on the constraint that the insurance business is not actually positively concerned by the recipients.

[0075] With the help of the insurance business push algorithm, the original push parsing algorithm is optimized. It is also necessary to obtain the insurance business vector representations corresponding to multiple first insurance business samples that the first recommended recipient sample pays attention to, and at the same time, it is necessary to obtain the annotations corresponding to the multiple first insurance business samples, that is, the marks representing the actual results. The annotations of the first insurance business samples indicate whether the first insurance business samples pay positive attention to the corresponding first recommended recipient samples, such as consulting, browsing or purchasing.

[0076] For each first recommended person sample, the insurance business vector representation of the first insurance business sample that the first recommended person sample is concerned about and the targeted business list sample of the first recommended person sample are input into the insurance business push algorithm to obtain a first estimated inference result of whether the first insurance business sample is successfully converted to the first recommended person sample, and each behavior corresponds to an estimated inference result. A cost function is generated by the cost between the first estimated inference result and the annotation of each first insurance business sample, and the original push parsing algorithm is optimized according to the cost function, and the push parsing algorithm is obtained when the original push parsing algorithm converges. The inference result represents the probability that the recommended person will generate corresponding behavior (consultation, browsing or purchase, etc.) for the insurance business.

[0077] Exemplarily, when estimating the consultation behavior and browsing behavior, the insurance business vector representation of the first insurance business sample and the corresponding targeted business list of the first recommended person sample are input into the insurance business push algorithm, the targeted business list of the first recommended person sample is obtained through the original push analysis algorithm, and the insurance business push algorithm estimates the inferred result of the first insurance business sample being consulted by the first recommended person sample, and the inferred result of the first insurance business sample being browsed by the first recommended person sample. The consultation cost function is established by the cost between whether the first insurance business sample included in the annotation of the first insurance business sample is followed by the first recommended person sample, and the estimated inferred result of the first insurance business sample being consulted by the first recommended person sample. Similarly, a browsing cost function can also be established, and the algorithm configuration variables (i.e., algorithm parameters) of the original push analysis algorithm are adjusted according to the consultation cost function and the browsing cost function. The push parsing algorithm is obtained based on the optimization of several positive insurance business vector representation sets in the same way, that is, the push parsing algorithm is determined as a pre-part of the insurance business push algorithm, and the insurance business push algorithm is used to estimate whether the insurance business is performed in one or more of the multiple behaviors, and then the original push parsing algorithm is optimized by the cost of the insurance business being actually executed, so as to obtain an optimized push parsing algorithm. Among them, in the optimization link of the push parsing algorithm, the insurance business push algorithm is already optimized or optimized together with the push parsing algorithm, and the optimized insurance business push algorithm is used to help the push parsing algorithm to optimize, thereby improving the optimization speed of the push parsing algorithm, and when pushing insurance business, the optimized insurance business push algorithm is called. Based on this, since the targeted business list obtained by the push parsing algorithm is difficult to optimize directly, the insurance business push algorithm is used to help the push parsing algorithm to optimize, so that the push parsing algorithm can establish an accurate targeted business list of the recipients, and the obtained targeted business list is more in line with the requirements of the subsequent insurance business push algorithm.

[0078] When the insurance business push algorithm determines to push the insurance business, the insurance business push algorithm estimates the first estimated inference results of each behavior of the alternative insurance business for each alternative insurance business according to the input features. For example, for the alternative insurance business P, the estimated inference result of the alternative insurance business P being consulted by the pushee is 0.1, and the estimated inference result of the alternative insurance business P being browsed by the pushee is 0.2. For the alternative insurance business Q, the estimated inference result of the alternative insurance business Q being consulted by the pushee is 0.3, and the estimated inference result of the alternative insurance business Q being browsed by the pushee is 0.4. The first selected neutrality of the alternative insurance business can be obtained through the first estimated inference results corresponding to the behavior corresponding to each alternative insurance business, and the first estimated inference results corresponding to each behavior can be assigned different weights. For example, the weight of consultation is 2, and the weight of browsing is 1, then the first selected neutrality of the alternative insurance business P is 2×0.1+1×0.2=0.4, and the first selected neutrality of the alternative insurance business Q is 2×0.3+1×0.4=1.

[0079] The insurance business to be pushed to the recipient is determined from among the several alternative insurance businesses by using the first selected median values ​​corresponding to the several alternative insurance businesses. For example, the alternative insurance business with the largest first selected median value among the several alternative insurance businesses is determined as the insurance business to be pushed to the recipient. In this way, the insurance business is pushed to the recipient through the targeted business list of the recipient, and the selected insurance business is more in line with the conversion coefficient distribution of the recipient.

[0080] Based on the above implementation, the present disclosure can also establish a significance weighted list of the pushee, which is a list obtained by weighting the positive insurance business vector representation set of the pushee, and the corresponding weight reflects the attention degree of the pushee to each positive insurance business in the positive insurance business vector representation set. The significance weighted list of the pushee includes the attention of the pushee to each positive insurance business in the positive insurance business vector representation set. The significance weighted list of the pushee is the significance weighted list of the pushee obtained by the significance determination algorithm after the pushee vector representation and the positive insurance business vector representation set of the pushee are input into the significance determination algorithm, and the significance determination algorithm is obtained through supervised learning.

[0081] The present disclosure introduces pushing insurance services to a pushed recipient based on a targeted service list of the pushed recipient, inputting insurance service vector representations of several candidate insurance services and the targeted service list of the pushed recipient into an insurance service pushing algorithm, and obtaining the pushed insurance services of the pushed recipient determined among several candidate insurance services, and also pushing insurance services to the pushed recipient based on a significance weighted list of the pushed recipient, thereby inputting insurance service vector representations of several candidate insurance services and the significance weighted list of the pushed recipient into an insurance service pushing algorithm, and obtaining the pushed insurance services of the pushed recipient determined among several candidate insurance services, or pushing insurance services to the pushed recipient based on both the targeted service list and the significance weighted list of the pushed recipient, that is, inputting the insurance service vector representations of several candidate insurance services, the targeted service list and the significance weighted list of the pushed recipient into an insurance service pushing algorithm, and obtaining the pushed insurance services of the pushed recipient determined among several candidate insurance services. Because the significance weighted list of the recommended recipient includes the recommended recipient's attention to each positive insurance business in the positive insurance business vector representation set, pushing insurance business to the recommended recipient based on the significance weighted list, compared with pushing insurance business directly based on the recommended recipient's positive insurance business vector representation set, can make the determined pushed insurance business and the positive insurance business that the recommended recipient pays more attention to in the positive insurance business vector representation set have a strong correlation, and the push is more accurate.

[0082] For the establishment of the significance weighted list, the vector representation of the pushee includes the information of the pushee that can be obtained. The pushee vector representation of the pushee can be processed to obtain the acceptance degree of the pushee to each positive insurance business in the positive insurance business vector representation set. The acceptance degree of the pushee to each positive insurance business in the positive insurance business vector representation set is arranged according to the order of each positive insurance business in the positive insurance business vector representation set, and the significance value of the positive insurance business vector representation set (which is an attention parameter used to adjust the significance degree) can be obtained. The positive insurance business vector representation set of the pushee is weighted by the significance value of the positive insurance business vector representation set to obtain the significance weighted list of the pushee.

[0083] The significance weighted list of the recommendees is obtained by inputting the recommendee vector representation and the positive insurance business vector representation set of the recommendees into the significance determination algorithm. The significance determination algorithm obtains the recommendee vector representation sample of the second recommendee sample and the positive insurance business vector representation set sample by transforming the recommendee vector representation sample into a transformation vector of the same dimension as the insurance business vector representation of the positive insurance business. Optionally, the pushee vector representation and the positive insurance business vector representation set of the pushee are input into a significance determination algorithm, and the significance determination algorithm transforms the pushee vector representation of the pushee into a transformation vector of the pushee (a mapped vector) of the same dimension as the insurance business vector representation of the positive insurance business; obtains the spatial proximity (for example, vector distance, which can reflect the degree of similarity between the two, and the vector distance can be the calculated Euclidean distance) between the insurance business vector representation of each positive insurance business in the positive insurance business vector representation set and the transformation vector of the pushee, and determines the spatial proximity corresponding to each positive insurance business as the weight corresponding to the positive insurance business when establishing a significance weighted list through the positive insurance business vector representation set. The positive insurance business has an order in the positive insurance business vector representation set, and the weight of each positive insurance business also corresponds to the same order. The weight with the order is determined as the significance value of the positive insurance business vector representation set; the positive insurance business vector representation set of the pushee is weighted by the significance value to establish a significance weighted list of the pushee. In this way, the significance weighted list of the recommended recipients obtained includes the acceptance level of the recommended recipients to each positive insurance business in the positive insurance business vector representation set. When the insurance business is pushed through the significance weighted list of the recommended recipients, the selected insurance business is more similar to each insurance business in the positive insurance business vector representation set, and the selected insurance business is more consistent with the acceptance level of the recommended recipients.

[0084] When optimizing the significance determination algorithm, first obtain the recommendee vector characterization samples of multiple second recommendee samples and the positive insurance business vector characterization set samples of the multiple second recommendee samples, the second recommendee samples are the same as the first recommendee samples; the recommendee vector characterization samples and the positive insurance business vector characterization set samples of the multiple second recommendee samples are input into the original significance determination algorithm to obtain the significance weighted list samples of the multiple second recommendee samples. The original significance determination algorithm is the significance determination algorithm to be optimized, and includes the same algorithm composition as the optimized significance determination algorithm; the method of obtaining the significance weighted list samples of the second recommendee samples through the original significance determination algorithm is the same as the method of obtaining the significance weighted list of recommendees by the above-mentioned significance determination algorithm of the present disclosure.

[0085] With the help of the insurance business push algorithm, the optimization of the original significance determination algorithm also requires obtaining the insurance business vector representations corresponding to the multiple second insurance business samples that the second pushee sample pays attention to and the annotations corresponding to the multiple second insurance business samples. The annotations of the second insurance business samples indicate whether the second insurance business sample pays positive attention to each behavior of the corresponding second pushee sample. If the second pushee sample is the first pushee sample, then the first insurance business sample of the first pushee sample is used as the second insurance business sample of the second pushee sample. For each second pushee sample, the insurance business vector representation of the second insurance business sample that the second pushee sample pays attention to and the significance weighted list sample of the second pushee sample are input into the insurance business push algorithm to obtain a second estimated inference result of whether the second insurance business sample successfully converts the second pushee sample. For example, a second estimated inference result is included for each behavior. A cost function is established through the cost between the second estimated inference result of each second insurance business sample and the annotation. At the same time, the original significance determination algorithm is optimized through the cost function until convergence to obtain the significance determination algorithm.

[0086] In this way, because the significance weighted list obtained by the significance determination algorithm is difficult to optimize directly, the significance determination algorithm is optimized based on the insurance business push algorithm, so that the significance determination algorithm can establish an accurate significance weighted list of the recipients. The obtained significance weighted list is more in line with the needs of the subsequent insurance business push algorithm. Optionally, the significance determination algorithm is optimized when a targeted business list is input.

[0087] The insurance business vector representation of the second insurance business sample that each second recommendee sample pays attention to, as well as the significance weighted list sample and the targeted business list sample of the second recommendee sample, are simultaneously input into the insurance business push algorithm to obtain a second estimated inference result of whether the second insurance business sample is successfully converted into the second recommendee sample. The method of obtaining the targeted business list sample of the second recommendee sample is the same as that of obtaining the targeted business list sample of the first recommendee sample.

[0088] Optionally, the significance determination algorithm is optimized together with the push parsing algorithm, that is, the significance weighted list sample of the second pushee sample input to the insurance business push algorithm is obtained based on the original significance determination algorithm, and the targeted business list sample of the second pushee sample input to the insurance business push algorithm is obtained through the original push parsing algorithm. Optionally, the significance determination algorithm and the push parsing algorithm are optimized separately, that is, the significance weighted list sample of the second pushee sample input to the insurance business push algorithm is obtained through the original significance determination algorithm, and the targeted business list of the second pushee sample input to the insurance business push algorithm is obtained through the optimized push parsing algorithm, or the significance determination algorithm is optimized first, and then the push parsing algorithm is optimized, that is, the significance weighted list of the second pushee sample input to the insurance business push algorithm is obtained through the optimized significance determination algorithm, and the targeted business list sample of the second pushee sample input to the insurance business push algorithm is obtained through the original push parsing algorithm.

[0089] In this way, when the insurance business is pushed based on the targeted business list and the significance weighted list, the push accuracy is high. Compared with the case where the targeted business list is not input, the significance determination algorithm is optimized. When the targeted business list is input, the significance determination algorithm is optimized, which can make the insurance business push based on the targeted business list and the significance weighted list more accurate. At the same time, the insurance business push algorithm helps the significance determination algorithm to optimize, which can make the significance determination algorithm obtain the significance weighted list of the pushee more meet the insurance business push algorithm requirements. In the disclosed embodiment, the push parsing algorithm, the insurance business push algorithm and the significance determination algorithm are all neural network algorithms, such as deep neural networks, which can be convolutional neural network algorithms, recurrent neural network algorithms, residual network algorithms, etc. The basic architecture of the algorithm is used for optimization training, and the corresponding execution algorithm is obtained after adjusting the algorithm parameters. The basic composition architecture of the algorithm (such as the number of input layers, intermediate layers, and output layers, the specific composition of each layer, etc.) is not repeated.

[0090] When determining to push insurance business, the insurance business push algorithm may input insurance business vector representations of several candidate insurance businesses, a targeted business list of the pushee, and a significance weighted list into the insurance business push algorithm to obtain a second estimated inference result of the successful conversion of several candidate insurance businesses to the pushee. If the insurance business vector representations of several candidate insurance businesses and a targeted business list of the pushee are input into the insurance business push algorithm, refer to the above content. If the insurance business vector representations of several candidate insurance businesses and a significance weighted list of the pushee are input into the insurance business push algorithm, the acceptance degree of each positive insurance business in the positive insurance business vector representation set by the pushee is evaluated. If the insurance business vector representations of several candidate insurance businesses, a targeted business list of the pushee, and a significance weighted list are input into the insurance business push algorithm, the conversion coefficient and acceptance degree of each insurance business in the positive insurance business vector representation set by the pushee are evaluated, and the pushed insurance business obtained is more accurate.

[0091] The insurance business push algorithm estimates the second inference results of various behaviors of the alternative insurance business for each alternative insurance business. The second selected neutrality of the alternative insurance business is obtained through the second inference results corresponding to each behavior of the pushee corresponding to each alternative insurance business, and different weights can be assigned to the second inference results corresponding to different behaviors. The push insurance business of the pushee is determined among several alternative insurance businesses through the second selected neutrality corresponding to several alternative insurance businesses. Optionally, the alternative insurance business with the highest second selected neutrality among several alternative insurance businesses can be determined as the push insurance business of the pushee, or the first N alternative insurance businesses in descending order of the second selected neutrality among several alternative insurance businesses can be determined as the push insurance business.

[0092] The second inference result is obtained when the insurance business push algorithm is input with the insurance business vector representations of several alternative insurance businesses, and the targeted business list and significance weighted list of the pushee. It can also be obtained when the insurance business push algorithm is input with the insurance business vector representations of several alternative insurance businesses, and one of the targeted business list or significance weighted list of the pushee. In this way, the insurance business is pushed to the pushee based on the targeted business list and significance weighted list of the pushee. The selected insurance business not only conforms to the conversion coefficient distribution of the pushee, but also conforms to the positive insurance business in the positive insurance business vector representation concentration that the pushee is concerned about, and the accuracy is improved.

[0093] Based on the above embodiments, the present disclosure provides an insurance business push device. Figure 3 An insurance business push device 340 provided by the embodiment of the present disclosure is as follows: Figure 3 As shown, the device 340 includes:

[0094] A vector representation acquisition module 341 is used to acquire insurance business vector representations of several candidate insurance businesses and a pushee vector representation of a pushee; wherein the pushee vector representation of the pushee is obtained through the pushee description information of the pushee and the insurance business evaluation behavior of the pushee;

[0095] A positive feature acquisition module 342, configured to acquire a positive insurance business vector representation set of the recommended person by acquiring insurance business vector representations of the multiple positive insurance businesses concerned by the recommended person;

[0096] A service list acquisition module 343 is used to input the pushee vector representation and the positive insurance service vector representation set of the pushee into the push parsing algorithm to obtain a targeted service list of the pushee, wherein the targeted service list of the pushee includes a ratio of the conversion coefficient of the pushee to each of the positive insurance services to the total conversion coefficient of the pushee to the multiple positive insurance services;

[0097] The push service selection module 344 is used to input the insurance service vector representations of the several alternative insurance services and the targeted service list of the pushee into the insurance service push algorithm, obtain the first inference result of the insurance service push algorithm based on the successful conversion of the several alternative insurance services to the pushee, and determine the push insurance service for the pushee among the several alternative insurance services.

[0098] The description of the above device embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the device embodiment of the present disclosure, please refer to the description of the method embodiment of the present disclosure for understanding.

[0099] If the disclosed technical solution involves personal or private information, the product using the disclosed technical solution has clearly informed the personal information processing rules and obtained the individual's voluntary consent before processing the personal information. If the disclosed technical solution involves sensitive personal information, the product using the disclosed technical solution has obtained the individual's separate consent before processing the sensitive personal information, and at the same time meets the requirements of "explicit consent", and collects within the scope of laws and regulations. For example, on personal information collection devices such as cameras, set clear and obvious signs to inform that the personal information collection scope has been entered and personal information will be collected. If the individual voluntarily enters the collection scope, it is deemed to agree to the collection of his or her personal information; or on the device for personal information processing, when the personal information processing rules are notified by obvious signs / information, the individual's authorization is obtained through pop-up information or by asking the individual to upload his or her personal information; among them, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the type of personal information processed.

[0100] It should be noted that in the embodiments of the present disclosure, if the above method is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present disclosure is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions to enable an electronic device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk. In this way, the embodiments of the present disclosure are not limited to any specific combination of hardware and software.

[0101] An embodiment of the present disclosure provides an electronic device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor implements the above method when executing the computer program.

[0102] The embodiment of the present disclosure provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. The computer-readable storage medium may be transient or non-transient.

[0103] The present disclosure provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and when the computer program is read and executed by a computer, some or all of the steps in the above method are implemented. The computer program product can be implemented specifically by hardware, software or a combination thereof. In an optional embodiment, the computer program product is specifically embodied as a computer storage medium, and in another optional embodiment, the computer program product is specifically embodied as a software product, such as a software development kit (Software Development Kit, SDK) and the like.

[0104] It should be noted that Figure 4 is a schematic diagram of a hardware entity of a server 300 provided in an embodiment of the present disclosure, such as Figure 4As shown, the hardware entity of the server 300 includes: a processor 310, a communication interface 320 and a memory 330, wherein: the processor 310 generally controls the overall operation of the server 300. The communication interface 320 enables the electronic device to communicate with other terminals or servers through a network. The memory 330 is configured to store instructions and applications executable by the processor 310, and can also cache data to be processed or processed by the processor 310 and each module in the server 300 (for example, image data, audio data, voice communication data and video communication data), which can be implemented by flash memory (FLASH) or random access memory (Random Access Memory, RAM). Data can be transmitted between the processor 310, the communication interface 320 and the memory 330 through the bus 340. It should be pointed out here that the description of the above storage medium and device embodiments is similar to the description of the above method embodiments, and has similar beneficial effects as the method embodiments. For technical details not disclosed in the storage medium and device embodiments of the present disclosure, please refer to the description of the method embodiments of the present disclosure for understanding.

[0105] It should be understood that "one embodiment" or "an embodiment" mentioned throughout the specification means that specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present disclosure. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in the various embodiments of the present disclosure, the size of the serial number of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present disclosure. The serial numbers of the embodiments of the present disclosure are for description only and do not represent the advantages and disadvantages of the embodiments.

[0106] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0107] In the several embodiments provided in the present disclosure, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0108] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0109] In addition, all functional units in the embodiments of the present disclosure may be integrated into one processing unit, or each unit may be separately configured as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0110] A person skilled in the art can understand that all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM), magnetic disks or optical disks, etc., various media that can store program codes.

[0111] Alternatively, if the above-mentioned integrated unit of the present disclosure is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present disclosure can essentially or in other words, the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for an electronic device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0112] The above description is only an implementation mode of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present disclosure, which should be included in the protection scope of the present disclosure.

Claims

1. A method for pushing insurance business based on big data analysis, characterized in that: Applied to the server, the method includes: Obtaining insurance business vector representations of several candidate insurance businesses and a recommended person vector representation of a recommended person; wherein the recommended person vector representation of the recommended person is obtained through recommended description information of the recommended person and insurance business evaluation behavior of the recommended person; The insurance business vector representation set of the recommended person is obtained by obtaining the insurance business vector representations of the multiple positive insurance businesses that the recommended person is concerned about, and forming the insurance business vector representation set of the multiple positive insurance businesses; Inputting the pushee vector representation and the positive insurance business vector representation set of the pushee into the significance determination algorithm, transforming the pushee vector representation of the pushee into a transformation vector of the same dimension as the insurance business vector representation of the positive insurance business; obtaining the spatial proximity between the insurance business vector representation of each positive insurance business in the positive insurance business vector representation set of the pushee and the transformation vector, and obtaining the significance value of the positive insurance business vector representation set of the pushee; establishing a significance weighted list of the pushee through the significance value and the positive insurance business vector representation set of the pushee; Input the pushee vector representation and the positive insurance business vector representation set of the pushee into the push analysis algorithm to obtain the ratio of the conversion coefficient of the pushee to each of the positive insurance businesses to the total conversion coefficient of the pushee to the multiple positive insurance businesses; obtain the list adjustment sequence of the positive insurance business vector representation set of the pushee through the ratios corresponding to the multiple positive insurance businesses; establish the targeted business list of the pushee through the list adjustment sequence and the positive insurance business vector representation set of the pushee, the targeted business list of the pushee including the ratio of the conversion coefficient of the pushee to each of the positive insurance businesses to the total conversion coefficient of the pushee to the multiple positive insurance businesses; Input the insurance business vector representations of the several alternative insurance businesses, the targeted business list and the significance weighted list of the promoted person into the insurance business push algorithm to obtain the pushed insurance business for the promoted person determined among the several alternative insurance businesses; or; input the insurance business vector representations of the several alternative insurance businesses and the targeted business list of the promoted person into the insurance business push algorithm to obtain the first inferred results that each of the several alternative insurance businesses has successfully converted the promoted person; determine the first selected neutrals corresponding to the several alternative insurance businesses respectively through the first inferred results corresponding to the several alternative insurance businesses; determine the pushed insurance business for the promoted person among the several alternative insurance businesses through the first selected neutrals corresponding to the several alternative insurance businesses.

2. The method according to claim 1, characterized in that The method also includes an optimization step of the push parsing algorithm, including: Obtaining a plurality of first recommendee samples' recommendee vector representation samples and a plurality of first recommendee samples' respective positive insurance business vector representation set samples; Inputting the recommended person vector representation samples and the positive insurance business vector representation set samples of the plurality of first recommended person samples into the original push parsing algorithm to obtain the targeted business list samples of the plurality of first recommended person samples respectively; Obtaining an insurance business vector representation of a first insurance business sample that each first referee sample is interested in and a comment reflecting whether the first insurance business sample has been successfully converted to the first referee sample; Inputting the insurance business vector representation of the first insurance business sample concerned by each first referee sample and the targeted business list sample of the first referee sample into the insurance business push algorithm to obtain a first estimated inference result of whether the first insurance business sample is successfully converted to the first referee sample; The original push parsing algorithm is optimized through the first estimated inference results and annotations of each of the first insurance business samples to obtain the push parsing algorithm.

3. The method according to claim 1, characterized in that The method also includes an optimization step of the saliency determination algorithm, including: Obtaining a plurality of recommendee vector representation samples of second recommendee samples and positive insurance business vector representation set samples of the plurality of second recommendee samples respectively; Inputting the referee vector representation samples and the positive insurance business vector representation set samples of the plurality of second referee samples into the original significance determination algorithm to obtain significance weighted list samples of the plurality of second referee samples respectively; Obtaining the insurance business vector representation of the second insurance business sample that each second referee sample is concerned about and a comment reflecting whether the second insurance business sample has been successfully converted to the second referee sample; Inputting the insurance business vector representation of the second insurance business sample concerned by each second recommendee sample and the significance weighted list sample of the second recommendee sample into the insurance business push algorithm to obtain a second estimated inference result of whether the second insurance business sample is successfully converted to the second recommendee sample; The original significance determination algorithm is optimized through the second estimated inference results and annotations of each of the second insurance business samples to obtain the significance determination algorithm.

4. The method according to claim 3, characterized in that The step of inputting the insurance business vector representation of the second insurance business sample concerned by each second recommendee sample and the significance weighted list sample of the second recommendee sample into the insurance business push algorithm to obtain a second estimated inference result of whether the second insurance business sample is successfully converted to the second recommendee sample includes: Obtaining a targeted business list sample for each of the second recommended person samples; The insurance business vector representation of the second insurance business sample that each second recommended person sample pays attention to, the significance weighted list sample and the targeted business list sample of the second recommended person sample are input into the insurance business push algorithm to obtain a second estimated inference result of whether the second insurance business sample is successfully converted to the second recommended person sample.

5. The method according to claim 1, characterized in that: The step of inputting the insurance service vector representations of the plurality of candidate insurance services, the targeted service list and the significance weighted list of the recommended recipient into the insurance service push algorithm to obtain the pushed insurance service of the recommended recipient determined from the plurality of candidate insurance services comprises: Inputting the insurance business vector representations of the plurality of candidate insurance businesses, the targeted business list and the significance weighted list of the recommended person into the insurance business push algorithm to obtain a second inference result that the plurality of candidate insurance businesses have successfully converted the recommended person; Determining second selected neutrals corresponding to the plurality of candidate insurance businesses respectively through second inference results corresponding to the plurality of candidate insurance businesses respectively; The insurance service pushed to the recipient is determined from among the several candidate insurance services by using the second selection median corresponding to the several candidate insurance services respectively.

6. An insurance business push system, characterized in that: The method comprises a client and a server connected to each other for communication, wherein the server comprises a processor and a memory, wherein the memory stores computer instructions, and when the processor executes the computer instructions, the method according to any one of claims 1 to 5 is implemented.

Images