CN117593039A - Method for predicting yield, method for training model, device, equipment and medium - Google Patents
Method for predicting yield, method for training model, device, equipment and medium Download PDFInfo
- Publication number
- CN117593039A CN117593039A CN202311632800.0A CN202311632800A CN117593039A CN 117593039 A CN117593039 A CN 117593039A CN 202311632800 A CN202311632800 A CN 202311632800A CN 117593039 A CN117593039 A CN 117593039A
- Authority
- CN
- China
- Prior art keywords
- product
- predicted
- sample
- model
- product channel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 108
- 238000012549 training Methods 0.000 title claims abstract description 77
- 238000006243 chemical reaction Methods 0.000 claims abstract description 55
- 238000004364 calculation method Methods 0.000 claims abstract description 34
- 230000015654 memory Effects 0.000 claims description 19
- 238000004590 computer program Methods 0.000 claims description 14
- 238000010276 construction Methods 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 abstract description 13
- 238000013461 design Methods 0.000 description 16
- 238000004458 analytical method Methods 0.000 description 13
- 230000008569 process Effects 0.000 description 9
- 230000006870 function Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 7
- 238000013135 deep learning Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000011161 development Methods 0.000 description 4
- 230000014759 maintenance of location Effects 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 230000007423 decrease Effects 0.000 description 3
- 230000003247 decreasing effect Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000013468 resource allocation Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000010219 correlation analysis Methods 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000002194 synthesizing effect Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 230000001747 exhibiting effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000012384 transportation and delivery Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0203—Market surveys; Market polls
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Data Mining & Analysis (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Game Theory and Decision Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application provides a prediction method, a model training method, a device, equipment and a medium for actual yield, and relates to the technical field of data processing, wherein the method comprises the following steps: and inputting the multi-factor time sequence of the product channel to be predicted into a real yield prediction model by acquiring the multi-factor time sequence of the product channel to be predicted, and acquiring the real yield of the product channel to be predicted output by the real yield prediction model. The multi-factor time sequence comprises a product name feature, a purchase times feature, a product channel feature and a policy issuing date feature, and the actual yield prediction model is a model obtained by model training based on full-chain insurance customer conversion link data and payment calculation related data of each sample product. In the technical scheme, the accuracy of the calculated actual yield is effectively improved.
Description
Technical Field
The present disclosure relates to the field of data processing technologies, and in particular, to a method for predicting yield, a method for model training, a device, equipment, and a medium.
Background
In the full-chain user value project, the actual yield prediction is performed on products of different business channels, and the prediction of the sales and income conditions of the products of each business channel (such as agent channels, direct sales channels and the like) can be evaluated, so that the company is helped to plan and manage the sales activities of each business channel more accurately, and the establishment of corresponding sales targets and resource allocation strategies is very important.
Currently, the actual yield is calculated by calculating the premium continuation rate by a chain step method or a differential method, and calculating the actual yield from the premium continuation rate. However, the accuracy of the actual yield calculated by the prior art is low.
Disclosure of Invention
The application provides a prediction method, a model training method, a device, equipment and a medium for actual yield, which are used for solving the problem of lower accuracy of actual yield calculation in the prior art.
In a first aspect, an embodiment of the present application provides a method for predicting a real yield, including:
acquiring a multi-factor time sequence of a product channel to be predicted, wherein the multi-factor time sequence comprises a product name feature, a purchase frequency feature, a product channel feature and a policy issuing date feature;
inputting the multi-factor time sequence of the product channel to be predicted into a yield prediction model, and obtaining the yield of the product channel to be predicted output by the yield prediction model, wherein the yield prediction model is a model obtained by model training based on full-chain insuring customer conversion link data and payment calculation related data of each sample product.
In one possible design, the inputting the multi-factor time series of the product channel to be predicted into a yield prediction model to obtain the yield of the product channel to be predicted output by the yield prediction model includes:
Predicting the continuation rate of the product channel to be predicted according to the multi-factor time sequence of the product channel to be predicted;
and predicting the actual yield of the product channel to be predicted according to the external factor weight and the continuation rate.
In one possible design, the method further comprises:
and if the difference of the actual yield of the product channel to be predicted subtracted from the actual yield of the product channel to be predicted in the last period is larger than a preset difference, early warning is carried out on the product channel to be predicted.
In one possible design, the method further comprises:
determining target actual yield of each policy issuing date in a preset time period in the product channel to be predicted;
and displaying the target actual yield of the product channel.
In a second aspect, an embodiment of the present application provides a model training method, including:
according to the full-chain insurance customer conversion link data of each sample product and payment calculation related data of each sample product, constructing a sample multi-factor time sequence of each product channel, wherein the sample multi-factor time sequence comprises a purchasing frequency characteristic, a product name characteristic, a product channel characteristic and a policy issuing date characteristic;
model training is carried out according to the sample multi-factor time sequence to obtain a first model, and the first model is used for calculating the continuation rate of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted;
Model training is carried out according to the external factors and the continuation rates corresponding to the sample multi-factor time sequence, so that a second model is obtained, and the second model contains the weights of the external factors;
correcting the first model according to the weight of the external factor to obtain a yield prediction model, wherein the yield prediction model is used for calculating the yield of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted.
In one possible design, the constructing a sample multi-factor time sequence of each product channel according to the full-chain insuring client conversion link data of each sample product and the payment calculation related data of each sample product includes:
constructing customer conversion detail data according to full-chain insurance customer conversion link data of each sample product, wherein the customer conversion detail data is used for representing the first purchased sample product and the second purchased sample product of each customer;
classifying sample products in the customer conversion detail data according to the service requirements and product channels to determine sample products contained in each product channel;
determining the purchase times characteristic of each sample product corresponding to each transaction from the customer conversion detail data;
And calculating related data, the purchase times characteristic of each sample product corresponding to each transaction and the product channel characteristic of each sample product according to the payment calculation of each transaction, and constructing a sample multi-factor time sequence of each product channel.
In a third aspect, an embodiment of the present application provides a device for predicting a recovery rate, including:
the acquisition module is used for acquiring a multi-factor time sequence of a product channel to be predicted, wherein the multi-factor time sequence comprises a product name feature, a purchase frequency feature, a product channel feature and a policy issuing date feature;
the input module is used for inputting the multi-factor time sequence of the product channel to be predicted into a yield prediction model, obtaining the yield of the product channel to be predicted output by the yield prediction model, wherein the yield prediction model is a model obtained by model training based on full-chain insurance customer conversion link data and payment calculation related data of each sample product.
In one possible design, the yield prediction model is specifically used for:
predicting the continuation rate of the product channel to be predicted according to the multi-factor time sequence of the product channel to be predicted;
and predicting the actual yield of the product channel to be predicted according to the external factor weight and the continuation rate.
In one possible design, the device for predicting the actual yield further includes an early warning module, configured to:
and if the difference of the actual yield of the product channel to be predicted subtracted from the actual yield of the product channel to be predicted in the last period is larger than a preset difference, early warning is carried out on the product channel to be predicted.
In one possible design, the device for predicting the yield further includes a display module for:
determining target actual yield of each policy issuing date in a preset time period in the product channel to be predicted;
and displaying the target actual yield of the product channel.
In a fourth aspect, an embodiment of the present application provides a model training apparatus, including:
the construction module is used for constructing a sample multi-factor time sequence of each product channel according to full-chain insurance customer conversion link data of each sample product and payment calculation related data of each sample product, wherein the sample multi-factor time sequence comprises a purchase time feature, a product name feature, a product channel feature and an insurance policy issuing date feature;
the training module is used for carrying out model training according to the sample multi-factor time sequence to obtain a first model, and the first model is used for calculating the continuation rate of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted;
The training module is further used for performing model training according to the external factors and the continuation rates corresponding to the sample multi-factor time sequence to obtain a second model, wherein the second model comprises the weights of the external factors;
the correction module is used for correcting the first model according to the weight of the external factor to obtain a yield prediction model, and the yield prediction model is used for calculating the yield of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted.
In one possible design, the building block is specifically configured to:
constructing customer conversion detail data according to full-chain insurance customer conversion link data of each sample product, wherein the customer conversion detail data is used for representing the first purchased sample product and the second purchased sample product of each customer;
classifying sample products in the customer conversion detail data according to the service requirements and product channels to determine sample products contained in each product channel;
determining the purchase times characteristic of each sample product corresponding to each transaction from the customer conversion detail data;
and calculating related data, the purchase times characteristic of each sample product corresponding to each transaction and the product channel characteristic of each sample product according to the payment calculation of each transaction, and constructing a sample multi-factor time sequence of each product channel.
In a fifth aspect, embodiments of the present application provide an electronic device, including: a processor, a memory and computer program instructions stored on the memory and executable on the processor for implementing the method provided by the first aspect and each possible design when the processor executes the computer program instructions.
In a sixth aspect, embodiments of the present application may provide a computer-readable storage medium having stored therein computer-executable instructions which, when executed by a processor, are configured to implement the method provided by the first aspect and each possible design.
In a seventh aspect, embodiments of the present application provide a computer program product comprising a computer program for implementing the method provided by the first aspect and each possible design when executed by a processor.
According to the actual yield prediction method, the model training method, the device, the equipment and the medium, in the method, the multi-factor time sequence of the product channel to be predicted is input into the actual yield prediction model by acquiring the multi-factor time sequence of the product channel to be predicted, and the actual yield of the product channel to be predicted, which is output by the actual yield prediction model, is acquired. The multi-factor time sequence comprises a product name feature, a purchase times feature, a product channel feature and a policy issuing date feature, and the actual yield prediction model is a model obtained by model training based on full-chain insurance customer conversion link data and payment calculation related data of each sample product. In the technical scheme, the actual yield prediction model comprises the weight of the external factor, the weight of the external factor is obtained through model training, and the accuracy is high, so that the accuracy of the actual yield calculated based on the weight of the external factor is effectively improved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application.
FIG. 1 is a schematic flow chart of a first embodiment of a method for predicting the actual yield according to the embodiment of the present application;
FIG. 2 is a schematic flow chart of a first embodiment of a model training method according to the present application;
fig. 3 is a schematic flow chart of a second embodiment of a model training method provided in the embodiment of the present application;
FIG. 4 is a schematic structural diagram of a device for predicting the actual yield according to an embodiment of the present disclosure;
fig. 5 is a schematic structural diagram of a model training device according to an embodiment of the present application;
fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Specific embodiments thereof have been shown by way of example in the drawings and will herein be described in more detail. These drawings and the written description are not intended to limit the scope of the inventive concepts in any way, but to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments.
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
Before describing the embodiments of the present application, an application background of the embodiments of the present application will be explained first:
full chain user value items refer to improving sales and retention of insurance products by analyzing customer's lifetime value and purchasing behavior to gain more insight into customers and formulating corresponding marketing strategies. The actual yield prediction of different service channels of the product in the full-chain user value project means the prediction of the sales of insurance products and the evaluation of income conditions of each service channel (such as agent channels, direct sales channels and the like), so that the sales strategy is optimized according to the actual yield obtained by prediction, the design accuracy of the product is improved, and the management and control risks are improved, thereby improving the competitiveness and profitability of enterprises.
Specifically, after predicting the actual yield of different service channels of the product, the insurance company can:
channel resource allocation optimization: by exhibiting a continuous prediction of the actual yield of products from different time intervals and different channels, and comparing with the actual yield and the expected yield, the insurer can learn which business channels perform better or worse in terms of sales and renewal. Based on the data, the company can reasonably adjust the resource allocation and put more resources into a channel with good performance, thereby improving the overall yield.
Early warning function: the method can realize the monitoring of the achievement of the actual yield and provide an early warning function. If the yield of a certain business channel is low or abnormal conditions occur, the system can send out early warning signals in time. Thus, the manager and related departments can quickly take action, investigate and solve the problem, and ensure the achievement of the objective of the actual yield.
Decision analysis support: according to the predicted actual yield and the monitoring data, different business departments can obtain actual yield early warning reports. These reports may provide information to the decision maker regarding current yield conditions and support subsequent factor decision analysis. For example, an insurance company may decide whether to adjust rewards policies, sales policies or service levels of a channel based on the actual yield predictions and monitoring results to optimize business performance and increase actual yield.
Performance assessment and incentive: the performance evaluation method can be combined with performance evaluation and incentive of insurance companies, performance of different channels is evaluated through yield prediction and monitoring data, and incentive measures are provided for business personnel. This may motivate sales teams to improve performance and to agree with corporate targets.
From the above, it is important for insurance companies to have important strategic and decision significance for predicting the actual yield of different service channels of products.
The traditional method for calculating the actual yield is to calculate the premium continuation rate by a chain ladder method or a difference method so as to calculate the actual yield. Next, the calculation of the actual yield of the product by the chain step method and the differential method will be explained
Chain ladder Method (English: chain-ladder Method):
historical policy data is collected, including year of sale and corresponding premium revenue or policy numbers. The data are grouped according to the year and date of sale to form a data chain. The data at each time point represents the premium revenue or policy number from the beginning of the year of sale to that time point. The chain ladder method is used for estimating the development factor of each time point, wherein the development factor refers to the growth rate from one time point to the next time point, and reflects the growth or descending trend of the business. And gradually calculating the premium continuation rate of the future time point according to the development factor, wherein the premium continuation rate represents the probability that the policy at a certain time point continues to pay the premium after the policy is paid. And synthesizing the premium continuation rate at each time point to obtain the overall real yield.
Difference mode (english: difference Method):
historical policy data is collected, including the year of the policy and the corresponding premium revenue or policy number. The data is grouped by year of policy and the total premium revenue or policy count for each year is calculated. The increment of each year is obtained by calculating the difference between two adjacent years, and the difference represents the new application and the renewal between different years. And calculating the rate of the continuous premium in the future year according to the proportional relation of the difference value, wherein the rate of the continuous premium represents the probability that the policy continues to pay the premium after the policy is purchased in the specific year. And synthesizing the premium continuation rate of each year to obtain the overall real yield.
However, both the above two methods are based on analysis of historical data to calculate the premium continuation rate and the real yield, and the parameters need to be manually introduced to be calculated, and the values of the parameters are determined by experience or professional judgment, so that the service requirement capability of the personnel to be calculated is high, and the accuracy is low.
Based on this, the inventors found that, when the related method for calculating the actual yield is studied, since the manually introduced parameters directly participate in the calculation of the actual yield, the accuracy of the calculated actual yield is low when the service level of the computerist is low or experience is insufficient. Based on the technical problems, the application provides a prediction method and a model training method of actual yield, which are characterized in that a first model for calculating the product continuation rate is trained, then external factors are introduced by referring to a traditional fine calculation chain ladder method, model training is carried out according to the external factors and the product continuation rate to obtain a second model, and the first model is corrected according to the weight of the external factors of the second model, so that the actual yield prediction model is obtained. Therefore, in practical application, the subsequent continuous rate curve can be predicted according to the previous n-period continuous rate of the product, and the actual yield can be calculated according to the weight of the external factors in the actual yield prediction model. Because the weight of the external factor is obtained through model training in advance, the method replaces a manual determination mode in the prior art, and the accuracy of the actual yield of the product obtained through calculation according to the weight of the external factor is effectively improved.
The following describes the technical scheme of the present application in detail through specific embodiments.
It should be noted that the following embodiments may be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.
Fig. 1 is a schematic flow chart of a first embodiment of a method for predicting the actual yield according to the embodiment of the present application.
As shown in fig. 1, the method for predicting the actual yield may include the steps of:
s11, acquiring a multi-factor time sequence of a product channel to be predicted.
The execution main body of the embodiment of the application is an electronic device, and the electronic device may be a terminal device, for example, a mobile phone, a notebook computer, a desktop computer, etc., and may also be a server. In practical application, the electronic device is specifically a terminal device or a server, and may be determined according to practical situations, which is not specifically limited in the embodiments of the present application.
The multi-factor time sequence comprises a product name feature, a purchase frequency feature, a product channel feature and a policy issuing date feature.
Optionally, the date of issuing the policy is the month of the underwriting year.
Optionally, the feature value of the purchase number feature is first-time insurance or repurchase.
It should be appreciated that the multi-factor time series may also include other features such as product type features, installment features, continuation rate features, and the like. The embodiments of the present application do not specifically limit the features contained in the multi-factor time series.
In practical application, the multi-factor time sequence of the product channel to be predicted can be obtained from the data of the multi-factor time sequence of the product channel to be predicted, and the full-chain application client conversion link data and payment plan related data of each product in the product channel to be predicted can be obtained from the database of the full-chain application client conversion link data and payment plan related data of each product in the product channel to be predicted, so that the multi-factor time sequence of the product channel to be predicted is constructed according to the full-chain application client conversion link data and payment plan related data of each product in the product channel to be predicted.
According to the conversion link data and the payment plan related data of all-chain insurance customers of each product in the product channel to be predicted, the construction of the multi-factor time sequence of the product channel to be predicted can be realized through the following processes:
for each product in the product channel to be predicted, determining whether each user is first insuring or repurchase according to the conversion link data of the full-chain insuring clients, namely determining the purchasing frequency characteristics of the product. And then, determining the date characteristic of the policy issuing of each product according to the related data of the payment plan of each product. Further, according to the product names of the products and the product channels to be predicted, a multi-factor time sequence of the product channels to be predicted is constructed.
Optionally, the product related to the embodiment of the application may be a month payment product.
S12, inputting the multi-factor time sequence of the product channel to be predicted into a yield prediction model, and obtaining the yield of the product channel to be predicted output by the yield prediction model.
The actual yield prediction model is a model obtained by model training based on full-chain insurance customer conversion link data and payment calculation related data of each sample product, and the actual yield prediction model comprises weights of external factors, wherein the weights of the external factors are obtained through model training.
In practical application, the actual yield prediction model can be obtained in advance from a database storing the actual yield prediction model, and model training can be performed in advance based on full-chain insurance customer conversion link data and payment calculation related data of each sample product, so that the actual yield prediction model is obtained. It should be understood that the embodiment of the present application is not limited to a specific manner of obtaining the actual yield prediction model, and may be determined according to actual situations.
It should be appreciated that the model training process may be explained with reference to the embodiment shown in fig. 2, and will not be described here again.
In one possible implementation, after inputting the multi-factor time sequence of the product channel to be predicted into the actual yield prediction model, the actual yield prediction model predicts the continuation rate of the product channel to be predicted according to the multi-factor time sequence of the product channel to be predicted. And then, the actual yield prediction model predicts the actual yield of the product channel to be predicted according to the external factor weight and the continuation rate.
According to the actual yield prediction method provided by the embodiment of the application, the multi-factor time sequence of the product channel to be predicted is obtained, the multi-factor time sequence of the product channel to be predicted is input into the actual yield prediction model, and the actual yield of the product channel to be predicted, which is output by the actual yield prediction model, is obtained. The multi-factor time sequence comprises a product name feature, a purchase times feature, a product channel feature and a policy issuing date feature, and the actual yield prediction model is a model obtained by model training based on full-chain insurance customer conversion link data and payment calculation related data of each sample product. In the technical scheme, the actual yield prediction model comprises the weight of the external factor, the weight of the external factor is obtained through model training, and the accuracy is high, so that the accuracy of the actual yield calculated based on the weight of the external factor is effectively improved.
Optionally, in some embodiments, the actual yield early warning function may be opened, and relevant early warning may be performed on a product channel in which the actual yield has a significant tendency to decrease. Specifically, if the difference of the actual yield of the product channel to be predicted minus the actual yield of the product channel to be predicted in the last period is larger than a preset difference, early warning is carried out on the product channel to be predicted.
The preset difference is exemplified by the difference of the decrease with the maximum actual yield under normal conditions, such as 10%, 15% or 20%, etc., and the specific value can be determined according to the actual situation.
In a specific implementation manner, the early warning can be performed by sending short messages, mails, push messages and the like to terminal equipment of related staff.
In the embodiment, the product channel to be predicted is monitored in real time according to the actual yield of the product channel to be predicted obtained through prediction, so that early warning can be performed in time when the actual yield has a significant decreasing trend, and the product channel to be predicted can be processed.
Optionally, in some embodiments, the predicted actual yields for the various underwriting years and months for the different product channels over a particular time period may also be presented in a customer retention analysis page of the user value system. Specifically, determining the target actual yield of each policy issuing date in a preset time period in the product channel to be predicted, and displaying the target actual yield of the product channel.
Alternatively, the specific time period may be preset, or may be acquired in response to a query operation by the relevant staff.
Model training is also required to obtain a yield prediction model for predicting yield before predicting yield. The model training process is described in detail below in connection with specific embodiments. The following embodiments may be combined with each other, and some embodiments may not be repeated for the same or similar concepts or processes.
In particular, the execution subject of the model training method may be an electronic device having processing capability, such as a terminal or a server. It should be understood that the electronic device for the model training method and the electronic device for performing the above-described prediction method of the actual yield may be the same device or may be different devices.
Fig. 2 is a schematic flow chart of a first embodiment of a model training method provided in the embodiment of the present application. As shown in fig. 2, the model training method may be implemented by:
s21, according to the full-chain insurance customer conversion link data of each sample product and the payment calculation related data of each sample product, a sample multi-factor time sequence of each product channel is constructed.
The sample multi-factor time sequence comprises a purchasing times feature, a product name feature, a product channel feature and a policy issuing date feature.
In one possible implementation, S21 may be implemented with reference to the following S221 to S214:
s211, constructing customer conversion detail data according to full-chain insurance customer conversion link data of each sample product.
Wherein the customer conversion details data are used to represent the first purchased sample product and the second purchased sample product for each customer.
S212, classifying sample products in the customer conversion detail data according to the service requirements and determining the sample products contained in each product channel.
S213, determining the purchase times characteristic of each sample product corresponding to each transaction from the customer conversion detail data.
S214, calculating relevant data, the purchase times characteristic of each sample product corresponding to each transaction and the product channel characteristic of each sample product according to payment of each transaction, and constructing a sample multi-factor time sequence of each product channel.
S22, performing model training according to the sample multi-factor time sequence to obtain a first model.
The first model is used for calculating the continuation rate of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted.
In one possible implementation, a sample multi-factor time sequence is modeled by using a Long Short-Term Memory (LSTM) mode, and a model parameter optimization model is adjusted by using a continuation rate as a target value to obtain a first model.
S23, performing model training according to the external factors and the continuation rates corresponding to the sample multi-factor time sequence to obtain a second model.
Wherein the second model contains weights for external factors.
In one possible implementation, the actual yield corresponding to the sample multi-factor time sequence is used as a target, the external factor and the continuation rate corresponding to the sample multi-factor time sequence are fitted through logistic regression (Logistic Regression, LR), and the model parameter optimization model is adjusted to obtain a second model.
Alternatively, the actual yield corresponding to the sample multifactor time series may be generated based on a refined chain ladder method.
S24, correcting the first model according to the weight of the external factors to obtain a yield prediction model.
The actual yield prediction model is the actual yield prediction model in the embodiment shown in fig. 1, and is used for calculating the actual yield of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted. The yield prediction model includes weights for the external factors.
According to the model training method provided by the embodiment of the application, the sample multi-factor time sequence of each product channel is constructed according to the full-chain insurance customer conversion link data of each sample product and the payment calculation related data of each sample product. And then, performing model training according to the sample multi-factor time sequence to obtain a first model. And then, performing model training according to the external factors and the continuation rates corresponding to the sample multi-factor time sequence to obtain a second model. And finally, correcting the first model according to the weight of the external factors to obtain a yield prediction model. The sample multi-factor time sequence comprises a purchase times feature, a product name feature, a product channel feature and a policy issuing date feature, the first model is used for calculating the continuation rate of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted, the second model comprises the weight of an external factor, and the actual yield prediction model is used for calculating the actual yield of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted. In the technical scheme, an original manual actual yield calculation mode is replaced by an innovative mode of adding LR into LSTM, external factors are introduced by referring to a traditional refined chain ladder method through deep learning modeling of time sequence curves of continuous rates of different product channels, and the model of deep learning training is corrected by using the external factors, so that the actual predicted actual yield is more in line with an actual expected value, and is equivalent to season or holiday factors introduced in a time sequence model.
Based on the prediction method and model training method of the actual yield shown in the above embodiments, the prediction process of the actual yield will be explained by a specific example.
Fig. 3 is a schematic flow chart of a second embodiment of a model training method provided in the embodiment of the present application. As shown in fig. 3, the model training method may be implemented by:
and step 1, according to full-chain insuring customer conversion link data of each sample product, combining the first purchased sample product and the second purchased sample product of the customer in the dimension of the customer to construct a customer conversion list.
And step 2, classifying the sample products, and determining the purchasing frequency characteristics of the sample products according to the first purchase and the repurchase labels.
And step 3, expanding a customer conversion list according to the payment period by combining the payment plan related data to generate a sample multi-factor time sequence of each product channel.
And 4, modeling the sample multi-factor time sequence in an LSTM mode, adjusting a model parameter optimization model, and generating a first model.
And 5, taking the actual yield of the sample multi-factor time sequence generated by the accurate chain ladder method as a target value, and fitting the external factor and the continuation rate of the sample multi-factor time sequence together through LR to obtain a second model.
And 6, correcting the weight of the external factors in the second model to the first model to generate a yield prediction model.
It is understood that the yield prediction model can be used as a model of the bottom layer dependence of the yield prediction display value in the staged product retention analysis page.
Summarizing, the landing solution of the embodiments of the present application may include the following steps:
processing bottom data related to product conversion (namely link data converted by full-chain insurance clients) in the full-link user value service system to generate related characteristics such as first purchase/repurchase, products, channels, and underwriting years, months, times, and continuing rate.
Expanding the processing data according to the period to generate a sample multi-factor time sequence.
And thirdly, training the sample multi-factor time sequence model by utilizing the LSTM, and constructing a first model.
And fourthly, performing LR model training on the recovery rate through the external factor and the first model to obtain the weight of the external factor.
And fifthly, reconstructing a yield prediction model on the basis of the first model according to the weight of the external factors.
And step six, predicting a subsequent continuous rate curve through the previous n-phase continuous rate after the model is online, and calculating the actual yield according to the weight of the external factors in the model.
And step seven, displaying actual yield predicted values of all the underwriting years and months in specific time periods of different product channels in a customer retention analysis page in a customer value system.
And step eight, opening a real yield early warning function, and carrying out related early warning on a product channel with obvious decreasing trend of the real yield.
In summary, the effect of the correlation analysis of the product actual yield in the full-chain user value project is expected to be realized, the actual yield predicted value of each product and each product channel is provided for each business department, and the product channel with obvious decreasing trend of the expected actual yield is correspondingly reminded, however, the prior art cannot cover the full-chain and cannot perform the first purchase and the second purchase correlation analysis through manual experience analysis, and the manual chain ladder method analysis period is longer, so that the time response to the change data cannot be performed. Aiming at the problems, the method expands a month-to-month continuous rate curve according to the underwriting year and month by introducing a time sequence deep learning mode, learns the change trend of the month-to-month continuous rate curve, corrects the learned prediction curve by combining an external factor, achieves the effect rate prediction function for different product channels in a rapid and automatic mode, can perfectly replace a manual mode and realize on-line deployment of a full-link effect rate model, and solves the problems that the prior art depends on the business familiarity of a computerer and the processing degree of bottom data, and can not realize free configuration and effective combination with product conversion link formation.
According to the method, the continuous rate of the full-scale month-to-month product of the company is learned in a deep learning mode, different product channels are aggregated according to the bearing year and month, the original data are expanded according to the 12-period continuous rate, a 12-period continuous rate curve (namely a multi-factor time sequence) of the historical bearing year and month is generated, learning modeling is conducted on the curve in the deep learning mode, the output result of the first model is corrected in the mode of introducing external factors and the artificial actual yield result, and the first model is subjected to relevant adjustment in combination with the weight of the external factors, so that the final actual yield prediction model is obtained.
In the method, the yield prediction index of different product channels in different underwriting years and months can be provided, and the method is used for a product yield analysis page in a full-chain user value system, is used as a reference value for the yield prediction of a business department, and can provide index early warning of yield and acquisition and repudiation scene yield index analysis for business. In the existing fine calculation report, a real yield calculation mode of a repurchase channel cannot be provided, and related prediction cannot be carried out on the real yield of each product channel of a full chain. Meanwhile, for the actual yield analysis of first purchase and repurchase, an accurate actual yield prediction index of repurchase can be provided for an accurate calculation business department, and references are provided for the actual conversion efficiency of each product channel and future policy delivery.
Meanwhile, the method can solve the problems of larger limitation, lower efficiency and accuracy existing in the traditional calculation by adopting a chain ladder method to calculate the actual yield from the technical aspect. The traditional mode is relatively dependent on bottom processing data, needs data developers to closely cooperate, processes the bottom data according to different requirements, needs a large amount of time to manually develop a related actual yield model, needs independent analysis aiming at different data, and has higher requirements on professional analysis and business capability of the computerised personnel. The method can directly analyze the bottom dimension layer data in a machine learning mode, data development personnel are not required to independently process the data according to requirements, a whole set of processes of model construction and model online are realized through automatic feature engineering, relevant prediction of the actual yield of short-risk month-paying products of the whole company can be realized in extremely short time, and a prediction result is automatically generated and a relevant early warning function is provided.
According to the method, a sample multi-factor time sequence based on the continuous rate of different product channels is established in the early stage, the continuous rate of the product channels is predicted by using a deep learning mode, the external factors are introduced to correct relevant parameters of the first model, meanwhile, model training can be popularized to the product channels of all companies by using an automatic mode, rapid deployment and automatic prediction are performed, and a traditional chain ladder method training mode aiming at limited data by manpower is comprehensively replaced. Meanwhile, the method and the device can effectively save labor cost, reduce the prediction threshold of the actual yield model, quickly perform model training and prediction, and perform model training on different product channels, and greatly improve the efficiency and diversity of the overall actual yield prediction model in an automatic mode. The problems that in the prior art, a computerist needs to accurately judge the continuous rate and the actual yield prediction trend by combining actual experience and business standards, different judgment standards cannot be given to different business types of different products, the utilization scene is limited, the computerised product cannot be popularized to different channel products of the whole company, and the related actual yield of the first-purchased product and the second-purchased product of the user cannot be combined to conduct distinguishing prediction are effectively solved.
In summary, the present application shares four key technical points:
firstly, constructing a sample multi-factor time sequence for a short-risk month paying product in a mode of expanding the stage continuous rate, and performing model training by utilizing an LSTM (least squares) to obtain a first model.
And secondly, carrying out target value correction on the first model by combining external factors in the accurate calculation chain ladder method to obtain a yield prediction model. The LSTM+LR mode is used to make the actual yield prediction model more approximate to the actual business requirement.
The third and actual yield prediction models can be based on a full-chain user value product transfer matrix, multi-factor first purchase and repurchase continuation rate model training can be performed, and automatic modeling can be performed on the actual yield of first purchase and repurchase of different product channels.
Fourth, the yield prediction model can also combine the later yield early warning function, can provide the product channel yield decline trend early warning for business department.
The following are device embodiments of the present application, which may be used to perform method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.
Fig. 4 is a schematic structural diagram of a device for predicting the actual yield according to an embodiment of the present application. As shown in fig. 4, the actual yield prediction apparatus 40 includes:
The obtaining module 41 is configured to obtain a multi-factor time sequence of the product channel to be predicted, where the multi-factor time sequence includes a product name feature, a purchase number feature, a product channel feature, and a policy issuing date feature.
The input module 42 is configured to input the multi-factor time sequence of the product channel to be predicted into a yield prediction model, and obtain the yield of the product channel to be predicted output by the yield prediction model, where the yield prediction model is a model obtained by performing model training based on full-chain insurance customer conversion link data and payment calculation related data of each sample product.
In one possible design, the yield prediction model is specifically used for:
and predicting the continuous rate of the product channel to be predicted according to the multi-factor time sequence of the product channel to be predicted.
And predicting the actual yield of the product channel to be predicted according to the external factor weight and the continuation rate.
In one possible design, the prediction apparatus 40 of the actual yield further includes an early warning module for:
and if the difference of the actual yield of the product channel to be predicted minus the actual yield of the product channel to be predicted in the last period is larger than the preset difference, early warning is carried out on the product channel to be predicted.
In one possible design, the apparatus 40 for predicting the yield of a product further comprises a display module for:
and determining the target actual yield of each policy issuing date in the preset time period in the product channel to be predicted.
And displaying the target actual yield of the product channel.
The prediction device for the actual yield provided in the embodiment of the present application may be used to execute the prediction method for the actual yield in any of the above embodiments, and its implementation principle and technical effects are similar, and are not described herein again.
Fig. 5 is a schematic structural diagram of a model training device according to an embodiment of the present application. As shown in fig. 5, the model training apparatus 50 includes:
the construction module 51 is configured to construct a sample multi-factor time sequence of each product channel according to the full-chain insurance customer conversion link data of each sample product and the payment calculation related data of each sample product, where the sample multi-factor time sequence includes a purchase number feature, a product name feature, a product channel feature, and an insurance policy issue date feature.
The training module 52 is configured to perform model training according to the sample multi-factor time sequence, so as to obtain a first model, where the first model is used to calculate a continuation rate of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted.
The training module 52 is further configured to perform model training according to the external factor and the continuation rate corresponding to the sample multi-factor time sequence, to obtain a second model, where the second model includes the weight of the external factor.
The correction module 53 is configured to correct the first model according to the weight of the external factor, and obtain a yield prediction model, where the yield prediction model is used to calculate the yield of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted.
In one possible design, the module 51 is constructed specifically for:
and constructing customer conversion detail data according to the full-chain insurance customer conversion link data of each sample product, wherein the customer conversion detail data is used for representing the first purchased sample product and the second purchased sample product of each customer.
And classifying sample products in the customer conversion detail data according to the service requirements and determining the sample products contained in each product channel.
From the customer conversion profile data, a number of purchases characteristic for each sample product corresponding to each transaction is determined.
And calculating related data, the purchase times characteristic of each sample product corresponding to each transaction and the product channel characteristic of each sample product according to the payment calculation of each transaction, and constructing a sample multi-factor time sequence of each product channel.
The model training device provided in the embodiment of the present application may be used to execute the model training method in any of the above embodiments, and its implementation principle and technical effects are similar, and are not described herein again.
It should be noted that, it should be understood that the division of the modules of the above apparatus is merely a division of a logic function, and may be fully or partially integrated into a physical entity or may be physically separated. And these modules may all be implemented in the form of software calls through the processing elements. Or may be implemented entirely in hardware. The method can also be realized in a form of calling software by a processing element, and the method can be realized in a form of hardware by a part of modules. In addition, all or part of the modules may be integrated together or may be implemented independently. The processing element here may be an integrated circuit with signal processing capabilities. In implementation, each step of the above method or each module above may be implemented by an integrated logic circuit of hardware in a processor element or an instruction in a software form.
Fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application. As shown in fig. 6, the electronic device 60 may include: the system comprises a processor 61, a memory 62 and computer program instructions stored on the memory 62 and capable of running on the processor 61, wherein the processor 61 executes the computer program instructions to realize the prediction method or the model training method of the actual yield provided by any of the previous embodiments.
Alternatively, the above-mentioned respective devices of the electronic apparatus 60 may be connected by a system bus.
The memory 62 may be a separate memory unit or may be a memory unit integrated into the processor. The number of processors is one or more.
Optionally, the electronic device 60 may also include a communication interface to interact with other devices.
It should be appreciated that the processor 61 may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the present application may be embodied directly in a hardware processor or in a combination of hardware and software modules within a processor.
The system bus may be a peripheral component interconnect standard (peripheral component interconnect, PCI) bus or an extended industry standard architecture (extended industry standard architecture, EISA) bus, among others. The system bus may be classified into an address bus, a data bus, a control bus, and the like. For ease of illustration, the figures are shown with only one bold line, but not with only one bus or one type of bus. The memory may include random access memory (random access memory, RAM) and may also include non-volatile memory (NVM), such as at least one disk memory.
All or part of the steps for implementing the method embodiments described above may be performed by hardware associated with program instructions. The foregoing program may be stored in a readable memory. The program, when executed, performs steps including the method embodiments described above; and the aforementioned memory (storage medium) includes: read-only memory (ROM), RAM, flash memory, hard disk, solid state disk, magnetic tape, floppy disk, optical disk (optical disc), and any combination thereof.
The electronic device provided in the embodiment of the present application may be used to execute the prediction method or the model training method of the actual yield provided in any of the above method embodiments, and its implementation principle and technical effects are similar and are not described herein again.
Embodiments of the present application provide a computer-readable storage medium having stored therein computer-executable instructions that, when executed on a computer, cause the computer to perform the above-described yield prediction method or model training method.
The computer readable storage medium described above may be implemented by any type of volatile or non-volatile memory device or combination thereof, such as static random access memory, electrically erasable programmable read-only memory, magnetic memory, flash memory, magnetic disk or optical disk. A readable storage medium can be any available medium that can be accessed by a general purpose or special purpose computer.
In the alternative, a readable storage medium is coupled to the processor such that the processor can read information from, and write information to, the readable storage medium. In the alternative, the readable storage medium may be integral to the processor. The processor and the readable storage medium may reside in an application specific integrated circuit (Application Specific Integrated Circuits, ASIC). The processor and the readable storage medium may reside as discrete components in a device.
Embodiments of the present application also provide a computer program product, where the computer program product includes a computer program, where the computer program is stored in a computer readable storage medium, and at least one processor may read the computer program from the computer readable storage medium, where the at least one processor may implement the prediction method or the model training method of the actual yield when executing the computer program.
It is to be understood that the present application is not limited to the precise arrangements and instrumentalities shown in the drawings, which have been described above, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the application is limited only by the appended claims.
Claims (10)
1. A method for predicting the actual yield is characterized by comprising the following steps:
acquiring a multi-factor time sequence of a product channel to be predicted, wherein the multi-factor time sequence comprises a product name feature, a purchase frequency feature, a product channel feature and a policy issuing date feature;
inputting the multi-factor time sequence of the product channel to be predicted into a yield prediction model, and obtaining the yield of the product channel to be predicted output by the yield prediction model, wherein the yield prediction model is a model obtained by model training based on full-chain insuring customer conversion link data and payment calculation related data of each sample product.
2. The method according to claim 1, wherein the inputting the multi-factor time series of the product channel to be predicted into the actual yield prediction model to obtain the actual yield of the product channel to be predicted output by the actual yield prediction model includes:
predicting the continuation rate of the product channel to be predicted according to the multi-factor time sequence of the product channel to be predicted;
and predicting the actual yield of the product channel to be predicted according to the external factor weight and the continuation rate.
3. The method according to claim 1 or 2, characterized in that the method further comprises:
And if the difference of the actual yield of the product channel to be predicted subtracted from the actual yield of the product channel to be predicted in the last period is larger than a preset difference, early warning is carried out on the product channel to be predicted.
4. The method according to claim 1 or 2, characterized in that the method further comprises:
determining target actual yield of each policy issuing date in a preset time period in the product channel to be predicted;
and displaying the target actual yield of the product channel.
5. A method of model training, comprising:
according to the full-chain insurance customer conversion link data of each sample product and payment calculation related data of each sample product, constructing a sample multi-factor time sequence of each product channel, wherein the sample multi-factor time sequence comprises a purchasing frequency characteristic, a product name characteristic, a product channel characteristic and a policy issuing date characteristic;
model training is carried out according to the sample multi-factor time sequence to obtain a first model, and the first model is used for calculating the continuation rate of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted;
model training is carried out according to the external factors and the continuation rates corresponding to the sample multi-factor time sequence, so that a second model is obtained, and the second model contains the weights of the external factors;
Correcting the first model according to the weight of the external factor to obtain a yield prediction model, wherein the yield prediction model is used for calculating the yield of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted.
6. The method of claim 5, wherein the constructing the sample multi-factor time series for each product channel based on the full chain insurance customer conversion link data for each sample product and the payment calculation related data for each sample product comprises:
constructing customer conversion detail data according to full-chain insurance customer conversion link data of each sample product, wherein the customer conversion detail data is used for representing the first purchased sample product and the second purchased sample product of each customer;
classifying sample products in the customer conversion detail data according to the service requirements and product channels to determine sample products contained in each product channel;
determining the purchase times characteristic of each sample product corresponding to each transaction from the customer conversion detail data;
and calculating related data, the purchase times characteristic of each sample product corresponding to each transaction and the product channel characteristic of each sample product according to the payment calculation of each transaction, and constructing a sample multi-factor time sequence of each product channel.
7. An apparatus for predicting a yield, comprising:
the acquisition module is used for acquiring a multi-factor time sequence of a product channel to be predicted, wherein the multi-factor time sequence comprises a product name feature, a purchase frequency feature, a product channel feature and a policy issuing date feature;
the input module is used for inputting the multi-factor time sequence of the product channel to be predicted into a yield prediction model, obtaining the yield of the product channel to be predicted output by the yield prediction model, wherein the yield prediction model is a model obtained by model training based on full-chain insurance customer conversion link data and payment calculation related data of each sample product.
8. A model training device, comprising:
the construction module is used for constructing a sample multi-factor time sequence of each product channel according to full-chain insurance customer conversion link data of each sample product and payment calculation related data of each sample product, wherein the sample multi-factor time sequence comprises a purchase time feature, a product name feature, a product channel feature and an insurance policy issuing date feature;
the training module is used for carrying out model training according to the sample multi-factor time sequence to obtain a first model, and the first model is used for calculating the continuation rate of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted;
The training module is further used for performing model training according to the external factors and the continuation rates corresponding to the sample multi-factor time sequence to obtain a second model, wherein the second model comprises the weights of the external factors;
the correction module is used for correcting the first model according to the weight of the external factor to obtain a yield prediction model, and the yield prediction model is used for calculating the yield of the product channel to be predicted according to the sample multi-factor time sequence of the product channel to be predicted.
9. An electronic device, comprising: a processor, a memory and computer program instructions stored on the memory and executable on the processor, wherein the processor is adapted to implement the method of any one of claims 1 to 7 when executing the computer program instructions.
10. A computer readable storage medium having stored therein computer executable instructions which when executed by a processor are adapted to carry out the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311632800.0A CN117593039A (en) | 2023-11-29 | 2023-11-29 | Method for predicting yield, method for training model, device, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311632800.0A CN117593039A (en) | 2023-11-29 | 2023-11-29 | Method for predicting yield, method for training model, device, equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117593039A true CN117593039A (en) | 2024-02-23 |
Family
ID=89909720
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311632800.0A Pending CN117593039A (en) | 2023-11-29 | 2023-11-29 | Method for predicting yield, method for training model, device, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117593039A (en) |
-
2023
- 2023-11-29 CN CN202311632800.0A patent/CN117593039A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8014983B2 (en) | Computer-implemented system and method for storing data analysis models | |
| US7945472B2 (en) | Business management tool | |
| US8219440B2 (en) | System for enhancing business performance | |
| CN109583729B (en) | Data processing method and device for platform online model | |
| WO2020024718A1 (en) | Method and device for predicting foreign exchange transaction volume | |
| CN112541817A (en) | Marketing response processing method and system for potential customers of personal consumption loan | |
| AU2014201264A1 (en) | Scenario based customer lifetime value determination | |
| JP2001125962A (en) | Support system for management consulting and decision making in management | |
| CN111353901A (en) | Risk identification monitoring method and device and electronic equipment | |
| US11037236B1 (en) | Algorithm and models for creditworthiness based on user entered data within financial management application | |
| US10178241B2 (en) | Telecommunication price-based routing apparatus, system and method | |
| US8589191B1 (en) | Home resale price protection plan | |
| CN108416662B (en) | Data verification method and device | |
| CN110610415B (en) | Method and device for updating model | |
| CN117011031A (en) | Intelligent financial risk monitoring and early warning platform using big data information processing technology | |
| Shahnazarian et al. | FORECASTING AND ANALYSING CORPORATE TAX REVENUES IN SWEDEN USING BAYESIAN VAR MODELS. | |
| US7702555B1 (en) | Vintage maturation analytics for predicting behavior and projecting cash flow for customer communities and their responses to economic, competitive, or management changes | |
| CN117593039A (en) | Method for predicting yield, method for training model, device, equipment and medium | |
| CN112200340A (en) | Block chain system for predicting escaping waste and debt | |
| Guerini et al. | Zombification of the economy? Assessing the effectiveness of French government support during COVID-19 lockdown | |
| KR101954132B1 (en) | A delinquent disposal system | |
| JP7298286B2 (en) | Model providing program, model providing method and model providing apparatus | |
| KR101954131B1 (en) | An unmanned payment system | |
| JP2021502653A (en) | Systems and methods for automated preparation of visible representations regarding the achievability of goals | |
| KR101954129B1 (en) | Smart Communication FinTech Payment System |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |