CN110717806B - Product information pushing method, device, equipment and storage medium - Google Patents
Product information pushing method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN110717806B CN110717806B CN201910847224.9A CN201910847224A CN110717806B CN 110717806 B CN110717806 B CN 110717806B CN 201910847224 A CN201910847224 A CN 201910847224A CN 110717806 B CN110717806 B CN 110717806B
- Authority
- CN
- China
- Prior art keywords
- information
- preset
- client
- decision tree
- sub
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 94
- 238000007637 random forest analysis Methods 0.000 claims abstract description 50
- 238000003066 decision tree Methods 0.000 claims description 128
- 239000011159 matrix material Substances 0.000 claims description 59
- 230000010354 integration Effects 0.000 claims description 33
- 238000012360 testing method Methods 0.000 claims description 24
- 238000012549 training Methods 0.000 claims description 22
- 238000004891 communication Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 abstract description 10
- 238000012216 screening Methods 0.000 abstract 1
- 238000004364 calculation method Methods 0.000 description 2
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 1
- 239000000463 material Substances 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
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- 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
-
- 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
Abstract
The invention provides a product information pushing method, a device, equipment and a readable storage medium, wherein the method comprises the following steps: collecting client information of each client received in a history preset time period; performing dimension reduction processing on the client information based on a preset manifold learning algorithm model to obtain dimension reduction information; inputting the dimension reduction information into a preset random forest algorithm model to classify each client and obtain a target classification result; and acquiring the association relation between the preset client type and different product information, and directionally pushing the corresponding product information to each client according to the association relation and the target classification result. The invention solves the technical problem that the information pushing of insurance products is inaccurate because of difficult accurate screening of useful customer information based on an intelligent decision mode.
Description
Technical Field
The present invention relates to the field of computer technologies, and in particular, to a method, an apparatus, a device, and a readable storage medium for pushing product information.
Background
In the big data age, relevant information of clients is more and more, the dimension is high, the complexity is high, in the prior art, useful client information is obtained manually by using an original data analysis method, the useful client information is difficult to obtain from a large amount of redundant data by using the existing original data analysis method, manpower, material resources and financial resources are wasted, and the useful client information is difficult to accurately obtain, so that accurate recommendation is difficult to carry out according to the useful client information.
Disclosure of Invention
The invention mainly aims to provide a product information pushing method, device and equipment and a readable storage medium, and aims to solve the technical problem that in the prior art, because useful customer information is difficult to screen accurately, insurance product information pushing is inaccurate.
In order to achieve the above object, the present invention provides a product information pushing method, including:
collecting client information of each client received in a history preset time period;
acquiring a preset number of neighbor samples of each piece of customer information from a preset manifold learning algorithm model, and acquiring sample information corresponding to the neighbor samples and a local reconstruction weight matrix of each piece of customer information;
determining a low-dimensional matrix of the client information according to the local reconstruction weight matrix and the sample information so as to obtain dimension reduction information according to the low-dimensional matrix;
inputting the dimension reduction information into a preset random forest algorithm model to classify each client and obtain a target classification result;
and acquiring the association relation between the preset client type and different product information, and directionally pushing the corresponding product information to each client according to the association relation and the target classification result.
Optionally, the step of obtaining a preset number of neighbor samples of each piece of customer information from the manifold learning algorithm model, and obtaining sample information corresponding to the neighbor samples and a local reconstruction weight matrix of each piece of customer information includes:
acquiring each Euclidean distance of each piece of customer information and sample information corresponding to each sample in the manifold learning algorithm model according to a preset Euclidean distance algorithm;
acquiring a preset number of neighbor samples of each piece of customer information according to the size of each Euclidean distance;
determining a weight coefficient vector of each piece of customer information and each piece of sample information according to a preset local covariance algorithm;
and determining the local reconstruction weight matrix of the sample information corresponding to the neighbor sample and each piece of client information according to the weight coefficient vector and the association relation between the preset weight coefficient vector and the local reconstruction weight matrix.
Optionally, before the step of obtaining a preset number of neighbor samples of each piece of customer information from the manifold learning algorithm model, the method includes:
acquiring the name of each sub-dimension of the client information, and determining the integration sequence of the client information dimension integration according to the name of each sub-dimension, wherein the integration sequence comprises the sequence determination in an alphabet according to the first letter of the name of each sub-dimension;
And dimension integration is carried out on the client information according to the integration sequence, so that integrated client information is obtained.
Optionally, the step of inputting the dimension reduction information into a preset random forest algorithm model to classify the clients and obtain the target classification result includes:
acquiring preset use cases and each sub-dimension of the dimension reduction information;
carrying out random selection of different numbers of sub-dimensions with different times from the dimension reduction information so as to construct different decision tree clusters, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random;
the decision tree cluster generation process comprises the following steps: selecting N sub-dimensions from all the N sub-dimensions of the dimension reduction information, wherein N is smaller than N, randomly determining the decision influence degree of the N sub-dimensions, and obtaining a decision tree based on the decision influence degree of the N sub-dimensions and the decision influence degree of the N sub-dimensions, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random, establishing m decision trees, forming random forests by the m decision trees, and searching the number of each decision tree by a grid to obtain each decision tree cluster;
Setting the use cases with preset proportion in the use cases as first use cases, and setting other use cases with first exception as second use cases;
and carrying out adjustment training on decision influence degree in different decision tree clusters by taking the first example as a training example, carrying out testing on the decision tree clusters by taking the second example as a testing example, selecting an optimal target decision tree cluster with testing error smaller than a preset error value from all the decision tree clusters, and setting the target decision tree cluster as a preset random forest algorithm model.
Optionally, the step of inputting the dimension reduction information into a preset random forest algorithm model to classify the clients and obtain a target classification result includes:
inputting the dimension reduction information into a preset random forest algorithm model, and acquiring classification of different decision trees in the preset random forest algorithm model on the dimension reduction information to obtain different initial target classification results;
and determining the target classification result according to the classification duty ratio of different initial target classification results.
The invention also provides a product information pushing device, which comprises:
The collection module is used for collecting the client information of the clients received in the historical preset time period;
the first acquisition module is used for acquiring a preset number of neighbor samples of each piece of client information from a preset manifold learning algorithm model, and acquiring sample information corresponding to the neighbor samples and a local reconstruction weight matrix of each piece of client information;
the second acquisition module is used for determining a low-dimensional matrix of the client information according to the local reconstruction weight matrix and the sample information so as to obtain dimension reduction information according to the low-dimensional matrix;
the classification module is used for inputting the dimension reduction information into a preset random forest algorithm model so as to classify the clients and obtain a target classification result;
and the third acquisition module is used for acquiring the association relation between the preset client type and different insurance products, and directionally recommending the insurance products to each client according to the association relation and the target classification result. Optionally, the first acquisition module includes:
the first acquisition unit is used for acquiring each Euclidean distance of each piece of customer information and sample information corresponding to each sample in the manifold learning algorithm model according to a preset Euclidean distance algorithm;
The second acquisition unit is used for acquiring a preset number of neighbor samples of each piece of customer information according to the size of each Euclidean distance;
the first determining unit is used for determining a weight coefficient vector of each piece of client information and each piece of sample information according to a preset local covariance algorithm;
and the second determining unit is used for determining the sample information corresponding to the neighbor samples and the local reconstruction weight matrix of each piece of client information according to the weight coefficient vector and the association relation between the weight coefficient vector and the preset weight coefficient vector and the local reconstruction weight matrix.
Optionally the product information pushing device comprises:
a fourth obtaining module, configured to obtain a name of each sub-dimension of the client information, and determine an integration sequence of dimension integration of the client information according to the name of each sub-dimension, where the integration sequence includes determining according to a sequence of initial letters of the name of each sub-dimension in an alphabet;
and the integration module is used for carrying out dimension integration on the client information according to the integration sequence to obtain integrated client information.
Optionally, the product information pushing device includes:
a fifth obtaining module, configured to obtain a preset use case and each sub-dimension of the dimension reduction information;
The random selection module is used for carrying out random selection of different numbers of sub-dimensions with different times from the dimension reduction information so as to construct different decision tree clusters, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random;
the decision tree cluster generation process comprises the following steps: selecting N sub-dimensions from all the N sub-dimensions of the dimension reduction information, wherein N is smaller than N, randomly determining the decision influence degree of the N sub-dimensions, and obtaining a decision tree based on the decision influence degree of the N sub-dimensions and the decision influence degree of the N sub-dimensions, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random, establishing m decision trees, forming random forests by the m decision trees, and searching the number of each decision tree by a grid to obtain each decision tree cluster;
the setting module is used for setting the use cases with preset proportion in the use cases as first use cases, and setting other use cases with the first exception as second use cases;
the adjustment training module is used for performing adjustment training on the decision influence degree in the different decision tree clusters by taking the first case as a training case, performing testing on the decision tree clusters by taking the second case as a testing case, selecting an optimal target decision tree cluster with testing error smaller than a preset error value from all the decision tree clusters, and setting the target decision tree cluster as a preset random forest algorithm model.
Optionally, the classification module includes:
the third acquisition unit is used for inputting the dimension reduction information into a preset random forest algorithm model, and acquiring classification of different decision trees in the preset random forest algorithm model on the dimension reduction information to obtain different initial target classification results;
and the classification unit is used for determining the target classification result according to the classification duty ratio of different initial target classification results.
In addition, to achieve the above object, the present invention also provides a readable storage medium storing one or more programs executable by one or more processors for:
collecting client information of each client received in a history preset time period;
performing dimension reduction processing on the client information based on a preset manifold learning algorithm model to obtain dimension reduction information;
inputting the dimension reduction information into a preset random forest algorithm model to classify each client and obtain a target classification result;
and acquiring the association relation between the preset client type and different product information, and directionally pushing the corresponding product information to each client according to the association relation and the target classification result.
The method comprises the steps of collecting client information of clients received in a historical preset time period; acquiring a preset number of neighbor samples of each piece of customer information from a preset manifold learning algorithm model, and acquiring sample information corresponding to the neighbor samples and a local reconstruction weight matrix of each piece of customer information; determining a low-dimensional matrix of the client information according to the local reconstruction weight matrix and the sample information so as to obtain dimension reduction information according to the low-dimensional matrix; inputting the dimension reduction information into a preset random forest algorithm model to classify each client and obtain a target classification result; and acquiring the association relation between the preset client type and different product information, and directionally pushing the corresponding product information to each client according to the association relation and the target classification result. In the method, the useful customer information is screened according to the preset manifold learning algorithm model, then the classification of the customers is carried out according to the random forest algorithm model and the useful customer information, after the classification, the accurate recommendation of the insurance products is carried out on the customers of different types according to the association relation between the preset customer types and the different insurance products, so that the technical problem that the useful customer information is difficult to accurately screen and the insurance product information pushing is inaccurate is solved.
Drawings
FIG. 1 is a flowchart of a first embodiment of a product information pushing method according to the present invention;
FIG. 2 is a detailed flow chart of the steps of performing dimension reduction processing on the client information based on a preset manifold learning algorithm model to obtain dimension reduction information in the product information pushing method of the invention;
FIG. 3 is a schematic diagram of a device architecture of a hardware operating environment involved in a method according to an embodiment of the present invention.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The present invention provides a product information pushing method, in a first embodiment of the product information pushing method of the present invention, referring to fig. 1, the product information pushing method includes:
step S10, collecting the client information of the clients received in the history preset time period;
step S20, obtaining a preset number of neighbor samples of each piece of customer information from a preset manifold learning algorithm model, and obtaining sample information corresponding to the neighbor samples and a local reconstruction weight matrix of each piece of customer information;
Step S30, determining a low-dimensional matrix of the client information according to the local reconstruction weight matrix and the sample information so as to obtain dimension reduction information according to the low-dimensional matrix;
step S40, inputting the dimension reduction information into a preset random forest algorithm model to classify the clients and obtain a target classification result;
step S50, obtaining the association relation between the preset client type and different product information, and directionally pushing the corresponding product information to each client according to the association relation and the target classification result.
The method comprises the following specific steps:
step S10, collecting the client information of the clients received in the history preset time period;
in this embodiment, the product information pushing method is applied to the product information pushing system, and the collection of the client information of each client acquired by the product information pushing system in the past preset time period may be half a year, and the preset time period may be one year.
Specifically, the client information includes the sub-dimensions of client name, age, address, academic, wage, identification card number, etc., the purpose of collecting the client information is to ensure the timely effectiveness of the client information, so as to realize the purpose of accurate recommendation according to the freshest data, after the client information is obtained, the client information is integrated in dimensions, and the dimension integration is that: the client information of different sub-dimensions is ordered according to a certain sequence, for example, the client information comprises sub-dimensions of name, address, academic, age and the like, the client information is determined to be ordered according to the sequence of name, age, address and academic, for example, the client information is ordered according to the sequence of Zhang three, 25 years, the address Shenzhen mountain, university academy and the like, and the aim of ordering is to be able to orderly determine the coordinates of the client information when the dimension reduction processing is performed.
Step S20, obtaining a preset number of neighbor samples of each piece of customer information from a preset manifold learning algorithm model, and obtaining sample information corresponding to the neighbor samples and a local reconstruction weight matrix of each piece of customer information;
in this embodiment, after obtaining the client information, the dimension reduction processing is performed on the integrated information based on a preset manifold learning algorithm model to obtain dimension reduction information, where the preset manifold learning algorithm model is a pre-stored and trained model capable of performing dimension reduction processing on the integrated information, and in the preset manifold learning algorithm model, the dimension reduction processing on the integrated information is performed mainly based on a popular learning algorithm.
In this embodiment, the client information is input into the preset manifold learning algorithm model, the preset manifold learning algorithm model performs coordinate position determination processing on the client information of each client first, performs neighbor sample determination processing on a preset number of the client information after determining the coordinate position of the client information, and obtains sample information corresponding to the neighbor sample and a local reconstruction weight matrix of each piece of client information.
The step of obtaining a preset number of neighbor samples of each piece of customer information from the manifold learning algorithm model comprises the following steps:
Step A1, acquiring the name of each sub-dimension of the client information, and determining the integration sequence of the client information dimension integration according to the name of each sub-dimension, wherein the integration sequence comprises determining according to the sequence of the initial letter of the name of each sub-dimension in an alphabet;
in this embodiment, the integration order is determined according to the names of each of the sub-dimensions, specifically, the order of the first letter of the name of each of the sub-dimensions in the alphabet, and if the first letters of the names of different sub-dimensions are the same, the order of the second letters of the names of different sub-dimensions in the alphabet is determined, where the more front the first letter of the name is in the alphabet, the more front the integration order, for example, the learning X is in front of the address Z, and thus the learning is in front of the address in the integration order.
And step A2, dimension integration is carried out on the client information according to the integration sequence, and integrated client information is obtained.
And dimension integration is carried out on the client information according to the integration sequence, so that integrated client information is obtained.
Specifically, the step of obtaining a preset number of neighbor samples of each piece of customer information from the manifold learning algorithm model, and obtaining sample information corresponding to the neighbor samples and a local reconstruction weight matrix of each piece of customer information includes:
Step S211, according to a preset Euclidean distance algorithm, acquiring each Euclidean distance of each piece of customer information and sample information corresponding to each sample in the manifold learning algorithm model;
in this embodiment, the manifold learning algorithm model includes each customer sample, where each customer sample is simply referred to as a sample, the sample corresponds to customer information of the sample, and the customer information of the sample is simply referred to as sample information, and the sample information also includes sub-dimensions of customer name, age, address, academic, wage, id number, etc., where coordinates of the sample information corresponding to each sample in the manifold learning algorithm model are determined.
The manifold learning algorithm model includes a euclidean distance algorithm or a cosine algorithm (existing calculation formula) to obtain each euclidean distance between each piece of customer information and sample information corresponding to each sample in the manifold learning algorithm model, or obtain each cosine distance between each piece of customer information and sample information corresponding to each sample in the manifold learning algorithm model, where the calculation formula of the euclidean distance is as follows:
in the formula (1), D is a euclidean distance, and X and Y are sample information and client information in a certain dimension, respectively.
Step S212, obtaining a preset number of neighbor samples of each piece of customer information according to the size of each Euclidean distance;
after determining each Euclidean distance, obtaining a preset number of neighbor samples of each piece of customer information from low to high according to the size of each Euclidean distance, namely selecting the preset number of neighbor samples for each piece of customer information, wherein the preset number can be 7 or 8.
Step S213, determining a weight coefficient vector of each piece of customer information and each piece of sample information according to a preset local covariance algorithm;
according to a preset local covariance algorithm, determining a weight coefficient vector of each piece of client information and each piece of sample information, wherein the specific determining process comprises the following steps:
assuming that the neighbor samples of the client information xi are (xi 1, xi2,., xik), a preset local covariance algorithm is obtained, and a local covariance matrix zi= (xi-xj) is obtained T (xi-xj) based on the local covariance matrix zi= (xi-xj) T (xi-xj) obtaining a corresponding weight coefficient vector Wi;
Wi＝(Zi-11 k )/(1 k TZi-11 k )
wherein the nearest neighbor number is k.
Step S214, determining a local reconstruction weight matrix of each piece of customer information and sample information corresponding to the neighboring samples according to the weight coefficient vector and the association relation between the preset weight coefficient vector and the local reconstruction weight matrix.
After the weight coefficient vector is obtained, a weight coefficient matrix W is composed of the weight coefficient vector Wi, and a matrix m= (I-W) T is calculated.
Step S30, determining a low-dimensional matrix of the client information according to the local reconstruction weight matrix and the sample information so as to obtain dimension reduction information according to the low-dimensional matrix;
after obtaining the local reconstruction weight matrix, determining a low-dimensional matrix of the client information according to the sample matrix corresponding to the local reconstruction weight matrix and the sample information, so as to obtain dimension reduction information according to the low-dimensional matrix, wherein the dimension reduction information is information which is selected from the client information and is useful for accurate recommendation or has great influence on the client information.
Step S40, inputting the dimension reduction information into a preset random forest algorithm model to classify the clients and obtain a target classification result;
after the dimension reduction information is obtained, the dimension reduction information is input into a preset random forest algorithm model to classify the clients and obtain a target classification result, wherein the preset random forest algorithm model is a model capable of classifying the clients based on the dimension reduction information. In this embodiment, the dimension reduction information, but not all the client information, is input as input data into a preset random forest algorithm model, so that the accuracy of prediction can be improved.
Step S50, obtaining the association relation between the preset client type and different product information, and directionally pushing the corresponding product information to each client according to the association relation and the target classification result.
After the target classification result is obtained, the target type of the user is obtained, the association relation between the preset client type and different product information is obtained, and the corresponding product information is directionally pushed to each client according to the association relation and the target classification result, for example, the client type is high-knowledge type, the Q insurance is directionally recommended to each client, the client type is cautious type, and the T insurance is directionally recommended to each client.
The method comprises the steps of collecting client information of clients received in a historical preset time period; acquiring a preset number of neighbor samples of each piece of customer information from a preset manifold learning algorithm model, and acquiring sample information corresponding to the neighbor samples and a local reconstruction weight matrix of each piece of customer information; determining a low-dimensional matrix of the client information according to the local reconstruction weight matrix and the sample information so as to obtain dimension reduction information according to the low-dimensional matrix; inputting the dimension reduction information into a preset random forest algorithm model to classify each client and obtain a target classification result; and acquiring the association relation between the preset client type and different product information, and directionally pushing the corresponding product information to each client according to the association relation and the target classification result. In the method, the useful customer information is screened according to the preset manifold learning algorithm model, then the classification of the customers is carried out according to the random forest algorithm model and the useful customer information, after the classification, the accurate recommendation of the insurance products is carried out on the customers of different types according to the association relation between the preset customer types and the different insurance products, so that the technical problem that the useful customer information is difficult to accurately screen and the insurance product information pushing is inaccurate is solved.
Further, the present invention provides another embodiment of a product information pushing method, in this embodiment, the step of inputting the dimension reduction information into a preset random forest algorithm model to classify the clients and obtain a target classification result includes:
step B1, acquiring preset use cases and each sub-dimension corresponding to the dimension reduction information;
it should be noted that, in this embodiment, multiple training and testing processes are required before the preset random forest algorithm model is obtained, where, before the process, a pre-stored use case and each sub-dimension of the dimension reduction information are required to be obtained, where each sub-dimension of the dimension reduction information refers to a dimension, rather than specific dimension information, such as an academic dimension in the dimension reduction information, rather than specific content of a master.
The random forest is a forest formed by a plurality of independent decision trees, the decision trees are mutually independent, the weight of each tree is equal, a plurality of decision trees can be quickly and conveniently generated by using a sklearn library of Python, the decision trees are classified processes, specifically, each node of the decision trees represents a test on an attribute, each branch represents a test output, and each node on the decision tree represents a category.
Step B2, randomly selecting different numbers of sub-dimensions from the dimension reduction information for different times to construct different decision tree clusters, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random;
the decision tree cluster generation process comprises the following steps: selecting N sub-dimensions from all the N sub-dimensions of the dimension reduction information, wherein N is smaller than N, determining the decision influence degree of the N sub-dimensions randomly, and obtaining a decision tree based on the decision influence degree of the N sub-dimensions and the decision influence degree of the N sub-dimensions, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random, establishing m decision trees, forming random forests by the m decision trees, and searching the number of each decision tree by a grid to obtain each decision tree cluster.
After the dimension reduction information is obtained, carrying out random selection of different numbers of sub-dimensions with different times from the dimension reduction information so as to construct different decision tree clusters, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random.
The decision tree generation process may be: selecting N sub-dimensions from all the N sub-dimensions of the dimension reduction information, wherein N is smaller than N, randomly determining the decision influence degree of the N sub-dimensions, obtaining a decision tree based on the decision influence degree of the N sub-dimensions and the decision influence degree of the N sub-dimensions, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random, repeating the above m times to establish m decision trees, forming a random forest by the m decision trees, repeating the above two steps for w times to establish w decision trees, forming a random forest by the w decision trees, and searching the number of each decision tree by a grid to obtain each decision tree cluster.
Step B3, setting the use cases with preset proportion in the use cases as first use cases, and setting other use cases with the exception of the first training use cases as second use cases;
after each decision tree cluster is obtained, the use cases with the preset proportion in the use cases are set as first use cases, and other use cases except for the first training use cases are set as second use cases, and it is to be noted that the setting of a plurality of first use cases and a plurality of corresponding second use cases can be performed in a sampling-back mode so as to more accurately determine the target decision tree cluster based on the plurality of first use cases and the plurality of corresponding second use cases.
And B4, carrying out adjustment training on decision influence degree in the different decision tree clusters by taking the first case as a training case, carrying out testing on the decision tree clusters by taking the second case as a testing case, selecting an optimal target decision tree cluster with testing error smaller than a preset error value from all the decision tree clusters, and setting the target decision tree cluster as a preset random forest algorithm model.
After a first case and a second case are obtained, the first case is used as a training case to conduct adjustment training of decision influence degree in different decision tree clusters, specifically, the first case is input into different target decision tree clusters to obtain training results so as to judge how much of the first case is predicted correctly, so that a decision tree cluster closest to a real recording result is obtained, after the decision tree cluster closest to the real recording result is obtained, the first case is used as a training case to conduct adjustment training of decision influence degree in the different decision tree clusters, the second case is used as a test case to conduct testing of the decision tree clusters, an optimal target decision tree cluster with testing error smaller than a preset error value is selected from all the decision tree clusters, and the target decision tree cluster is set as a preset random forest algorithm model.
In this embodiment, the preset use case and each sub-dimension of the dimension reduction information are obtained; carrying out random selection of different numbers of sub-dimensions with different times from the dimension reduction information so as to construct different decision tree clusters, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random; setting the use cases with preset proportion in the use cases as first use cases, and setting other use cases except for the first training use cases as second use cases; and carrying out adjustment training on decision influence degree in different decision tree clusters by taking the first example as a training example, carrying out testing on the decision tree clusters by taking the second example as a testing example, selecting an optimal target decision tree cluster with testing error smaller than a preset error value from all the decision tree clusters, and setting the target decision tree cluster as a preset random forest algorithm model. In this embodiment, the random forest algorithm is used to construct a decision tree based on the dimension reduction information, so as to avoid that useless client information affects the recommendation of insurance products.
Further, in another embodiment of the product information pushing method, the step of inputting the dimension reduction information into a preset random forest algorithm model to classify the clients and obtain a target classification result includes:
Step C1, inputting the dimension reduction information into a preset random forest algorithm model, and obtaining classification of different decision trees in the preset random forest algorithm model on the dimension reduction information to obtain different initial target classification results;
in this embodiment, before the target classification result is obtained, the classification result of the dimension reduction information by different decision trees in the random forest algorithm model is determined, and the classification result of the dimension reduction information by different decision trees may be different, for example, the classification result of the dimension reduction information by the decision tree a is 0, and the classification result of the dimension reduction information by the decision tree b is 1.
And C2, determining a target classification result according to the classification duty ratio of different initial target classification results.
According to the classification duty ratio of different initial target classification results, determining a target classification node, wherein if the classification result of the dimension reduction information is 0, the duty ratio of the dimension reduction information is 80%, the classification result of the dimension reduction information is 1, the duty ratio of the dimension reduction information is 20%, and determining the target classification result to be 0.
In this embodiment, the dimension reduction information is input into a preset random forest algorithm model to obtain classification of different decision trees in the preset random forest algorithm model on the dimension reduction information, so as to obtain different initial target classification results; and determining the target classification result according to the classification duty ratio of different initial target classification results. And the classification accuracy is improved.
Referring to fig. 3, fig. 3 is a schematic device structure diagram of a hardware running environment according to an embodiment of the present invention.
The product information pushing device of the embodiment of the invention can be a PC, or can be terminal devices such as a smart phone, a tablet personal computer, an electronic book reader, an MP3 (Moving Picture Experts Group Audio Layer III, dynamic image expert compression standard audio layer 3) player, an MP4 (Moving Picture Experts Group Audio Layer IV, dynamic image expert compression standard audio layer 3) player, a portable computer and the like.
As shown in fig. 3, the product information pushing apparatus may include: a processor 1001, such as a CPU, memory 1005, and a communication bus 1002. Wherein a communication bus 1002 is used to enable connected communication between the processor 1001 and a memory 1005. The memory 1005 may be a high-speed RAM memory or a stable memory (non-volatile memory), such as a disk memory. The memory 1005 may also optionally be a storage device separate from the processor 1001 described above.
Optionally, the product information pushing device may further include a target user interface, a network interface, a camera, an RF (Radio Frequency) circuit, a sensor, an audio circuit, a WiFi module, and so on. The target user interface may comprise a Display screen (Display), an input unit such as a Keyboard (Keyboard), and the selectable target user interface may further comprise a standard wired interface, a wireless interface. The network interface may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface).
It will be appreciated by those skilled in the art that the product information pushing device structure shown in fig. 3 does not constitute a limitation of the product information pushing device, and may include more or fewer components than shown, or may combine certain components, or may be arranged in different components.
As shown in fig. 3, an operating system, a network communication module, and a product information push program may be included in the memory 1005, which is one type of computer storage medium. The operating system is a program that manages and controls the hardware and software resources of the product information pushing device, supporting the operation of the product information pushing program and other software and/or programs. The network communication module is used to implement communication between components within the memory 1005 and other hardware and software in the product information pushing device.
In the product information pushing device shown in fig. 3, the processor 1001 is configured to execute a product information pushing program stored in the memory 1005, to implement the steps of the product information pushing method described in any one of the above.
The specific implementation of the product information pushing device of the present invention is basically the same as the embodiments of the product information pushing method described above, and will not be described herein again.
The invention also provides a product information pushing device, which comprises:
The collection module is used for collecting the client information of the clients received in the historical preset time period;
the dimension reduction module is used for carrying out dimension reduction processing on the client information based on a preset manifold learning algorithm model to obtain dimension reduction information;
the classification module is used for inputting the dimension reduction information into a preset random forest algorithm model so as to classify the clients and obtain a target classification result;
the first acquisition module is used for acquiring the association relation between the preset client type and different product information, and directionally pushing the corresponding product information to each client according to the association relation and the target classification result.
The specific implementation of the product information pushing device is basically the same as the above embodiments of the product information pushing method, and will not be described herein again.
The present invention provides a readable storage medium storing one or more programs executable by one or more processors for implementing the steps of the product information pushing method of any one of the above.
The specific implementation manner of the readable storage medium of the present invention is basically the same as that of each embodiment of the product information pushing method, and will not be described herein again.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein, or any application, directly or indirectly, within the scope of the invention.
Claims (7)
1. The product information pushing method is characterized by comprising the following steps of:
collecting the information of each client received in a history preset time period;
acquiring a preset number of neighbor samples of each piece of customer information from a preset manifold learning algorithm model, and acquiring sample information corresponding to the neighbor samples and a local reconstruction weight matrix of each piece of customer information;
determining a low-dimensional matrix of the client information according to the local reconstruction weight matrix and the sample information so as to obtain dimension reduction information according to the low-dimensional matrix;
inputting the dimension reduction information into a preset random forest algorithm model to classify each client and obtain a target classification result;
acquiring association relations between preset client types and different product information, and directionally pushing corresponding product information to each client according to the association relations and the target classification result;
The step of inputting the dimension reduction information into a preset random forest algorithm model to classify the clients and obtaining a target classification result comprises the following steps:
acquiring preset use cases and each sub-dimension of the dimension reduction information;
carrying out random selection of different numbers of sub-dimensions with different times from the dimension reduction information so as to construct different decision tree clusters, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random;
the decision tree cluster generation process comprises the following steps: selecting N sub-dimensions from all the N sub-dimensions of the dimension reduction information, wherein N is smaller than N, randomly determining the decision influence degree of the N sub-dimensions, and obtaining a decision tree based on the decision influence degree of the N sub-dimensions and the decision influence degree of the N sub-dimensions, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random, establishing m decision trees, forming random forests by the m decision trees, and searching the number of each decision tree by a grid to obtain each decision tree cluster;
setting the use cases with preset proportion in the use cases as first use cases, and setting other use cases with first exception as second use cases;
The first example is used as a training example to carry out adjustment training of decision influence degree in different decision tree clusters, the second example is used as a test example to carry out testing of the decision tree clusters, so that an optimal target decision tree cluster with testing error smaller than a preset error value is selected from all decision tree clusters, and the target decision tree cluster is set as a preset random forest algorithm model;
the step of obtaining a preset number of neighbor samples of each piece of customer information from a preset manifold learning algorithm model and obtaining sample information corresponding to the neighbor samples and a local reconstruction weight matrix of each piece of customer information comprises the following steps:
acquiring each Euclidean distance of each piece of customer information and sample information corresponding to each sample in the manifold learning algorithm model according to a preset Euclidean distance algorithm;
acquiring a preset number of neighbor samples of each piece of customer information according to the size of each Euclidean distance;
determining a weight coefficient vector of each piece of customer information and each piece of sample information according to a preset local covariance algorithm; assuming that the neighbor samples of the client information xi are (xi 1, xi2, xik), a preset local protocol is acquired Variance algorithm, find local covariance matrix zi= (xi-xj) T (xi-xj) based on the local covariance matrix zi= (xi-xj) T (xi-xj), wherein xj represents client information different from xi;
and determining the local reconstruction weight matrix of the sample information corresponding to the neighbor sample and each piece of client information according to the weight coefficient vector and the association relation between the preset weight coefficient vector and the local reconstruction weight matrix.
2. The product information pushing method as claimed in claim 1, wherein the step of obtaining a preset number of neighbor samples of each piece of customer information from a preset manifold learning algorithm model includes:
acquiring the name of each sub-dimension of the client information, and determining the integration sequence of the client information dimension integration according to the name of each sub-dimension, wherein the integration sequence comprises the sequence determination in an alphabet according to the first letter of the name of each sub-dimension;
and dimension integration is carried out on the client information according to the integration sequence, so that integrated client information is obtained.
3. The product information pushing method as claimed in claim 1, wherein the step of inputting the dimension reduction information into a preset random forest algorithm model to classify the clients and obtain the target classification result comprises:
Inputting the dimension reduction information into a preset random forest algorithm model, and acquiring classification of different decision trees in the preset random forest algorithm model on the dimension reduction information to obtain different initial target classification results;
and determining the target classification result according to the classification duty ratio of different initial target classification results.
4. A product information pushing device, characterized in that the product information pushing device comprises:
the collection module is used for collecting the customer information received in the history preset time period;
the first acquisition module is used for acquiring a preset number of neighbor samples of each piece of client information from a preset manifold learning algorithm model, and acquiring sample information corresponding to the neighbor samples and a local reconstruction weight matrix of each piece of client information;
the second acquisition module is used for determining a low-dimensional matrix of the client information according to the local reconstruction weight matrix and the sample information so as to obtain dimension reduction information according to the low-dimensional matrix;
the classification module is used for inputting the dimension reduction information into a preset random forest algorithm model so as to classify the clients and obtain a target classification result;
the third acquisition module is used for acquiring the association relation between the preset client type and different product information, and directionally pushing the corresponding product information to each client according to the association relation and the target classification result;
The product information pushing device is used for realizing:
acquiring preset use cases and each sub-dimension of the dimension reduction information;
carrying out random selection of different numbers of sub-dimensions with different times from the dimension reduction information so as to construct different decision tree clusters, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random;
the decision tree cluster generation process comprises the following steps: selecting N sub-dimensions from all the N sub-dimensions of the dimension reduction information, wherein N is smaller than N, randomly determining the decision influence degree of the N sub-dimensions, and obtaining a decision tree based on the decision influence degree of the N sub-dimensions and the decision influence degree of the N sub-dimensions, wherein the decision influence degree of the sub-dimensions on the nodes of each decision tree in all the decision tree clusters is random, establishing m decision trees, forming random forests by the m decision trees, and searching the number of each decision tree by a grid to obtain each decision tree cluster;
setting the use cases with preset proportion in the use cases as first use cases, and setting other use cases with first exception as second use cases;
the first example is used as a training example to carry out adjustment training of decision influence degree in different decision tree clusters, the second example is used as a test example to carry out testing of the decision tree clusters, so that an optimal target decision tree cluster with testing error smaller than a preset error value is selected from all decision tree clusters, and the target decision tree cluster is set as a preset random forest algorithm model;
The first acquisition module includes:
the first acquisition unit is used for acquiring each Euclidean distance of each piece of customer information and sample information corresponding to each sample in the manifold learning algorithm model according to a preset Euclidean distance algorithm;
the second acquisition unit is used for acquiring a preset number of neighbor samples of each piece of customer information according to the size of each Euclidean distance;
the first determining unit is used for determining a weight coefficient vector of each piece of client information and each piece of sample information according to a preset local covariance algorithm; assuming that the neighbor samples of the client information xi are (xi 1, xi2,., xik), a preset local covariance algorithm is obtained, and a local covariance matrix zi= (xi-xj) is obtained T (xi-xj) based on the local covariance matrix zi= (xi-xj) T (xi-xj), wherein xj represents client information different from xi;
and the second determining unit is used for determining the sample information corresponding to the neighbor samples and the local reconstruction weight matrix of each piece of client information according to the weight coefficient vector and the association relation between the weight coefficient vector and the preset weight coefficient vector and the local reconstruction weight matrix.
5. The product information pushing apparatus of claim 4, wherein the product information pushing apparatus comprises:
A fourth obtaining module, configured to obtain a name of each sub-dimension of the client information, and determine an integration sequence of dimension integration of the client information according to the name of each sub-dimension, where the integration sequence includes determining according to a sequence of initial letters of the name of each sub-dimension in an alphabet;
and the integration module is used for carrying out dimension integration on the client information according to the integration sequence to obtain integrated client information.
6. A product information pushing apparatus, characterized in that the product information pushing apparatus comprises: a memory, a processor, a communication bus, and a product information push program stored on the memory,
the communication bus is used for realizing communication connection between the processor and the memory;
the processor is configured to execute the product information pushing program to implement the steps of the product information pushing method as claimed in any one of claims 1 to 3.
7. A readable storage medium, characterized in that the readable storage medium has stored thereon a product information pushing program, which when executed by a processor, implements the steps of the product information pushing method according to any one of claims 1 to 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910847224.9A CN110717806B (en) | 2019-09-06 | 2019-09-06 | Product information pushing method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910847224.9A CN110717806B (en) | 2019-09-06 | 2019-09-06 | Product information pushing method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110717806A CN110717806A (en) | 2020-01-21 |
| CN110717806B true CN110717806B (en) | 2024-04-02 |
Family
ID=69209738
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910847224.9A Active CN110717806B (en) | 2019-09-06 | 2019-09-06 | Product information pushing method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110717806B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111371853B (en) * | 2020-02-24 | 2022-12-20 | 中国银联股份有限公司 | Resource information pushing method, device, server and storage medium |
| CN111353860A (en) * | 2020-03-30 | 2020-06-30 | 中国建设银行股份有限公司 | Product information pushing method and system |
| CN112016098B (en) * | 2020-09-01 | 2023-09-08 | 中国平安财产保险股份有限公司 | Processing method and device of scanning result, computer equipment and storage medium |
| CN114266601A (en) * | 2021-12-24 | 2022-04-01 | 深圳前海微众银行股份有限公司 | Marketing strategy determination method and device, terminal equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105653670A (en) * | 2015-12-29 | 2016-06-08 | 中国电力科学研究院 | Intelligent electricity consumption data mining method based on manifold learning clustering algorithm |
| EP3048563A1 (en) * | 2015-01-22 | 2016-07-27 | Alcatel Lucent | Method and system for incremental manifold learning |
| CN106503205A (en) * | 2016-10-25 | 2017-03-15 | 四川大学 | A kind of song recommendations method based on non-linearity manifold study |
| CN109409923A (en) * | 2018-09-03 | 2019-03-01 | 平安普惠企业管理有限公司 | Distribution method, computer readable storage medium and the terminal device of sales region |
| CN110163657A (en) * | 2019-04-12 | 2019-08-23 | 中国平安财产保险股份有限公司 | Insurance products recommended method and relevant device based on manifold learning arithmetic |
-
2019
- 2019-09-06 CN CN201910847224.9A patent/CN110717806B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3048563A1 (en) * | 2015-01-22 | 2016-07-27 | Alcatel Lucent | Method and system for incremental manifold learning |
| CN105653670A (en) * | 2015-12-29 | 2016-06-08 | 中国电力科学研究院 | Intelligent electricity consumption data mining method based on manifold learning clustering algorithm |
| CN106503205A (en) * | 2016-10-25 | 2017-03-15 | 四川大学 | A kind of song recommendations method based on non-linearity manifold study |
| CN109409923A (en) * | 2018-09-03 | 2019-03-01 | 平安普惠企业管理有限公司 | Distribution method, computer readable storage medium and the terminal device of sales region |
| CN110163657A (en) * | 2019-04-12 | 2019-08-23 | 中国平安财产保险股份有限公司 | Insurance products recommended method and relevant device based on manifold learning arithmetic |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110717806A (en) | 2020-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110717806B (en) | Product information pushing method, device, equipment and storage medium | |
| CN109902708B (en) | Recommendation model training method and related device | |
| US11526799B2 (en) | Identification and application of hyperparameters for machine learning | |
| CN107153971B (en) | Method and device for identifying equipment cheating in APP popularization | |
| CN109697629B (en) | Product data pushing method and device, storage medium and computer equipment | |
| Libungan et al. | ShapeR: an R package to study otolith shape variation among fish populations | |
| CN109192242A (en) | Based on the microbial diversity interaction analysis system and method for calculating cloud platform | |
| CN106709318B (en) | A kind of recognition methods of user equipment uniqueness, device and calculate equipment | |
| CN110008397B (en) | Recommendation model training method and device | |
| US11562179B2 (en) | Artificial intelligence system for inspecting image reliability | |
| Maitra et al. | Bootstrapping for significance of compact clusters in multidimensional datasets | |
| CN111797927A (en) | Method and system for determining important features of machine learning samples | |
| JPWO2009031297A1 (en) | Image search device, image classification device and method, and program | |
| CN110059212A (en) | Image search method, device, equipment and computer readable storage medium | |
| CN110378516B (en) | Analyst portrait generation method, analyst portrait generation device, analyst portrait generation equipment and computer-readable storage medium | |
| CN112990318A (en) | Continuous learning method, device, terminal and storage medium | |
| CN114692889A (en) | Meta-feature training model for machine learning algorithm | |
| CN116894721A (en) | Index prediction method and device and computer equipment | |
| US20160357708A1 (en) | Data analysis method, data analysis apparatus, and recording medium having recorded program | |
| CN112783468A (en) | Target object sorting method and device | |
| CN113297411B (en) | Method, device and equipment for measuring similarity of wheel-shaped atlas and storage medium | |
| CN109993165A (en) | The identification of tablet plate medicine name and tablet plate information acquisition method, device and system | |
| CN111309608B (en) | Test case selection method and device, electronic equipment and readable storage medium | |
| US20170255691A1 (en) | Information processing system, information processing method, and program | |
| CN108198106A (en) | Job issuing method, equipment and computer readable storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |