Business opportunity recommendation method and device for B-type enterprises, equipment, medium and product

Abstract

The invention discloses a business opportunity recommendation method and device for B-type enterprises, equipment, a medium and a product, and relates to the field of data processing. The business opportunity recommendation method for B-type enterprises comprises the following steps: acquiring association relationship information, industry and commerce information and annual report information of each B-type enterprise in a customer relationship management system; obtaining product data of N historical bidding items with determined bidding results; performing feature extraction on the data to generate a feature set, and constructing a sample data set; and training a nonlinear prediction model based on the sample data set, predicting the bid reference probability of the enterprise for each target bid invitation project, and carrying out project recommendation. According to the application, the association relationship information, the industry and commerce information, the annual report information and the product data of the bid invitation project are used for training and learning, key customers and potential opportunities are mined, the nonlinear prediction model is constructed, then the bid invitation business opportunity is analyzed and predicted, the bid parametric probability is obtained, the business opportunity recommendation for B-type enterprises is given, and the business opportunity recommendation efficiency is improved. And comprehensive mining and quantitative analysis of business opportunities are realized.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a method, device, equipment, medium and product for recommending business opportunities for Class B enterprises. Background Art

[0002] Business opportunity is the abbreviation of "commercial opportunity". It represents the possibility of realizing market value by integrating resources and meeting market demand. Mature business opportunity forecasting can help companies quickly screen out the most likely business opportunities from a large amount of information, and pay more targeted attention to these more mature business opportunities, thereby acquiring project opportunities more efficiently.

[0003] Existing business opportunity analysis is mostly based on experience, which is a qualitative analysis rather than a quantitative analysis, and requires business personnel to invest a lot of energy in research. At the same time, the basis for judgment mostly relies on internal data such as net profit margin, project amount, etc., and rarely combines external information (such as publicly disclosed big data, such as historical bidding project information of related companies, project information of competing companies, etc.), which makes the judgment of business opportunities relatively one-sided. Summary of the invention

[0004] The purpose of this application is to provide a business opportunity recommendation method, device, equipment, medium and product for Class B enterprises, which can realize comprehensive mining and quantitative analysis of business opportunities.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a business opportunity recommendation method for Class B enterprises, the business opportunity recommendation method for Class B enterprises comprising:

[0007] Obtain the relationship information, industrial and commercial information and annual report information of each Class B enterprise in the customer relationship management system;

[0008] Obtain product data for N historical bidding projects whose bidding results have been determined;

[0009] Extract features from the association information, industrial and commercial information, and annual report information of each Class B enterprise in the customer relationship management system and the product data of N historical bidding projects, and generate the enterprise feature set of each Class B enterprise in the customer relationship management system and the product feature set of each historical bidding project;

[0010] Construct a sample data set; the sample data set includes M×N sample data, the input features of the sample data are the enterprise feature set and the product feature set, and the label of the sample data is the enterprise's bidding results for historical bidding projects; the bidding results include bidding and winning, bidding but not winning, and not bidding; M is the number of Class B enterprises in the customer relationship management system;

[0011] Training a nonlinear prediction model based on the sample data set to obtain a trained nonlinear prediction model;

[0012] The trained nonlinear prediction model is used to predict the probability of each Class B enterprise in the customer relationship management system to bid for each target bidding project; the target bidding project is a bidding project for which the bidding information has been made public but no bid has been made;

[0013] Recommend target bidding projects to each Class B enterprise in the customer relationship management system based on the bidding probability.

[0014] Optionally, feature extraction is performed on the association relationship information, industrial and commercial information, and annual report information of each Class B enterprise in the customer relationship management system and the product data of N historical bidding projects to generate an enterprise feature set of each Class B enterprise in the customer relationship management system and a product feature set of each historical bidding project, specifically including:

[0015] Using the first zero-shot classifier, the type of each Class B enterprise in the customer relationship management system is determined according to the association information, business information and annual report information of each Class B enterprise in the customer relationship management system;

[0016] Using the second zero-shot classifier, the industry type of each historical bidding project is determined according to the product data of each historical bidding project;

[0017] The LM model is used to extract keywords from the association information, industrial and commercial information, and annual report information of each Class B enterprise in the customer relationship management system to obtain a high-dimensional feature set of each Class B enterprise in the customer relationship management system;

[0018] Perform word frequency statistics and stop word library processing on the high-dimensional feature sets of each Class B enterprise in the customer relationship management system to obtain basic information of each Class B enterprise in the customer relationship management system;

[0019] Generate an enterprise feature set for each Class B enterprise in the customer relationship management system based on the type and basic information of each Class B enterprise in the customer relationship management system;

[0020] The LM model is used to extract keywords from the product data of each historical bidding project to obtain a high-dimensional feature set of each historical bidding project;

[0021] Perform word frequency statistics and stop word library processing on the high-dimensional feature sets of each historical bidding project to obtain the basic information of each historical bidding project;

[0022] According to the industry type of each historical bidding project and the basic information of each historical bidding project, a product feature set of each historical bidding project is generated.

[0023] Optionally, the nonlinear prediction model includes a convolutional neural network and a recurrent neural network;

[0024] The output end of the convolutional neural network is connected to the input end of the recurrent neural network.

[0025] Optionally, training a nonlinear prediction model based on the sample data set to obtain a trained nonlinear prediction model specifically includes:

[0026] The nonlinear prediction model is trained using the sample data set so that the loss function is less than a first threshold value, thereby obtaining a trained nonlinear prediction model;

[0027] The trained nonlinear prediction model is tested using a sample data set, and sample data with inaccurate test results are obtained to form a secondary training data set;

[0028] A cost-sensitive loss function is introduced, and the trained nonlinear prediction model is retrained using a secondary training data set, so that the cost-sensitive loss function is less than a second threshold, thereby obtaining a trained nonlinear prediction model.

[0029] Optionally, the cost-sensitive loss function is:

[0030] Loss1(x,i)=∑ j P(j|x)c(i,j);

[0031] Among them, Loss1(x,i) represents the cost-sensitive loss when the reference result of the nonlinear prediction model corresponding to the input feature x is i, P(j|x) represents the posterior probability that the reference result of the nonlinear prediction model corresponding to the input feature x is j, and c(i,j) is the cost loss weight factor of reference result i relative to reference result j.

[0032] Optionally, the loss function is:

[0033] Loss2(x,)=∑ j P(j|x);

[0034] Among them, Loss2(x,i) represents the loss when the parameter result of the nonlinear prediction model corresponding to the input feature x is i, and P(j|x) represents the posterior probability that the parameter result of the nonlinear prediction model corresponding to the input feature x is j.

[0035] In a second aspect, the present application provides a business opportunity recommendation device for Class B enterprises, wherein the business opportunity recommendation device for Class B enterprises applies the above-mentioned business opportunity recommendation method for Class B enterprises, and the business opportunity recommendation device for Class B enterprises comprises:

[0036] The enterprise information acquisition module is used to obtain the relationship information, industrial and commercial information and annual report information of each Class B enterprise in the customer relationship management system;

[0037] A product data acquisition module is used to acquire product data of N historical bidding projects whose bidding results have been determined;

[0038] The feature extraction module is used to extract features from the association relationship information, industrial and commercial information and annual report information of each Class B enterprise in the customer relationship management system and the product data of N historical bidding projects, and generate the enterprise feature set of each Class B enterprise in the customer relationship management system and the product feature set of each historical bidding project;

[0039] A sample data set construction module is used to construct a sample data set; the sample data set includes M×N sample data, the input features of the sample data are the enterprise feature set and the product feature set, and the labels of the sample data are the bidding results of the enterprise for historical bidding projects; the bidding results include bidding and winning, bidding but not winning, and not bidding; M is the number of Class B enterprises in the customer relationship management system;

[0040] A model training module, used for training a nonlinear prediction model based on the sample data set to obtain a trained nonlinear prediction model;

[0041] The prediction module is used to use the trained nonlinear prediction model to predict the probability of each Class B enterprise in the customer relationship management system to bid for each target bidding project; the target bidding project is a bidding project for which the bidding information has been made public but no bid has been made;

[0042] The recommendation module is used to recommend target bidding projects to each Class B enterprise in the customer relationship management system based on the bidding probability.

[0043] In a third aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned business opportunity recommendation method for Class B enterprises.

[0044] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned business opportunity recommendation method for Class B enterprises.

[0045] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned business opportunity recommendation method for Class B enterprises.

[0046] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0047] This application provides a method, device, equipment, medium and product for recommending business opportunities for Class B enterprises. This application uses association information, industrial and commercial information, annual report information and product data of bidding projects for training and learning, mines key customers and potential opportunities, builds a nonlinear prediction model, and then analyzes and predicts bidding opportunities, obtains the probability of bidding, and gives business opportunity recommendations for Class B enterprises, realizing comprehensive mining and quantitative analysis of business opportunities. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 A flowchart of a business opportunity recommendation method for Class B enterprises provided in one embodiment of the present application;

[0050] Figure 2 A schematic diagram of a business opportunity recommendation method for Class B enterprises provided in an embodiment of the present application;

[0051] Figure 3 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0053] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0054] In an exemplary embodiment, Figure 1 and Figure 2As shown, a method for recommending business opportunities for Class B enterprises is provided. The method for recommending business opportunities for Class B enterprises achieves the purpose of mining and recommending business opportunities for Class B enterprises (Class B enterprises are system integrators or operators, or large general contractors. They have sufficient bidding qualifications or many project integration experience companies in large projects. They cooperate with such enterprises to encourage system integrators or operators to select company products and then deliver them to the users of the products, i.e., end users. Class B enterprises are relative to Class A enterprises. Class A customers refer to the users of the products, i.e., end customers. In the bidding process, they generally correspond to the bidding enterprises (non-agents), direct demand parties) by quantitatively analyzing CRM (Customer Relationship Management) databases, open source bidding data, and enterprise data. The method includes the following steps 101 to 107. Among them:

[0055] Step 101, obtaining the relationship information, business information and annual report information of each Class B enterprise in the customer relationship management system.

[0056] Step 102, obtaining product data of N historical bidding projects whose bidding results have been determined.

[0057] Step 103, extracting features from the association information, business information and annual report information of each Class B enterprise in the customer relationship management system and the product data of N historical bidding projects, and generating an enterprise feature set of each Class B enterprise in the customer relationship management system and a product feature set of each historical bidding project.

[0058] Step 104, construct a sample data set; the sample data set includes M×N sample data, the input features of the sample data are the enterprise feature set and the product feature set, and the labels of the sample data are the bidding results of the enterprise for historical bidding projects; the bidding results include bidding and winning, bidding but not winning, and not bidding; M is the number of Class B enterprises in the customer relationship management system.

[0059] Step 105, training a nonlinear prediction model based on the sample data set to obtain a trained nonlinear prediction model;

[0060] Step 106, using the trained nonlinear prediction model, predicting the probability of each Class B enterprise in the customer relationship management system participating in the bidding for each target bidding project; the target bidding project is a bidding project for which the bidding information has been made public but no bid has been made.

[0061] Step 107, recommending target bidding projects to each Class B enterprise in the customer relationship management system according to the bidding probability.

[0062] Implement the above steps 101 to 107, use machine learning methods to use structured CRM data, open source bidding data, and enterprise data for training and learning, dig out key customers and potential opportunities, build a unique enterprise customer model, and then analyze and predict bidding opportunities, and give business opportunity recommendations for Class B companies. Then use external input data to continuously self-learn and optimize the model, so that the target customer participation rate can be analyzed through model prediction and initial business opportunity information at the beginning of the establishment of the business opportunity, and seize the opportunity in the early stage of the project or even in the early stage, so as to shorten the business opportunity tracking chain. Further, this application uses traditional mathematical statistical models and machine learning algorithms to learn the features of business opportunity information contained in the customer relationship management system, and obtains open source corresponding enterprise bidding information and industrial and commercial data, etc. to perform feature learning, and then predict the corresponding enterprise participation rate of new business opportunities such as bidding information and procurement intentions, and give business opportunity recommendations for Class B companies. Potential business opportunities here include direct project investment, direct procurement, and business opportunities for cooperative projects with the help of intermediaries.

[0063] In another exemplary embodiment, a specific implementation of step 101 and step 102 includes step 1.

[0064] Step 1: Processing bidding and enterprise data. Obtain multi-source data based on public bidding data sources, including internal CRM data, bidding project product data, enterprise data such as industrial and commercial information, annual report data, etc., and perform data screening, outlier processing and standardization on multi-source data.

[0065] In another exemplary embodiment, a specific implementation of step 103 includes step 2.

[0066] Step 2: Integrate multi-source data to obtain a Class B enterprise portrait model. Use the zero-shot model and LM model to extract features and classify labels for the data, build a keyword feature set, and analyze the basic statistical relationship between data. Specifically, based on the deep learning model, build a zero-shot classifier, input project information from different data sources into the zero-shot classifier, and give an algorithm label; use the LM model to extract and clean key words and sentences, perform word frequency statistics, and build a basic project knowledge base and enterprise feature vector set. The algorithm label is a classification label that classifies information according to analysis requirements, which is convenient for subsequent search for quantitative relationships between indicator categories. For example: classify the product content corresponding to the project by industry, such as the medical industry, tourism industry, media industry, etc.

[0067] Exemplarily, step 103 specifically includes:

[0068] The first zero-shot classifier is used to determine the type of each Class B enterprise in the customer relationship management system according to the association information, business information and annual report information of each Class B enterprise in the customer relationship management system.

[0069] The second zero-shot classifier is used to determine the industry type of each historical bidding project based on the product data of each historical bidding project.

[0070] The LM model is used to extract keywords from the association information, industrial and commercial information, and annual report information of each Class B enterprise in the customer relationship management system to obtain a high-dimensional feature set of each Class B enterprise in the customer relationship management system.

[0071] The high-dimensional feature sets of each Class B enterprise in the customer relationship management system are respectively subjected to word frequency statistics and stop word library processing to obtain the basic information of each Class B enterprise in the customer relationship management system.

[0072] According to the type and basic information of each Class B enterprise in the customer relationship management system, an enterprise feature set of each Class B enterprise in the customer relationship management system is generated.

[0073] The LM model is used to extract keywords from the product data of each historical bidding project to obtain a high-dimensional feature set of each historical bidding project.

[0074] The high-dimensional feature sets of each historical bidding project are subjected to word frequency statistics and stop word library processing to obtain the basic information of each historical bidding project.

[0075] According to the industry type of each historical bidding project and the basic information of each historical bidding project, a product feature set of each historical bidding project is generated.

[0076] Among them, the LM model is used to extract keywords from the obtained information: perform indicator analysis on the obtained bidding information and enterprise information, summarize the classification standards such as product information and industry information, and use the LM model to extract corresponding key points of the information to obtain a high-dimensional feature set.

[0077] The high-dimensional feature set obtained above is used for word frequency statistics, and stop word library processing such as modal particles and quantifiers is added to build a basic knowledge base of information for each Class B enterprise as the main reference set for subsequent predictions.

[0078] In another exemplary embodiment, a specific implementation of step 104 includes steps 3 to 5.

[0079] Step 3: Initial weighting of influencing factors analysis. Based on statistical regression model and association rule correlation analysis, a feature set of influencing factors and initial weights for Class B enterprise bidding are constructed. Specifically, it includes Step 3-1 and Step 3-2:

[0080] Step 3-1, use the association rule algorithm to determine the related items, call the apriori algorithm, determine the frequent subsets that affect the bidding results, and obtain the feature indicator set after preliminary dimensionality reduction.

[0081] Input the feature vector set of the winning projects cleaned in step 2.

[0082] Scan the relevant classification indicators in each feature vector, perform count comparison, and generate frequent item sets. Then perform link screening to generate candidate sets, that is, feature indicators with a high degree of influence.

[0083] The principle of generating frequent itemsets is as follows: for each vector set L to be analyzed, generate all its non-empty subsets S. For each non-empty subset S, if P(L) / P(S)≥min_conf, then output the subset in L excluding S, where min_conf is the minimum confidence threshold.

[0084] Step 3-2, use logistic regression based on multivariate linear statistical theory to conduct preliminary impact analysis and obtain variable impact factors. Take the historical bidding results as the dependent variable, the winning bid is 1, the unsuccessful bid is set to 0, and the bid review (i.e., no bid) is set to 0.5. The independent variable is the feature index set obtained after the above dimensionality reduction, and the quantitative feature set X is obtained, which is the basic model of enterprise portrait. Output X and the impact factor are the initial weight values ​​for the subsequent nonlinear conversion.

[0085] Step 4: nonlinear prediction model training simulation. The feature set obtained in step 3 is processed by learning, and the project bid participation rate is predicted by using keyword matching similarity learning and neural network nonlinear learning.

[0086] In the embodiment of the present application, a convolutional neural network CNN and a recurrent neural network RNN ​​are selected to construct a nonlinear prediction model, and an activation function is used to introduce nonlinearity to predict the winning rate of project business opportunities for Company B.

[0087] By adding a data fusion layer through weighted fusion, different information is merged together; the CNN structure is used for expansion, feature learning is performed, static weights and outputs are obtained, and then the results are extracted for RNN to perform dynamic prediction and process new project data in real time. That is, the CNN fully connected layer outputs the learned weight features, and then enters the RNN Concatenate layer to merge together to form a longer vector, and Attention is multiplied (merging process).

[0088] The feature set X from step 3 is encoded into an input feature vector, and the influence factor obtained is the initial training weight. The loss function is defined to quantify the gap between the neural network output and the true state. Specifically:

[0089] The weights and biases of the nonlinear prediction model are updated via the back-propagation algorithm.

[0090] The nonlinear prediction model is trained iteratively until the loss function reaches a satisfactory level and the fusion result is output.

[0091] Among them, the prediction result is the bid participation probability, which is similar to the regression result setting in step 3, and the bid participation probability of new project business opportunities is predicted.

[0092] Among them, the activation function is adjusted according to the model test using the hyperbolic tangent function, and the fitting result is the best.

[0093] Step 5, test data backtest optimization, based on the prediction results of step 4 and the actual observation data, the model is evaluated and the nonlinear prediction model is optimized. The model evaluation criteria are designed according to the needs of the enterprise, and then the difficult case mining screening strategy is used to correct the data set quantity ratio, change the weights that affect the feature ratio to obtain a data-biased backtest data set, perform feature analysis again, further optimize the learning rate, and reduce the loss value.

[0094] Step 5 backtests the data, extracts historical bidding contract data, performs model prediction, compares and analyzes the test results with the actual results, finds abnormal prediction values ​​for weighted process output, pre-marks abnormal weights, and subsequently corrects the screening threshold.

[0095] The cost-sensitive loss function is:

[0096] Loss1(x,i)=Σ j P(j|x)c(i,j)

[0097] Among them, Loss1(x,i) represents the cost-sensitive loss when the reference result of the nonlinear prediction model corresponding to the input feature x is i, P(j|x) represents the posterior probability that the reference result of the nonlinear prediction model corresponding to the input feature x is j, and c(i,j) is the cost loss weight factor of reference result i relative to reference result j.

[0098] In the historical loss judgment, the actual loss is not affected by the misjudgment method, that is, c(i,j) is considered to be 1. In the actual business process, especially for the model prediction of different Class B enterprises, the loss function actually has a specific direction bias. And this boundary condition design is also an important evidence for updating the classification of the model learning data set. After adding c(i,j), the original loss function Loss is converted from maximum likelihood estimation to maximum a posteriori estimation.

[0099] The loss function is:

[0100] Loss2(x,i)=Σ j P(j|x)

[0101] Among them, Loss2(x,i) represents the loss when the parameter result of the nonlinear prediction model corresponding to the input feature x is i.

[0102] Combined with B's business opportunity recommendation needs for Class B enterprises and the recommendation needs of different enterprises, different degrees of initial penalty values ​​are assigned according to different feature angles, which reduces the degree of false alarms and increases accuracy.

[0103] For example, the cost loss weight factor can be determined according to specific business needs and then slowly iterated linearly. In the initial business scenario, there will be a demand, and then in the case of prediction error correction, the value is linearly increased or decreased in the interval [-1, 1], and the step size is set to 0.1.

[0104] For example, after printing and comparing the preliminary prediction results of B, c(0,1) is updated to 0.7 and c(1,0) is updated to 0.3. Then, according to the weight features of the previous outliers, the classification algorithm is used to expand the same feature data set, and the corresponding feature data proportion weight is updated. This is used as a new preference for difficult examples to mine data, hoping to increase the sensitivity of the prediction model to related data, reduce the proportion of the original accurate data in the new data set, and construct a new preliminary test data set for backtesting.

[0105] In the above algorithm for screening new classified data sets, the top outlier annotation features are used as key classification features, the weights of enterprise feature vectors are adjusted, and the screening thresholds of difficult example data sets are continuously optimized. The test set is updated until the loss value is minimized, and the optimized prediction model is obtained.

[0106] In an exemplary embodiment, a business opportunity recommendation device for Class B enterprises is provided. The business opportunity recommendation device for Class B enterprises applies the above-mentioned business opportunity recommendation method for Class B enterprises. The business opportunity recommendation device for Class B enterprises includes:

[0107] The enterprise information acquisition module is used to obtain the relationship information, industrial and commercial information and annual report information of each Class B enterprise in the customer relationship management system.

[0108] The product data acquisition module is used to acquire product data of N historical bidding projects whose bidding results have been determined.

[0109] The feature extraction module is used to extract features from the relationship information, industrial and commercial information and annual report information of each Class B enterprise in the customer relationship management system and the product data of N historical bidding projects, and generate the enterprise feature set of each Class B enterprise in the customer relationship management system and the product feature set of each historical bidding project.

[0110] A sample data set construction module is used to construct a sample data set; the sample data set includes M×N sample data, the input features of the sample data are the enterprise feature set and the product feature set, and the labels of the sample data are the bidding results of the enterprise for historical bidding projects; the bidding results include bidding and winning, bidding but not winning, and not bidding; M is the number of Class B enterprises in the customer relationship management system.

[0111] The model training module is used to train a nonlinear prediction model based on the sample data set to obtain a trained nonlinear prediction model.

[0112] The prediction module is used to use the trained nonlinear prediction model to predict the probability of each Class B enterprise in the customer relationship management system to bid for each target bidding project; the target bidding project is a bidding project whose bidding information has been made public but no bid has been made.

[0113] The recommendation module is used to recommend target bidding projects to each Class B enterprise in the customer relationship management system based on the bidding probability.

[0114] The specific implementation method of this embodiment can be found in the above method embodiment, which will not be repeated here.

[0115] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data determined by complementarity. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a business opportunity recommendation method for Class B enterprises is implemented.

[0116] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0117] In an exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps in the above-mentioned business opportunity recommendation method embodiment for Class B enterprises are implemented.

[0118] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, which, when executed by a processor, implements the steps in the above-mentioned business opportunity recommendation method embodiment for Class B enterprises.

[0119] In an exemplary embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, implements the steps in the above-mentioned business opportunity recommendation method embodiment for Class B enterprises.

[0120] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0121] 1. This application is different from other business opportunity predictions that only analyze direct business opportunities. It focuses more on building an enterprise analysis model from external data. Therefore, the "winning rate" feature in the model results corresponds to the target Class B enterprise, and model optimization and updating also come from external data.

[0122] 2. Although the models involved in this application have a unified form, they are business opportunity prediction model clusters constructed for different enterprises based on their different characteristics and historical cooperation. That is, the model training parameters for each enterprise are different, and this result is also the core of the model-specific characteristics.

[0123] 3. This application aims to seek cooperative business opportunities, i.e. potential business opportunities, for B companies that do not directly have access to product information. LM and zero-shot obtain bidding data based on enterprise classification labels, extract keyword libraries, merge the information of data sources together, and use association rules and logical repurchase to capture the complex relationship between data.

[0124] In the nonlinear prediction simulation, based on the statistical model analysis, the project bid participation rate prediction is given by constructing a neural network model and keyword similarity matching.

[0125] The final simulation results are combined with the actual observed data to provide model evaluation values, correct model anomalies, and reflect the latest observed data. This forecasting process uses enterprise portraits to recommend business opportunities for Class B enterprises based on deep learning models.

[0126] 4. Due to the particularity of the data source, this model cannot adjust parameters from the internal enterprise perspective, so the learning level can only be adjusted from the external data optimization selection method. This application optimizes the screening ratio of the difficult case data set by adjusting the loss function, in order to increase the sensitivity of the prediction model to the corresponding features, and optimizes the new learning data set for retraining. This process is repeated until the loss value is minimized or tends to a stable state.

[0127] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0128] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, etc., but is not limited thereto.

[0129] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0130] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A business opportunity recommendation method for Class B enterprises, characterized in that: The business opportunity recommendation method for Class B enterprises includes: Obtain the relationship information, business information and annual report information of each Class B enterprise in the customer relationship management system; Class B enterprises refer to system integrators or operators; Obtain product data for N historical bidding projects whose bidding results have been determined; Extract features from the association information, industrial and commercial information, and annual report information of each Class B enterprise in the customer relationship management system and the product data of N historical bidding projects, and generate the enterprise feature set of each Class B enterprise in the customer relationship management system and the product feature set of each historical bidding project; Construct a sample data set; the sample data set includes M×N sample data, the input features of the sample data are the enterprise feature set and the product feature set, and the label of the sample data is the enterprise's bidding results for historical bidding projects; the bidding results include: bidding and winning, bidding but not winning, and not bidding; M is the number of Class B enterprises in the customer relationship management system; Training a nonlinear prediction model based on the sample data set to obtain a trained nonlinear prediction model; The trained nonlinear prediction model is used to predict the probability of each Class B enterprise in the customer relationship management system to bid for each target bidding project; the target bidding project is a bidding project for which the bidding information has been made public but no bid has been made; Recommend target bidding projects to each Class B enterprise in the customer relationship management system based on the bidding probability.

2. The business opportunity recommendation method for Class B enterprises according to claim 1, characterized in that: Feature extraction is performed on the association information, industrial and commercial information, and annual report information of each Class B enterprise in the customer relationship management system and the product data of N historical bidding projects to generate the enterprise feature set of each Class B enterprise in the customer relationship management system and the product feature set of each historical bidding project, including: Using the first zero-shot classifier, the type of each Class B enterprise in the customer relationship management system is determined according to the association information, business information and annual report information of each Class B enterprise in the customer relationship management system; Using the second zero-shot classifier, the industry type of each historical bidding project is determined according to the product data of each historical bidding project; The LM model is used to extract keywords from the association information, industrial and commercial information, and annual report information of each Class B enterprise in the customer relationship management system to obtain a high-dimensional feature set of each Class B enterprise in the customer relationship management system; Perform word frequency statistics and stop word library processing on the high-dimensional feature sets of each Class B enterprise in the customer relationship management system to obtain basic information of each Class B enterprise in the customer relationship management system; Generate an enterprise feature set for each Class B enterprise in the customer relationship management system based on the type and basic information of each Class B enterprise in the customer relationship management system; The LM model is used to extract keywords from the product data of each historical bidding project to obtain a high-dimensional feature set of each historical bidding project; Perform word frequency statistics and stop word library processing on the high-dimensional feature sets of each historical bidding project to obtain the basic information of each historical bidding project; According to the industry type of each historical bidding project and the basic information of each historical bidding project, a product feature set of each historical bidding project is generated.

3. The business opportunity recommendation method for Class B enterprises according to claim 1, characterized in that: The nonlinear prediction model includes a convolutional neural network and a recurrent neural network; The output end of the convolutional neural network is connected to the input end of the recurrent neural network.

4. The business opportunity recommendation method for Class B enterprises according to claim 1, characterized in that: Training a nonlinear prediction model based on the sample data set to obtain a trained nonlinear prediction model specifically includes: The nonlinear prediction model is trained using the sample data set so that the loss function is less than a first threshold value, thereby obtaining a trained nonlinear prediction model; The trained nonlinear prediction model is tested using a sample data set, and sample data with inaccurate test results are obtained to form a secondary training data set; A cost-sensitive loss function is introduced, and the trained nonlinear prediction model is retrained using a secondary training data set, so that the cost-sensitive loss function is less than a second threshold, thereby obtaining a trained nonlinear prediction model.

5. The business opportunity recommendation method for Class B enterprises according to claim 4, characterized in that: The cost-sensitive loss function is: Loss1(x,i)=∑ j P(j|x)c(i,j); Among them, Loss1(x,i) represents the cost-sensitive loss when the reference result of the nonlinear prediction model corresponding to the input feature x is i, P(j|x) represents the posterior probability that the reference result of the nonlinear prediction model corresponding to the input feature x is j, and c(i,j) is the cost loss weight factor of reference result i relative to reference result j.

6. The business opportunity recommendation method for Class B enterprises according to claim 4, characterized in that: The loss function is: Loss2(x,i)=∑ j P(j|x); Among them, Loss2(x,i) represents the loss when the parameter result of the nonlinear prediction model corresponding to the input feature x is i, and P(j|x) represents the posterior probability that the parameter result of the nonlinear prediction model corresponding to the input feature x is j.

7. A business opportunity recommendation device for Class B enterprises, characterized in that: The business opportunity recommendation device for Class B enterprises applies the business opportunity recommendation method for Class B enterprises described in any one of claims 1 to 6, and the business opportunity recommendation device for Class B enterprises comprises: The enterprise information acquisition module is used to obtain the relationship information, industrial and commercial information and annual report information of each Class B enterprise in the customer relationship management system; A product data acquisition module is used to acquire product data of N historical bidding projects whose bidding results have been determined; The feature extraction module is used to extract features from the association relationship information, industrial and commercial information and annual report information of each Class B enterprise in the customer relationship management system and the product data of N historical bidding projects, and generate the enterprise feature set of each Class B enterprise in the customer relationship management system and the product feature set of each historical bidding project; A sample data set construction module is used to construct a sample data set; the sample data set includes M×N sample data, the input features of the sample data are the enterprise feature set and the product feature set, and the labels of the sample data are the bidding results of the enterprise for historical bidding projects; the bidding results include bidding and winning, bidding but not winning, and not bidding; M is the number of Class B enterprises in the customer relationship management system; A model training module, used for training a nonlinear prediction model based on the sample data set to obtain a trained nonlinear prediction model; The prediction module is used to use the trained nonlinear prediction model to predict the probability of each Class B enterprise in the customer relationship management system to bid for each target bidding project; the target bidding project is a bidding project for which the bidding information has been made public but no bid has been made; The recommendation module is used to recommend target bidding projects to each Class B enterprise in the customer relationship management system based on the bidding probability.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the business opportunity recommendation method for Class B enterprises as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the business opportunity recommendation method for Class B enterprises described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the business opportunity recommendation method for Class B enterprises described in any one of claims 1 to 6 is implemented.

Images