Data recommendation method and device based on artificial intelligence, computer equipment and medium

Abstract

The invention provides a data recommendation method and device based on artificial intelligence, computer equipment and a storage medium. The method comprises the following steps: acquiring key business data of an initial crowd; analyzing the key business data based on an index analysis strategy to generate analysis reasons and optimization suggestions; performing data integration on the key business data, the analysis reasons, the optimization suggestions and the user behavior logs to obtain target data; based on the target data, performing crowd optimization processing on the initial crowd by using a preset large model, a diffusion model and a hybrid model to determine an optimized target crowd; performing effect prediction on the target crowd, and judging whether effect prediction data accords with expectation or not; if yes, binding the target crowd with a marketing strategy, and verifying the target crowd based on an AB experiment; and if the experiment result of the AB experiment is a forward result, performing product recommendation processing on the target crowd based on a total delivery strategy. According to the method, the marketing recommendation efficiency and accuracy are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, and in particular to a data recommendation method and device based on artificial intelligence, a computer device and a storage medium. BACKGROUND

[0002] In today's increasingly competitive market environment, targeted marketing, as the core means for enterprises to improve market share and efficiency, is becoming increasingly important. For example, in the process of determining the target audience of targeted marketing and carrying out marketing recommendations, traditional experience and simple data analysis methods are currently widely used. This traditional approach has obvious shortcomings. On the one hand, it lacks systematicity and scientificity, with fragmented data processing procedures and no unified standards. On the other hand, it is inefficient and difficult to quickly respond to market changes and customer needs.

[0003] Due to the above shortcomings, enterprises face great challenges in accurately positioning target customers, and a large amount of marketing resources is scattered to non-potential customer groups, resulting in resource waste. At the same time, the accuracy of marketing recommendations is low, which cannot effectively meet the actual needs of target customers, making it difficult for marketing activities to achieve the desired results, and restricting the expansion of the market and the improvement of efficiency.

[0004] Therefore, there is an urgent need for a more scientific, efficient and accurate method for determining targeted marketing target audiences and marketing recommendations. SUMMARY

[0005] The main purpose of the present application is to provide a data recommendation method and device based on artificial intelligence, a computer device and a storage medium, aiming to solve the technical problems of low processing efficiency and low accuracy of marketing recommendations in the prior art.

[0006] To achieve the above purpose, the present application provides a data recommendation method based on artificial intelligence, which comprises: determining an initial audience corresponding to a business target, and collecting key business data corresponding to the initial audience; analyzing and processing the key business data based on a preset index analysis strategy, generating corresponding analysis reasons and optimization suggestions; obtaining user behavior logs of the initial audience, and integrating the key business data, the analysis reasons, the optimization suggestions and the user behavior logs to obtain corresponding target data; based on the target data, using a preset large model, a diffusion model and a hybrid model to perform audience optimization processing on the initial audience to determine an optimized target audience; effect prediction is performed on the target audience, and it is determined whether the obtained effect prediction data meets the expectation; If yes, the target group is bound with a preset marketing strategy, and the target group is verified based on a preset AB experiment. If the experimental result of the AB experiment is a positive result, the target group is processed for corresponding product recommendation based on a preset full-quantity delivery strategy.

[0007] Optionally, the index analysis strategy includes a single-index analysis strategy and a multi-index analysis strategy; the key business data is analyzed based on the preset index analysis strategy to generate corresponding analysis reasons and optimization suggestions, including: The key business data is diagnosed based on the single-index analysis strategy to obtain a corresponding first attribution result; The key business data is diagnosed based on the multi-index analysis strategy to obtain a corresponding second attribution result; The analysis reasons corresponding to the key business data are generated based on the first attribution result and the second attribution result; The optimization suggestions are generated based on the analysis reasons to obtain corresponding optimization suggestions.

[0008] Optionally, the initial group is processed for group optimization based on the target data using a preset large model, a diffusion model, and a hybrid model to determine an optimized target group, including: The target data is processed for portrait generation based on the large model to obtain a group portrait corresponding to the initial group; The corresponding tree-shaped knowledge graph is constructed based on the group portrait; The initial group is processed for group expansion based on the tree-shaped knowledge graph and the diffusion model to obtain a corresponding first group; The first group is processed for group screening based on a hybrid model constructed by a random forest model and a graph neural network to obtain a screened second group; The second group is taken as the target group.

[0009] Optionally, the initial group is processed for group expansion based on the tree-shaped knowledge graph and the diffusion model to obtain a corresponding first group, including: The tree-shaped knowledge graph is input into the diffusion model to generate latent group data corresponding to the initial group through the diffusion model; A preset similarity calculation algorithm is obtained; The similarity between the latent group data and the initial group is calculated based on the similarity calculation algorithm to obtain a corresponding similarity calculation result; perform result analysis on the similarity calculation result to screen a specified crowd matching the initial crowd from the potential crowd data; take the specified crowd as the first crowd.

[0010] Optionally, the effect prediction is performed on the target crowd, and it is judged whether the obtained effect prediction data meets the expectation, comprising: calling a preset task prediction model; obtaining target crowd characteristics corresponding to the target crowd; performing effect prediction on the target crowd characteristics based on the task prediction model to obtain corresponding effect prediction data; obtaining preset historical average data; comparing the effect prediction data and the historical average data based on a preset expectation judgment strategy to generate an expectation judgment result corresponding to the effect prediction data; wherein the expectation judgment result comprises meeting the expectation or not meeting the expectation.

[0011] Optionally, after the corresponding product recommendation processing is performed on the target crowd based on the preset full-amount delivery strategy, the method further comprises: calling a preset data collection tool; using the data collection tool to collect monitoring data corresponding to the target crowd based on a preset monitoring index; obtaining a preset visual display strategy; performing display processing on the monitoring data based on the visual display strategy.

[0012] Optionally, after the monitoring data corresponding to the target crowd is collected based on the preset monitoring index using the data collection tool, the method further comprises: performing anomaly detection on the monitoring data; if it is detected that there is abnormal data in the monitoring data, obtaining a preset adjustment strategy; obtaining a preset intervention strategy; performing adjustment processing on the delivery strategy of the target crowd based on the adjustment strategy and the intervention strategy.

[0013] In addition, to achieve the above-mentioned purpose, the application further provides a data recommendation device based on artificial intelligence, comprising: a first collection module for determining an initial crowd corresponding to a business target and collecting key business data corresponding to the initial crowd; an analysis module for performing analysis processing on the key business data based on a preset index analysis strategy to generate corresponding analysis reasons and optimization suggestions; an integration module configured to acquire a user behavior log of the initial crowd and to integrate the key business data, the analysis reason, the optimization suggestion, and the user behavior log to obtain corresponding target data; an optimization module configured to perform crowd optimization processing on the initial crowd based on the target data using a preset large model, a diffusion model, and a hybrid model to determine an optimized target crowd; a judgment module configured to perform effect prediction on the target crowd and to determine whether the obtained effect prediction data meets an expectation; a verification module configured to, if so, bind the target crowd with a preset marketing strategy and to perform verification on the target crowd based on a preset AB experiment; a recommendation module configured to, if an experimental result of the AB experiment is a positive result, perform corresponding product recommendation processing on the target crowd based on a preset full-amount delivery strategy.

[0014] To solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the technical scheme as follows: The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the steps of the data recommendation method based on artificial intelligence proposed in any one of the embodiments of the present application when executing the computer program.

[0015] To solve the above technical problems, the embodiment of the present application further provides a computer readable storage medium, which adopts the technical scheme as follows: The computer readable storage medium stores a computer program, and the computer program implements the steps of the data recommendation method based on artificial intelligence proposed in any one of the embodiments of the present application when executed by a processor.

[0016] Compared with the prior art, the embodiment of the present application has the following beneficial effects: The application provides an artificial intelligence-based data recommendation method and device, computer equipment and a storage medium. The method comprises the following steps: first, determining an initial population corresponding to a business target, and collecting key business data corresponding to the initial population; then, analyzing and processing the key business data based on a preset index analysis strategy to generate corresponding analysis reasons and optimization suggestions; then, obtaining user behavior logs of the initial population, and integrating the key business data, the analysis reasons, the optimization suggestions and the user behavior logs to obtain corresponding target data; subsequently, based on the target data, using a preset large model, a diffusion model and a hybrid model to perform population optimization processing on the initial population to determine an optimized target population; further, performing effect prediction on the target population, and determining whether the obtained effect prediction data meets expectations; if yes, binding the target population with a preset marketing strategy, and verifying the target population based on a preset AB experiment; if the experimental result of the AB experiment is a positive result, performing corresponding product recommendation processing on the target population based on a preset full-quantity delivery strategy. The application realizes a highly automated marketing recommendation process by combining the steps of index analysis on the initial population, population optimization based on a large model, a diffusion model and a hybrid model, effect prediction, AB experiment verification and full-quantity delivery marketing recommendation processing, reduces manual participation and time consumption, and significantly improves the processing efficiency of marketing recommendation. In addition, the application realizes quantitative evaluation and optimization of the optimized target population based on machine learning algorithms such as a large model, a diffusion model and a hybrid model, and introduces an effect prediction and AB experiment verification mechanism to ensure high accuracy and effectiveness of target population recommendation, so that marketing decisions are more scientific and objective, the risk of subjective speculation and blind decision-making is reduced, automatic and accurate population marketing recommendation is realized, and the accuracy of marketing recommendation and marketing effect are effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the solutions in the application, the drawings needed in the description of the embodiments of the application will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Figure 1 is an exemplary system architecture diagram to which the application can be applied; Figure 2 is a flowchart of the artificial intelligence-based data recommendation method provided by the embodiment of the application; Figure 3 is a structural schematic diagram of one embodiment of the artificial intelligence-based data recommendation device according to the application; Figure 4A basic structure block diagram of a computer device for the embodiment is shown. DETAILED DESCRIPTION

[0019] The data recommendation method based on artificial intelligence provided by the embodiment of the application is applied to the data recommendation device based on artificial intelligence. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terms used in the specification of the application are only for the purpose of describing specific embodiments and are not intended to limit the application; the terms "include" and "have" and any variations thereof in the specification and claims of the application and the above description of drawings are intended to cover non-exclusive inclusion. The terms "first", "second" and the like in the specification and claims of the application or the above description of drawings are used to distinguish different objects, not to describe a specific order.

[0020] Reference herein to "embodiment" means that the specific features, structures, or characteristics described in connection with the embodiment can be included in at least one embodiment of the application. The phrase appears at various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily mutually exclusive of other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0021] In order to better understand the scheme of the application for those skilled in the art, the technical solutions in the embodiments of the application will be described clearly and completely in conjunction with the drawings below.

[0022] As shown in Figure 1 The system architecture 100 can include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used to provide a communication link medium between the terminal devices 101, 102, 103 and the server 105. The network 104 can include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0023] Users can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social online platform software, etc.

[0024] The terminal devices 101, 102, and 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smartphones, tablet computers, e-book readers, MP3 (Moving Picture Experts Group Audio Layer III) players, MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, and the like.

[0025] The server 105 can be a server providing various services, for example, a background server providing support for a page displayed on the terminal devices 101, 102, and 103.

[0026] It should be noted that the data recommendation method based on artificial intelligence provided in the embodiments of the present application is generally executed by a server / terminal device, and accordingly, the data recommendation apparatus based on artificial intelligence is generally arranged in a server / terminal device.

[0027] It should be understood that Figure 1 The number of terminal devices, networks, and servers in

[0028] With reference to Figure 2 , a flowchart of one embodiment of the data recommendation method based on artificial intelligence proposed in the present application is shown. The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology.

[0029] The data recommendation method based on artificial intelligence provided in the embodiments of the present application includes the following steps: S210, determining an initial population corresponding to a business target, and collecting key business data corresponding to the initial population.

[0030] In this step, the execution subject of the present application can be a data recommendation system, which can be referred to as a system for short. The present application can be applied to a product recommendation business scenario for targeted marketing of users, and exemplarily, can be applied to targeted marketing in a freight scenario, and can be extended to targeted marketing in multiple fields such as retail e-commerce, financial services (such as banks, securities, insurance, etc.), online education, tourism services, and medical and health services, for providing accurate target population optimization and marketing strategy support for various industries.

[0031] Among them, the operation team can determine the initial target group according to the business target and marketing strategy, and bind specific marketing strategies such as preferential activities, promotion information, etc. for the group, and then put the strategy online. Then connect with the business mysql database of the direct transaction system, configure the flink data collection task. Clearly, the key business data such as order quantity, matched order amount, transaction amount, etc. need to be collected, and the collection frequency is set, for example, once every minute. The collected key business data is stored in a suitable data warehouse or database for preliminary data cleaning and preprocessing, such as removing duplicate data, handling missing values, etc. to ensure the quality and consistency of the data.

[0032] S220, analyzing and processing the key business data based on the preset index analysis strategy to generate corresponding analysis reasons and optimization suggestions.

[0033] In this step, the specific implementation process of analyzing and processing the key business data based on the preset index analysis strategy to generate corresponding analysis reasons and optimization suggestions will be described in further detail in the subsequent specific embodiments, which will not be described here.

[0034] S230, obtaining the user behavior log of the initial group, and integrating the key business data, the analysis reasons, the optimization suggestions and the user behavior log to obtain corresponding target data.

[0035] In this step, the user behavior log of the initial group can be collected, including user browsing records, search keywords, favorite behaviors, etc. Then the key business data, analysis reasons, optimization suggestions and user behavior log are integrated, and the integrated multi-modal data is used as the target data.

[0036] S240, based on the target data, using a preset large model, diffusion model and hybrid model to perform group optimization processing on the initial group to determine the optimized target group.

[0037] In this step, the specific implementation process of using a preset large model, diffusion model and hybrid model to perform group optimization processing on the initial group based on the target data to determine the optimized target group will be described in further detail in the subsequent specific embodiments, which will not be described here.

[0038] S250, predicting the effect of the target group, and determining whether the obtained effect prediction data meets the expectation.

[0039] In this step, the effect prediction of the target population is performed, and whether the obtained effect prediction data meets the expectation is determined. The specific implementation process will be further described in the subsequent embodiments, and will not be described in detail here.

[0040] In S260, if yes, the target population is bound with the preset marketing strategy, and the target population is verified based on the preset AB experiment.

[0041] The verification process of the AB experiment includes: 1. Experimental group and control group configuration. Experimental group configuration: the optimized target population is used as the experimental group, and specific marketing strategies such as new preferential activities and personalized promotion information are bound. Control group configuration: the control group is set, and the control group adopts a blank strategy, that is, the natural flow configuration and the experimental group are configured with the same strategy (but the control group does not perform specific population optimization and strategy binding), so as to serve as a comparison benchmark. 2. Experiment development and report output. Experiment execution: start the marketing activities of the experimental group and the control group at the same time to ensure the consistency of the experimental environment. The experiment generally lasts for 2 weeks to one month to obtain sufficient data for analysis. Report output: after the experiment ends, the data of the experimental group and the control group are statistically analyzed, including the comparison of key business indicators, user behavior analysis, etc., and the AB experiment report is output. The report should clearly indicate whether the experimental results are positive, and the differences and effects between different strategies. 3. Experimental result judgment. Positive result determination: according to the AB experiment report, if the business indicators of the experimental group (such as order quantity, sales, user conversion rate, etc.) are obviously better than those of the control group, and the difference has statistical significance, it is considered that the experimental result is positive. Non-positive result processing: if the experimental result is not ideal, that is, the difference between the experimental group and the control group is not significant or the experimental group performs worse than the control group, the optimized population and strategy are not adopted, and the next population optimization scheme is continued to be analyzed.

[0042] In S270, if the experimental result of the AB experiment is a positive result, the corresponding product recommendation processing is performed on the target population based on the preset full-amount delivery strategy.

[0043] In this step, for the optimized population with a positive experimental result of the AB experiment, a corresponding full-amount delivery strategy is formulated, and the details such as the delivery time, channel, content, budget, etc. are determined. The full-amount delivery strategy can be submitted to relevant departments or personnel for approval to ensure the rationality and feasibility of the strategy.

[0044] The process of full-scale launch execution corresponding to the full-scale launch strategy includes: resource preparation: according to the full-scale launch strategy, the required resources are prepared, including human resources, technical resources, financial resources and the like. Launch implementation: according to the full-scale launch strategy, the marketing activities are fully promoted to the target population to make marketing recommendations to the target population and ensure the smooth progress of the marketing activities. Effect monitoring: continuously monitor the activity effect during the full-scale launch process, and timely adjust the strategy to cope with possible problems or challenges.

[0045] In the embodiment of the application, first, an initial population corresponding to a business target is determined, and key business data corresponding to the initial population is collected; then, the key business data is analyzed and processed based on a preset index analysis strategy to generate corresponding analysis reasons and optimization suggestions; thereafter, user behavior logs of the initial population are obtained, and the key business data, the analysis reasons, the optimization suggestions, and the user behavior logs are integrated to obtain corresponding target data; subsequently, based on the target data, a preset large model, a diffusion model, and a hybrid model are used to perform population optimization processing on the initial population to determine an optimized target population; further, the target population is predicted for effect, and it is determined whether the obtained effect prediction data meets expectations; if so, the target population is bound with a preset marketing strategy, and the target population is verified based on a preset AB experiment; if the experimental result of the AB experiment is a positive result, the target population is processed for corresponding product recommendation based on a preset full-scale launch strategy. Through the marketing recommendation processing steps of combining index analysis on the initial population, population optimization based on the large model, the diffusion model, and the hybrid model, effect prediction, AB experiment verification, and full-scale launch, the application realizes a highly automated marketing recommendation process, reduces manual participation and time consumption, and significantly improves the processing efficiency of marketing recommendation. In addition, the application realizes quantitative evaluation and optimization of the optimized target population based on machine learning algorithms such as the large model, the diffusion model, and the hybrid model, and introduces an effect prediction and AB experiment verification mechanism to ensure high accuracy and effectiveness of target population recommendation, so that marketing decisions are more scientific and objective, the risk of subjective speculation and blind decision-making is reduced, automatic and accurate population marketing recommendation is realized, and the accuracy of marketing recommendation and marketing effect are effectively improved.

[0046] Optionally, the index analysis strategy includes a single-index analysis strategy and a multi-index analysis strategy; the analysis and processing of the key business data based on the preset index analysis strategy to generate corresponding analysis reasons and optimization suggestions include: index diagnosis of the key business data based on the single-index analysis strategy to obtain a corresponding first attribution result.

[0047] In this step, the analysis and processing steps corresponding to the single-index analysis strategy include: selecting a single index that needs to be diagnosed, such as order conversion rate. Split the index according to the dimensions that may affect it, such as region, user age group, product category, etc. Calculate the contribution of each dimension to the index, which can be calculated by statistical methods (such as analysis of variance) or machine learning algorithms (such as decision tree feature importance). Take the dimension with the highest contribution as the attribution result of the index anomaly. Among them, the index diagnosis processing of the above key business data can be carried out according to the analysis and processing steps of the above single-index analysis strategy, thereby generating the corresponding first attribution result.

[0048] Based on the multi-index analysis strategy, the key business data is diagnosed to obtain the corresponding second attribution result.

[0049] In this step, the analysis and processing steps of the multi-index analysis strategy include: for a composite index such as customer lifetime value (CLV), first query its upstream indicators such as purchase frequency, average order amount, customer retention rate, etc. Through correlation analysis, regression analysis and other methods, calculate the influence factor of each upstream indicator on the composite index, and sort them. Take the upstream indicator with the largest influence factor as the attribution result of the composite index anomaly. Among them, the index diagnosis processing of the above key business data can be carried out according to the analysis and processing steps of the above multi-index analysis strategy, thereby generating the corresponding second attribution result.

[0050] Based on the first attribution result and the second attribution result, an analysis reason corresponding to the key business data is generated.

[0051] In this step, the integration result can be obtained by integrating the first attribution result and the second attribution result, and the integration result is taken as the analysis reason corresponding to the key business data.

[0052] Based on the analysis reason, a suggestion generation process is performed to obtain a corresponding optimization suggestion.

[0053] In this step, the suggestion generation process corresponding to the above analysis reason can be performed according to a preset suggestion generation strategy. Among them, the strategy content of the above suggestion generation strategy includes: according to the abnormal reason (i.e. the above analysis reason) corresponding to the above key business data diagnosed, combined with business line, city and other information, operation personnel and data analysts discuss together, output specific optimization suggestions for the abnormal reason and submit to the system. For example, if it is found that the low order conversion rate in a certain region is due to long logistics delivery time, then it can be suggested to cooperate with local logistics companies, optimize the delivery route, and shorten the delivery time; or provide additional logistics subsidies to users in the region.

[0054] In the embodiment of the present application, the single-index analysis strategy is used for index diagnosis on the key business data to obtain a corresponding first attribution result, and the multi-index analysis strategy is used for index diagnosis on the key business data to obtain a corresponding second attribution result. Then, the first attribution result and the second attribution result are used to generate an analysis reason corresponding to the key business data. Subsequently, the analysis reason is used for suggestion generation processing to obtain an optimization suggestion. The present application uses the single-index analysis strategy and the multi-index analysis strategy in combination to perform index diagnosis on the key business data, to obtain the first attribution result and the second attribution result. Then, the first attribution result and the second attribution result are used to automatically and accurately generate the analysis reason corresponding to the key business data, thereby improving the richness and accuracy of the generated analysis reason. In addition, the analysis reason is used to intelligently perform suggestion generation processing, so that the corresponding optimization suggestion can be quickly generated, thereby improving the generation efficiency of the optimization suggestion.

[0055] Optionally, the initial population is subjected to population optimization processing using a preset large model, a diffusion model, and a hybrid model based on the target data to determine an optimized target population, including: The target data is subjected to portrait generation processing based on the large model to obtain a population portrait corresponding to the initial population.

[0056] In this step, a Transformer architecture large model can be selected, and model training can be performed according to business requirements and data characteristics to construct a large model having the function of generating a coarse-grained dynamic population portrait from multi-modal data. In the model training process, the Transformer model can be trained by taking multi-modal data as input. In the training process, the model performance can be optimized by adjusting hyperparameters such as learning rate, batch size, and training rounds. In the training process, the model can be evaluated using a validation set to ensure that the model can accurately learn the behavior patterns and characteristics of users. Evaluation indicators can include accuracy, recall rate, and F1 score.

[0057] The trained large model can be used to extract key features from the target data, such as user interest preferences, behavior habits, and consumption capabilities. Then, a coarse-grained dynamic population portrait (i.e., the above-mentioned population portrait) can be generated based on the extracted features. For example, in the freight scenario, a portrait of "recently frequent cross-city transportation, price-sensitive freight drivers" can be generated.

[0058] The population portrait is used to construct a corresponding tree-shaped knowledge graph.

[0059] In this step, various features and attributes of the user included in the above people portrait can be organized in a tree structure to construct a tree knowledge graph. For example, "transportation route preference" is taken as a root node, and its child nodes can be "short-distance transportation", "long-distance transportation", "cross-city transportation", etc.

[0060] Based on the tree knowledge graph and the diffusion model, the initial people are processed for people expansion to obtain a corresponding first people.

[0061] In this step, the specific implementation process of the above people expansion processing based on the tree knowledge graph and the diffusion model on the initial people to obtain the corresponding first people will be further described in detail in subsequent specific embodiments, and will not be described in detail here.

[0062] Based on the hybrid model constructed by the random forest model and the graph neural network, the first people are processed for people screening to obtain a screened second people.

[0063] In this step, the construction process of the above hybrid model includes: first, a random forest model and a graph neural network are selected to construct a hybrid model. The random forest model is used for feature importance sorting, and the graph neural network is used for mining social relationship. Then, user feature data and user relationship data are prepared, and the user relationship data can come from social networks, transaction records, etc. Then, the random forest model and the graph neural network are trained respectively. The random forest model identifies key features through feature importance sorting; the graph neural network mines social relationship through user relationship data. The outputs of the random forest and the graph neural network are fused to construct a hybrid model, and the feature importance and the social relationship are comprehensively utilized for people screening.

[0064] The process of the mixed model for crowd screening of the first crowd includes: 1) setting an initial screening threshold according to business needs and real-time data; for example, a lower threshold can be set at the beginning of a marketing campaign to expand the target crowd. By monitoring the effect of the marketing campaign (such as click-through rate, conversion rate, etc.) in real time, dynamically adjust the screening threshold; according to the monitoring results, gradually adjust the threshold to cover enough target crowd and ensure marketing effect; establish a feedback loop mechanism to continuously optimize the threshold setting to ensure the accuracy and effectiveness of crowd screening. 2) Identify low-frequency high-value features through mixed model analysis. For example, in the freight scenario, some users may not transport frequently, but the value of the goods transported each time is high and the profit space is large; during the training process of the mixed model, adjust the feature weights to increase the weight of low-frequency high-value features, so that they play a greater role in crowd screening; according to the adjusted feature weights, filter out users with low-frequency high-value features as the optimal crowd; verify the filtered optimal crowd to ensure its high marketing value and conversion potential, and if the optimal crowd passes the verification, the optimal crowd is used as the second crowd and as the target crowd.

[0065] The second crowd is used as the target crowd.

[0066] In the embodiment of the application, the initial crowd corresponding to the initial crowd is obtained by generating a portrait of the target data based on the large model; then a corresponding tree knowledge graph is constructed based on the crowd portrait; then the initial crowd is expanded based on the tree knowledge graph and the diffusion model to obtain the first crowd; subsequently, the first crowd is screened based on a mixed model constructed by a random forest model and a graph neural network to obtain a second crowd that is screened, and the second crowd is used as the target crowd. The initial crowd corresponding to the initial crowd is obtained by generating a portrait of the target data based on the large model, and a tree knowledge graph is constructed based on the crowd portrait. Then, the initial crowd is expanded based on the combination of the tree knowledge graph and the diffusion model to obtain the first crowd, and subsequently, the first crowd is screened based on the mixed model constructed by the random forest model and the graph neural network to obtain the second crowd that is screened and used as the target crowd. The combination of the large model, the tree knowledge graph, the diffusion model and the mixed model can efficiently and accurately complete the crowd optimization of the initial crowd, and improve the accuracy of the generated target crowd.

[0067] Optionally, the initial crowd is expanded based on the tree knowledge graph and the diffusion model to obtain the corresponding first crowd, comprising: inputting the tree-shaped knowledge graph into the diffusion model, and generating latent population data corresponding to the initial population through the diffusion model.

[0068] In this step, a suitable diffusion model can be selected according to actual needs, such as variants of generative adversarial networks (GANs) (such as CycleGAN, StyleGAN, etc.), which can generate high-quality synthetic data. Specifically, the construction process of the diffusion model includes: generator training: the generator is responsible for generating new data samples, and by continuously learning the distribution of real data, the generated data is closer and closer to the real data. Discriminator training: the discriminator is responsible for judging whether the generated data sample is real, and through adversarial training with the generator, the judgment ability of the discriminator is improved. Adversarial training process: in the training process, the generator and the discriminator are trained alternately, and by minimizing the loss function of the generator and maximizing the loss function of the discriminator, the quality of the generated data is continuously improved. Training evaluation: during the training process, the quality of the generated data is evaluated regularly to ensure that the generated data is close to the real population in terms of feature distribution.

[0069] Among them, by taking the tree-shaped knowledge graph as input, the key nodes and features of the user behavior chain are extracted as input data of the diffusion model. Then the trained diffusion model is used to process the above input data to generate new latent population data, which has similar feature distribution to the real population (i.e. the initial population).

[0070] Obtain a preset similarity calculation algorithm.

[0071] In this step, the selection of the above similarity calculation algorithm is not specifically limited and can be determined according to actual business needs, for example, cosine similarity, Euclidean distance, etc. algorithm.

[0072] Calculate the similarity between the latent population data and the initial population based on the similarity calculation algorithm to obtain the corresponding similarity calculation result.

[0073] In this step, key features such as user's basic information, behavior characteristics, and consumption characteristics can be extracted from the generated latent population data and the original population data. Then, according to the selected similarity calculation algorithm, the extracted key features are calculated to calculate the similarity between the latent population data and the original population, i.e. the above similarity calculation result.

[0074] Perform result analysis on the similarity calculation result to filter out a specified population matching the initial population from the latent population data.

[0075] In this step, according to the similarity calculation result, a new target population similar to the original population (similarity greater than a preset similarity threshold) and possibly interested in the marketing strategy can be screened out. The similar population screened out is verified to ensure that it has high marketing value and conversion potential, and then the specific similar population that passes the verification is used as the above-mentioned specified population and used as the first population in the subsequent process. The value of the similarity threshold is not specifically limited, for example, it can be set to 0.9.

[0076] The specified population is used as the first population.

[0077] In the embodiments of the present application, the tree-shaped knowledge graph is input into the diffusion model, and the latent population data corresponding to the initial population is generated by the diffusion model. Then, a preset similarity calculation algorithm is obtained, and the similarity between the latent population data and the initial population is calculated based on the similarity calculation algorithm to obtain the corresponding similarity calculation result. Then, the similarity calculation result is analyzed to screen out a specified population matching the initial population from the latent population data. The specified population is used as the first population in the subsequent process. By inputting the tree-shaped knowledge graph into the diffusion model, the latent population data corresponding to the initial population is generated by using the diffusion model. Then, the similarity between the latent population data and the initial population is calculated based on the use of the similarity calculation algorithm to obtain the corresponding similarity calculation result. Then, the similarity calculation result is analyzed, so that the similar target population matching the initial population can be efficiently and accurately screened out from the latent population data and used as the first population, and the accuracy of the generated first population is effectively ensured.

[0078] Optionally, the effect prediction of the target population and the judgment of whether the obtained effect prediction data meets the expectation include: A preset task prediction model is called.

[0079] In this step, the above task prediction model can be a lightweight multi-task prediction model based on the XGBoost framework. Specifically, the model construction process of the task prediction model includes: collecting historical data containing user features, marketing strategies, business indicators, etc. as sample data, and dividing the sample data into training set, validation set and test set. Then use the training set to train the XGBoost model, adjust the hyperparameters (such as learning rate, number of trees, depth of tree, etc.) to optimize the model performance. During the training process, the validation set is used to monitor the model performance to prevent overfitting. Among them, the sparse key information enhancement mechanism can be introduced to adjust the loss function or feature weight to improve the model's attention to important features. For example, the sensitivity of the model to important features can be enhanced by increasing the weight of key features or using attention mechanisms. Subsequently, the test set is used to evaluate the model performance, and the prediction error rate is calculated until a model that meets the construction requirements, such as a prediction error rate ≤ 5%, is obtained as the above task prediction model. Evaluation indicators can include mean squared error (MSE), mean absolute error (MAE), R² score, etc.

[0080] Obtain target population characteristics corresponding to the target population.

[0081] In this step, the target population characteristics can be obtained by feature extraction on the target population. Among them, the target population characteristics can at least include user features, marketing strategy features, etc.

[0082] Based on the task prediction model, the effect prediction data corresponding to the target population characteristics is obtained.

[0083] In this step, the target population characteristics can be input into the task prediction model, and the model can predict multiple targets (such as order volume, sales, user retention rate, etc.). The accuracy and reliability of the prediction results are ensured, the prediction accuracy is improved by continuously adjusting the model parameters and optimizing the feature engineering, and the final output of the task prediction model is obtained as the corresponding effect prediction data.

[0084] Obtain pre-set historical average data.

[0085] In this step, historical average data is collected in advance, including historical average values of key business indicators (such as order volume, sales, user retention rate, etc.).

[0086] Compare the effect prediction data and the historical average data based on the pre-set expected judgment strategy to generate an expected judgment result corresponding to the effect prediction data; wherein the expected judgment result includes meeting expectations or not meeting expectations.

[0087] In this step, the policy content of the above-mentioned expected determination strategy includes: comparing the current effect prediction data with the historical average data, calculating the proportion of key business indicator data reaching or exceeding the historical average data. Then set a threshold, for example, require more than 80% of the key business indicator data to reach or exceed the historical average data. Then according to the data comparison result, judge whether the prediction result meets the expectation. If more than 80% of the key business indicator data reaches or exceeds the historical average data, it is considered that the prediction result meets the expectation, and the next step can be entered. Otherwise, it is considered that the prediction result does not meet the expectation, and the next person group optimization scheme needs to be analyzed, which may need to adjust the algorithm model, optimize the person group characteristics or redevelop the marketing strategy.

[0088] Among them, the comparison processing between the effect prediction data and the historical average data can be performed according to the policy content of the above-mentioned expected determination strategy, and the expected determination result of the effect prediction data is generated.

[0089] In the embodiment of the application, the preset task prediction model is called; then the target person group characteristics corresponding to the target person group are obtained; then the target person group characteristics are predicted based on the task prediction model to obtain corresponding effect prediction data; subsequently, preset historical average data is obtained; finally, the effect prediction data and the historical average data are compared based on the preset expected determination strategy to generate an expected determination result corresponding to the effect prediction data; wherein, the expected determination result includes meeting the expectation or not meeting the expectation. Through obtaining the target person group characteristics corresponding to the target person group, and then predicting the target person group characteristics based on the use of the task prediction model to obtain the effect prediction data, and then comparing the effect prediction data with the obtained historical average data based on the use of the expected determination strategy, the expected determination result of the effect prediction data can be efficiently and accurately generated, the generation efficiency of the expected determination result is improved, and the accuracy of the obtained expected determination result is ensured.

[0090] Optionally, after the corresponding product recommendation processing of the target person group based on the preset full-amount delivery strategy, the method further comprises: Calling a preset data collection tool.

[0091] In this step, the selection of the above-mentioned data collection tool is not specifically limited, and can be determined according to actual business needs. For example, a suitable data collection tool can be selected, such as a log collection tool, an API interface, a database query, etc., to ensure that the required data can be collected in real time or periodically.

[0092] Using the data collection tool, monitoring data corresponding to the target person group is collected based on a preset monitoring index.

[0093] In this step, a monitoring index system containing monitoring indicators can be established according to actual business needs and business objectives. Specifically, monitoring indicators can include order conversion rate, user satisfaction, complaint rate, user retention rate, number of active users, etc. Among them, first of all, based on the above monitoring indicators, the data types that need to be collected are determined, including transaction data (such as order volume, sales, conversion rate, etc.), user feedback data (such as user evaluation, complaints, satisfaction survey, etc.) and other related data (such as user behavior logs, system logs, etc.). Further, based on the above data types, monitoring data matching the target population is collected from the corresponding data sources. In addition, the collected monitoring data can be stored in appropriate data storage media, such as data warehouse, database or data lake, to ensure the security of the monitoring data and facilitate subsequent analysis and monitoring.

[0094] Obtain a preset visual display strategy.

[0095] In this step, the display processing steps of the visual display strategy include: 1. Visual tool selection: select appropriate visual tools or platforms, such as Tableau, Power BI, Grafana, etc., to display monitoring data in the form of charts, reports, etc. 2. Dashboard design: design an intuitive and easy-to-understand dashboard to display key monitoring indicators and trend analysis. The dashboard should include real-time data updates, historical data comparisons, abnormal data alerts, etc. 3. Data update mechanism: ensure that the data displayed in the visualization can be updated in real time or periodically, so that the operation personnel can understand the effect of marketing activities and user feedback in a timely manner. 4. User permission management: set user permissions according to the needs of different roles to ensure that only authorized personnel can view and operate related data.

[0096] Based on the visual display strategy, the monitoring data is displayed.

[0097] In this step, based on the display processing steps corresponding to the above visual display strategy, the display processing of the above monitoring data is performed.

[0098] In the embodiment of the present application, a preset data collection tool is called, then the data collection tool is used to collect monitoring data corresponding to the target population based on a preset monitoring index, then a preset visual display strategy is obtained, and subsequently the monitoring data is displayed based on the visual display strategy. After the target population is recommended corresponding products based on the use of the full-amount delivery strategy, the data collection tool is automatically used to collect monitoring data corresponding to the target population based on the monitoring index, and then the monitoring data is displayed intelligently based on the use of the visual display strategy, so that the collection efficiency of the monitoring data and the display intelligence are improved, so that the operation personnel can timely understand the effect of the marketing activities and the user feedback, and the work efficiency and work experience of the operation personnel can be improved.

[0099] Optionally, after the data collection tool is used to collect monitoring data corresponding to the target population based on a preset monitoring index, the method further comprises: Abnormal detection is performed on the monitoring data.

[0100] In this step, the abnormal situation in the monitoring data can be detected in real time by a statistical method or a machine learning algorithm, such as a sudden decrease in order quantity, an increase in user complaint rate, etc. If an abnormality is detected or an expected target is not reached, it is determined that there is abnormal data in the monitoring data.

[0101] If it is detected that there is abnormal data in the monitoring data, a preset adjustment strategy is obtained.

[0102] In this step, the strategy content of the above adjustment strategy includes that when it is detected that the monitoring data is abnormal, the system automatically adjusts the target population or the marketing strategy. For example, the promotion strength, the preferential policy, the target population screening condition, etc. are automatically adjusted. After the automatic adjustment, the adjustment effect is continuously monitored, the effectiveness of the adjustment strategy is evaluated, and further optimization is performed according to the need.

[0103] A preset intervention strategy is obtained.

[0104] In this step, the intervention strategy can be formulated according to the monitoring data and the actual business needs, such as adjusting the screening condition of the target population, optimizing the marketing copy, adjusting the promotion channel, etc. By implementing the formulated intervention strategy into the marketing activities, it is ensured that the adjustment measures can take effect in time. After the intervention, the intervention effect is continuously monitored, the feedback data is collected, the effectiveness of the intervention strategy is evaluated, and reference is provided for subsequent automatic adjustment.

[0105] Based on the adjustment strategy and the intervention strategy, an adjustment processing of the delivery strategy for the target population is performed.

[0106] In this step, the target population can be adjusted according to the policy content of the above adjustment strategy and the policy content of the above intervention strategy, so as to ensure that the marketing activity can be dynamically optimized according to real-time data, and improve the effect of the marketing activity and the satisfaction of the user.

[0107] In the embodiment of the application, the monitoring data is detected for abnormalities, if abnormal data is detected in the monitoring data, a preset adjustment strategy is obtained, then a preset intervention strategy is obtained, and subsequently, the target population is adjusted according to the adjustment strategy and the intervention strategy. After the data collection tool is used to collect the monitoring data corresponding to the target population, the application intelligently detects the monitoring data for abnormalities, and when it is detected that there is abnormal data in the monitoring data, the application automatically adjusts the target population according to the adjustment strategy and the intervention strategy, so as to effectively ensure that the marketing activity can be dynamically optimized according to real-time data, and to improve the effect of the marketing activity and the satisfaction of the user.

[0108] In some optional implementations, the obtained user information is subject to user consent and complies with relevant laws and relevant policies.

[0109] The application aims to provide a marketing recommendation method for optimizing target population in freight scenarios. Through five steps of diagnostic analysis, population optimization, effect prediction, AB experiment verification and population online, the optimal target population is recommended, the efficiency and scientificity of target population optimization decision are improved, the reliability of marketing results is enhanced, the dependence on business experience and manual time consumption is reduced, and the enterprise is assisted to adjust marketing strategy in time.

[0110] In addition to marketing recommendation in the freight industry, the application can be expanded from marketing objects and business types, and the following are other expandable scenarios: 1. E-commerce industry: targeted marketing for users of e-commerce platforms. Diagnostic analysis can be performed by analyzing user shopping history, browsing preferences, and collection behavior data to identify high-value customer characteristics. Similar potential users can be automatically optimized using hybrid algorithms to predict purchase conversion rates and other effect indicators under different marketing activities (such as full price reduction, discount, and gift). The effects of different marketing plans (such as different advertising copy and promotion combinations) are verified through AB experiments, and finally, accurate product recommendations and marketing are launched.

[0111] 2. Financial Industry: Targeted marketing to customers in the banking, securities, and insurance sectors. For example, banks can segment customers based on data such as asset status, transaction records, and risk preferences to optimize their target audience. Predicting customer purchase intentions and amounts under different financial product promotion schemes, and comparing the effectiveness of different marketing channels (online advertising, offline activities) and content (product presentation methods) through A / B experiments, can improve the sales efficiency of financial products.

[0112] 3. Tourism Industry: For tourism enterprises, diagnostic analysis can be conducted based on data such as tourists' travel history, destination preferences, and budget range. Hybrid algorithms can be used to automatically optimize the search for potential tourists and predict the attractiveness and booking rates of different tour packages and promotional activities. A / B testing can be used to compare the effectiveness of different marketing methods (social media promotion, travel agency partnerships) to achieve precise marketing of tourism products.

[0113] In summary, the processing method of the present invention is not only applicable to targeted marketing in freight scenarios, but can also be extended to multiple fields such as retail e-commerce, financial services, online education, tourism services, and healthcare, providing precise target audience optimization and marketing strategy support for various industries.

[0114] In addition, the main innovative points of this invention include: 1. Full-process automatic audience optimization method: This paper proposes a full-process optimization method from diagnostic analysis to audience deployment, which covers the design and implementation of five steps: diagnostic analysis, audience optimization, effect prediction, A / B test verification and target audience deployment.

[0115] 2. Efficient audience validation mechanism: Effect prediction and A / B test validation through algorithm model: By introducing an effect prediction model and A / B test validation mechanism, this invention ensures high accuracy and effectiveness of target audience recommendations, and improves the scientificity and reliability of marketing decisions.

[0116] 3. Multi-model fusion for audience screening: Combining random forest (RF) and graph neural network (GNN) for audience screening, random forest is used to determine feature importance, while graph neural network is used to mine community relationships, dynamically adjust the screening threshold, capture low-frequency but high-value features, and accurately screen the optimal audience.

[0117] Furthermore, compared to the prior art, the advantages of the present invention are as follows: 1. Improve operational efficiency: This invention realizes closed-loop management from diagnostic analysis, A / B verification, algorithm recommendation to user deployment and effect feedback. The entire process is highly automated, reducing manual participation and time consumption, minimizing human intervention, and significantly improving operational efficiency.

[0118] 2. Data-driven scientific decision-making: The present application is based on big data analysis and machine learning algorithms, which realizes the quantitative evaluation and optimization of the target population effect, makes the marketing decision more scientific and objective, reduces the risk of subjective speculation and blind decision-making, realizes automatic population recommendation and continuous improvement, provides scientific and accurate decision-making basis for enterprises, and reduces the risk of decision-making.

[0119] 3. Reducing cost and improving efficiency: The present application can adjust the operation strategy according to real-time data, optimize resource allocation, and avoid resource waste. By means of data-driven operation mode, enterprises can deeply understand user demand and behavior habits, tap the potential value of users, and improve user repurchase rate and consumption frequency through optimizing operation strategy, so as to maximize user value.

[0120] Further referring to Figure 3 , as an implementation of the method shown in Figure 2 , the present application provides an embodiment of a data recommendation device 300 based on artificial intelligence. The device embodiment corresponds to the method embodiment shown in Figure 2 , and the device can be applied to various electronic devices.

[0121] The data recommendation device 300 based on artificial intelligence provided by the embodiment of the present application comprises: The first acquisition module 310 is used for determining an initial population corresponding to a business target, and acquiring key business data corresponding to the initial population; The analysis module 320 is used for analyzing and processing the key business data based on a preset index analysis strategy, and generating corresponding analysis reasons and optimization suggestions; The integration module 330 is used for acquiring user behavior logs of the initial population, and performing data integration on the key business data, the analysis reasons, the optimization suggestions and the user behavior logs to obtain corresponding target data; The optimization module 340 is used for performing population optimization processing on the initial population based on the target data using a preset large model, a diffusion model and a hybrid model, to determine an optimized target population; The judgment module 350 is used for predicting the effect of the target population, and judging whether the obtained effect prediction data meets the expectation; The verification module 360 is used for binding the target population with a preset marketing strategy if yes, and verifying the target population based on a preset AB experiment; The recommendation module 370 is used for performing corresponding product recommendation processing on the target population based on a preset full-amount delivery strategy if the experimental result of the AB experiment is a positive result.

[0122] Optionally, the analysis module 320 comprises: a first processing submodule, configured to perform index diagnosis on the key business data based on the single-index analysis strategy, to obtain a corresponding first attribution result; a second processing submodule, configured to perform index diagnosis on the key business data based on the multi-index analysis strategy, to obtain a corresponding second attribution result; a first generation submodule, configured to generate an analysis reason corresponding to the key business data based on the first attribution result and the second attribution result; a second generation submodule, configured to perform suggestion generation processing based on the analysis reason, to obtain a corresponding optimization suggestion.

[0123] Optionally, the optimization module 340 comprises: a third generation submodule, configured to perform portrait generation processing on the target data based on the large model, to obtain a population portrait corresponding to the initial population; a construction submodule, configured to construct a corresponding tree-shaped knowledge graph based on the population portrait; an expansion submodule, configured to perform population expansion processing on the initial population based on the tree-shaped knowledge graph and the diffusion model, to obtain a corresponding first population; a screening submodule, configured to perform population screening processing on the first population based on a hybrid model constructed by a random forest model and a graph neural network, to obtain a screened second population; a determination submodule, configured to determine the second population as the target population.

[0124] Optionally, the expansion submodule comprises: a generation unit, configured to input the tree-shaped knowledge graph into the diffusion model, and generate latent population data corresponding to the initial population through the diffusion model; an acquisition unit, configured to acquire a preset similarity calculation algorithm; a calculation unit, configured to calculate a similarity between the latent population data and the initial population based on the similarity calculation algorithm, to obtain a corresponding similarity calculation result; an analysis unit, configured to perform result analysis on the similarity calculation result, to screen out a specified population matching the initial population from the latent population data; a determination unit, configured to determine the specified population as the first population.

[0125] Optionally, the judgment module 350 comprises: a calling submodule, configured to call a preset task prediction model; The first obtaining sub-module is configured to obtain target crowd characteristics corresponding to the target crowd. The prediction sub-module is configured to perform effect prediction on the target crowd characteristics based on the task prediction model, to obtain corresponding effect prediction data. The second obtaining sub-module is configured to obtain preset historical average data. The comparison sub-module is configured to compare the effect prediction data and the historical average data based on a preset expected decision strategy, to generate an expected decision result corresponding to the effect prediction data; the expected decision result includes meeting an expectation or not meeting the expectation.

[0126] Optionally, the data recommendation apparatus 300 based on artificial intelligence further includes: The calling module is configured to call a preset data collection tool. The second collection module is configured to collect monitoring data corresponding to the target crowd based on a preset monitoring index by using the data collection tool. The first obtaining module is configured to obtain a preset visual display strategy. The display module is configured to perform display processing on the monitoring data based on the visual display strategy.

[0127] Optionally, the data recommendation apparatus 300 based on artificial intelligence further includes: The detection module is configured to perform abnormality detection on the monitoring data. The second obtaining module is configured to obtain a preset adjustment strategy if it is detected that there is abnormal data in the monitoring data. The third obtaining module is configured to obtain a preset intervention strategy. The adjustment module is configured to perform adjustment processing on a delivery strategy of the target crowd based on the adjustment strategy and the intervention strategy.

[0128] To solve the above technical problems, the embodiments of the present application further provide a computer device. For details, please refer to Figure 4 , Figure 4 The basic structure block diagram of the computer device of the present embodiment is shown in FIG. 1. The computer device 4 includes a memory 41, a processor 42, and a network interface 43, which are communicatively connected to each other through a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure, but it should be understood that all the shown components are not required to be implemented, and more or fewer components can be alternatively implemented. Among them, those skilled in the art can understand that the computer device herein is a device capable of automatically performing numerical calculation and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0129] The computer device can be a desktop computer, a notebook computer, a palm computer, a cloud server, and the like. The computer device can interact with the user through a keyboard, a mouse, a remote controller, a touchpad, a voice control device, and the like.

[0130] The memory 41 includes at least one type of readable storage medium, which includes a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, and the like. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or a memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, and the like. Of course, the memory 41 can also include both the internal storage unit and the external storage device of the computer device 4. In the present embodiment, the memory 41 is generally used to store an operating system and various application software installed in the computer device 4, such as program codes of the data recommendation method based on artificial intelligence, and the like. In addition, the memory 41 can also be used to temporarily store various data that have been output or will be output.

[0131] The processor 42 may, in some embodiments, be a Central Processing Unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 42 is generally used to control the overall operation of the computer device 4. In the present embodiment, the processor 42 is configured to execute program code stored in the memory 41 or to process data, such as to execute program code of the artificial intelligence-based data recommendation method.

[0132] The network interface 43 may include a wireless network interface or a wired network interface, and is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0133] The present application also provides another embodiment, i.e., to provide a computer readable storage medium storing the application crash processing program, which can be executed by at least one processor to enable the at least one processor to perform the steps of the artificial intelligence-based data recommendation method as described above.

[0134] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of software and necessary general hardware online platform, of course, it can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes a plurality of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to perform the methods described in the various embodiments of the present application.

[0135] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected by a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0136] Obviously, the above-described embodiments are only some embodiments but not all embodiments of the present application, the preferred embodiments of the present application are shown in the drawings, but do not limit the patent scope of the present application. The present application can be implemented in many different forms, and conversely, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent replacements to some technical features therein. Any equivalent structure made by using the content of the specification and drawings, directly or indirectly applied to other related technical fields, is also within the patent protection scope of the present application.

Claims

1. An artificial intelligence-based data recommendation method, characterized by, The method comprises the following steps: determining an initial population corresponding to a business target, and collecting key business data corresponding to the initial population; analyzing and processing the key business data based on a preset index analysis strategy, generating corresponding analysis reasons and optimization suggestions; obtaining user behavior logs of the initial population, and integrating the key business data, the analysis reasons, the optimization suggestions, and the user behavior logs to obtain corresponding target data; based on the target data, using a preset large model, a diffusion model, and a hybrid model to perform population optimization processing on the initial population to determine an optimized target population; performing effect prediction on the target population and determining whether the obtained effect prediction data meets expectations; if yes, binding the target population with a preset marketing strategy and verifying the target population based on a preset AB experiment; if the experimental result of the AB experiment is a positive result, performing corresponding product recommendation processing on the target population based on a preset full-amount delivery strategy.

2. The method of claim 1, wherein, The index analysis strategy includes a single-index analysis strategy and a multi-index analysis strategy; the analysis and processing of the key business data based on the preset index analysis strategy to generate corresponding analysis reasons and optimization suggestions includes: performing index diagnosis on the key business data based on the single-index analysis strategy to obtain a corresponding first attribution result; performing index diagnosis on the key business data based on the multi-index analysis strategy to obtain a corresponding second attribution result; generating an analysis reason corresponding to the key business data based on the first attribution result and the second attribution result; generating an optimization suggestion based on the analysis reason.

3. The method of claim 1, wherein, The population optimization processing on the initial population based on the target data using the preset large model, diffusion model, and hybrid model to determine the optimized target population includes: performing portrait generation processing on the target data based on the large model to obtain a population portrait corresponding to the initial population; constructing a corresponding tree-shaped knowledge graph based on the population portrait; performing population expansion processing on the initial population based on the tree-shaped knowledge graph and the diffusion model to obtain a corresponding first population; performing population screening processing on the first population based on a hybrid model constructed by a random forest model and a graph neural network to obtain a screened second population; the second population is taken as the target population.

4. The method of claim 3, wherein, The population expansion processing on the initial population based on the tree-shaped knowledge graph and the diffusion model to obtain the corresponding first population includes: inputting the tree-shaped knowledge graph into the diffusion model to generate latent population data corresponding to the initial population through the diffusion model; obtaining a preset similarity calculation algorithm; calculating the similarity between the latent population data and the initial population based on the similarity calculation algorithm to obtain a corresponding similarity calculation result; performing result analysis on the similarity calculation result to screen out a specified population matching the initial population from the latent population data; the specified population is taken as the first population.

5. The method of claim 1, wherein, The effect prediction is performed on the target population, and whether the obtained effect prediction data meets the expectation is judged, comprising: calling a preset task prediction model; obtaining target population characteristics corresponding to the target population; based on the task prediction model, the target population characteristics are predicted to obtain corresponding effect prediction data; obtaining preset historical average data; based on the preset expected judgment strategy, the effect prediction data and the historical average data are compared to generate an expected judgment result corresponding to the effect prediction data; wherein the expected judgment result includes meeting the expectation or not meeting the expectation.

6. The method of claim 1, wherein, After the corresponding product recommendation processing of the target population based on the preset full-amount delivery strategy, the method further comprises: calling a preset data collection tool; using the data collection tool, monitoring data corresponding to the target population is collected based on a preset monitoring index; obtaining a preset visual display strategy; based on the visual display strategy, the monitoring data is displayed.

7. The method of claim 6, wherein, After using the data collection tool to collect monitoring data corresponding to the target population based on a preset monitoring index, the method further comprises: abnormal detection is performed on the monitoring data; if it is detected that there is abnormal data in the monitoring data, a preset adjustment strategy is obtained; obtaining a preset intervention strategy; based on the adjustment strategy and the intervention strategy, the target population is adjusted for delivery strategy processing. 8.A data recommendation apparatus based on artificial intelligence, characterized by, comprising: a first acquisition module for determining an initial population corresponding to a business target and acquiring key business data corresponding to the initial population; an analysis module for analyzing the key business data based on a preset index analysis strategy to generate corresponding analysis reasons and optimization suggestions; an integration module for obtaining user behavior logs of the initial population and integrating the key business data, the analysis reasons, the optimization suggestions and the user behavior logs to obtain corresponding target data; an optimization module for using a preset large model, diffusion model and hybrid model to perform population optimization processing on the initial population based on the target data to determine an optimized target population; a judgment module for predicting the effect of the target population and judging whether the obtained effect prediction data meets the expectation; a verification module for binding the target population with a preset marketing strategy if so, and verifying the target population based on a preset AB experiment; a recommendation module for performing corresponding product recommendation processing on the target population based on a preset full-amount delivery strategy if the experimental result of the AB experiment is a positive result.

9. A computer device, comprising: comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to realize the steps of the artificial intelligence-based data recommendation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer program is stored on the computer readable storage medium, and the computer program is executed by the processor to realize the steps of the artificial intelligence-based data recommendation method according to any one of claims 1 to 7.

Images