Automatic information pushing method and system and application thereof in green financial field

Abstract

The invention relates to the technical field of computers, and discloses an automatic information pushing method and system and application thereof in the field of green finance, and the method comprises the steps: collecting to-be-pushed target information through a preset data source, and detecting the information updating condition of the data source; setting a weight for the target information according to the information updating condition of the data source; extracting features of the target information from the target information, and extracting features of historical behaviors of the user from the historical behavior data of the user; calculating the matching degree between the target information and the user according to the features of the target information and the features of the historical behaviors of the user; according to the weight of the target information and the matching degree between the target information and the user, whether the target information is pushed to the user is judged; and when the judgment result is yes, pushing the target information to the user. Through the dynamic weight mechanism, the feature-driven matching model and the multi-dimensional decision logic, the information pushing accuracy, timeliness and user experience are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and more specifically, to an information automatic push method and system and their application in the field of green finance. Background Art

[0002] With the rapid development of Internet technology, information push systems have become the core hub connecting users and information resources. Traditional information push methods mainly rely on static rule engines, and perform content recommendation through predefined keyword matching or collaborative filtering algorithms based on users' explicit behaviors (such as clicks, favorites). Although such methods can achieve basic push functions, there are problems such as lag in data source updates and single user interest modeling. In recent years, although some technologies have tried to introduce machine learning models to extract text keywords of information and perform dynamic push according to the relationship between keywords and users, the accuracy of matching between users and information is limited. Therefore, there is an urgent need for an information push method and system that can adapt to information content and dynamically and accurately push information to users. Summary of the Invention

[0003] In order to solve the above technical problems, the present application is proposed to provide an information automatic push method and system that can adapt to information content and dynamically and accurately push information to users and their application in the field of green finance.

[0004] In a first aspect, the present invention provides an information automatic push method, including: collecting target information to be pushed through a preset data source, and detecting the information update situation of the data source; setting a weight for the target information according to the information update situation of the data source; extracting features of the target information from the target information, and extracting features of the user's historical behavior from the user's historical behavior data; calculating a matching degree between the target information and the user according to the features of the target information and the features of the user's historical behavior; judging whether to push the target information to the user according to the weight of the target information and the matching degree between the target information and the user; when the judgment result is yes, pushing the target information to the user.

[0005] Optionally, in the foregoing information automatic push method, setting a weight for the target information according to the information update situation of the data source includes: extracting the number of information updates of the data source within a preset time interval and the last update time from the information update situation of the data source; calculating the weight of the target information according to the number of information updates of the data source within the preset time interval and the last update time , where is the number of information updates of the data source within the time interval, is a set of the number of information updates of a preset plurality of data sources within the time interval, is the number of information updates of any one of the plurality of data sources within the time interval, is a preset time decay factor, is the current time, is the last update time of the data source within the time interval.

[0006] Optionally, in the foregoing information automatic push method, before calculating the weight of the target information according to the number of information updates and the last update time of the data source within a preset time interval, the information automatic push method further includes: obtaining the historical time decay factor recorded most recently ; counting the number of information refreshes of the user within the time interval ; obtaining the hot information within the time interval; recalculating the time decay factor , where is a preset threshold value of the number of refreshes, represents the target information, represents the set of the hot information, represents any one of the hot information, represents the total number of the hot information.

[0007] Optionally, in the foregoing information automatic push method, extracting the features of the target information from the target information includes: identifying the field to which the target information belongs; obtaining the corresponding knowledge base according to the field to which the target information belongs; fusing the features of the target information with the knowledge base to update the features of the target information.

[0008] Optionally, in the foregoing information automatic push method, extracting the features of the target information from the target information and extracting the features of the historical behavior of the user from the historical behavior data of the user includes: extracting the dense vector of the target information as the spatial vector feature of the target information through the trained neural network, and extracting the dense vector of the historical behavior data of the user as the spatial vector feature of the historical behavior of the user; counting the discrete distribution of the target information as the probability distribution feature of the target information, and counting the discrete distribution of the historical behavior data of the user as the probability distribution feature of the historical behavior of the user.

[0009] Optionally, in the aforementioned information automatic pushing method, calculating a matching degree between the target information and the user according to characteristics of the target information and characteristics of the user's historical behavior includes: calculating a similarity degree between the target information and the user's historical behavior according to a spatial vector feature of the target information and a spatial vector feature of the user's historical behavior , where is a preset number of modalities, is a spatial vector feature of the th modality in the spatial vector feature of the target information, represents the user's historical behavior, is a spatial vector feature of the th modality in the spatial vector feature of the user's historical behavior, represents the target information, The function is the Jaccard function for calculating the similarity degree; calculating a difference degree between the target information and the user's historical behavior according to a probability distribution feature of the target information and a probability distribution feature of the user's historical behavior , where is a probability distribution feature of the th modality in the probability distribution feature of the target information, is a probability distribution feature of the th modality in the probability distribution feature of the user's historical behavior, The function is used to calculate the JS divergence; calculating the matching degree between the target information and the user according to the similarity degree and the difference degree between the target information and the user's historical behavior , where is a preset smoothing factor, , are a preset first balance coefficient and second balance coefficient respectively

[0010] Optionally, in the aforementioned information automatic pushing method, pushing the target information to the user includes: converting non-text modality content in the target information into a text modality; compressing the target information; calculating a probability that each word in the target information is retained after compression, where the probability that the jth word is retained after compression is , is the target information after compression; calculating a compression coefficient of the target information , where is a set of words included in the target information, is the length of the target information after compression; determining the compression coefficient Whether it exceeds a preset threshold; when the judgment result is yes, perform the step of pushing the target information to the user; when the judgment result is no, return to the step of compressing the target information.

[0011] Optionally, in the foregoing information automatic pushing method, the target information may be advertising information or course information in the field of green finance.

[0012] In a second aspect, the present invention provides an information automatic pushing system, including: an information collection module, which collects target information to be pushed through a preset data source and detects the information update situation of the data source; a weight setting module, which sets a weight for the target information according to the information update situation of the data source; a feature extraction module, which extracts the features of the target information from the target information and extracts the features of the historical behavior of the user from the historical behavior data of the user; a matching degree calculation module, which calculates the matching degree between the target information and the user according to the features of the target information and the features of the historical behavior of the user; a judgment module, which judges whether to push the target information to the user according to the weight of the target information and the matching degree between the target information and the user; a pushing module, which pushes the target information to the user when the judgment result is yes.

[0013] One or more of the above technical solutions of the present invention have at least one or more of the following beneficial effects:

[0014] According to the technical solution of the present invention, through a dynamic weight mechanism, a feature-driven matching model, and a multi-dimensional decision logic, while improving the pushing accuracy, timeliness, and user experience, the efficient utilization of resources is realized. It not only solves the "one-size-fits-all" problem of traditional pushing systems, but also provides an extensible solution for personalized services and information distribution optimization, and is applicable to scenarios such as e-commerce, social media, news media, and online education that require accurate information pushing. Description of the Drawings

[0015] By describing the embodiments of the present application in more detail in combination with the drawings, the above and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. They are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 It is a flowchart of an information pushing method according to an embodiment of the present application;

[0017] Figure 2 It is a partial flowchart of an information pushing method according to an embodiment of the present application;

[0018] Figure 3 Another layout flowchart of the information push method according to an embodiment of the present application;

[0019] Figure 4 Another partial flowchart of the information push method according to an embodiment of the present application;

[0020] Figure 5 A block diagram of the information push system according to an embodiment of the present application. Detailed implementation manners

[0021] Some implementation manners of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these implementation manners are only used to explain the technical principle of the present invention and are not intended to limit the protection scope of the present invention.

[0022] As Figure 1 shown, in an embodiment of the present invention, an information automatic push method is provided, including:

[0023] Step S110, collecting target information to be pushed through a preset data source and detecting the information update situation of the data source.

[0024] In this embodiment, the type of the target information is not limited. For example, it can be advertising information or course information in the field of green finance. The target information can be text information, text information, image information, voice information, video information or other types of information, etc. The data source can be various media platforms, public platforms (including multimedia platforms), public service platforms, etc. When the solution of the present application is specifically applied to the field of green finance, the target information can include, but is not limited to, advertising information or course information in the field of green finance, and the relevant intermediate data or result data can be data related to advertising information or course information in the field of green finance, etc., which will not be elaborated here.

[0025] Step S120, setting a weight for the target information according to the information update situation of the data source.

[0026] In this embodiment, the weight of the target information is dynamically set according to the update frequency or importance of the data source. The system can give priority to pushing high-priority or frequently updated information (such as breaking news, emergency notifications), and at the same time suppress the push of redundant or outdated information, solving the problem that the importance of information is not quantified in the traditional push system, ensuring that users can obtain key information in time and reducing information overload.

[0027] Step S130, extracting the features of the target information from the target information and extracting the features of the user's historical behavior from the user's historical behavior data.

[0028] Step S140: Calculate the matching degree between the target information and the user according to the characteristics of the target information and the characteristics of the user's historical behavior.

[0029] In this embodiment, by extracting the characteristics of the target information and user behavior (such as keywords, interest tags, behavior patterns) and combining with the matching degree calculation, the relevance between the user's potential needs and the information can be more accurately identified. Compared with the rough push based on rules, this method reduces mis-pushes, improves the user's acceptance and click-through rate of the pushed content, and reduces the waste of resources for ineffective pushes.

[0030] Step S150: Determine whether to push the target information to the user according to the weight of the target information and the matching degree between the target information and the user.

[0031] In this embodiment, by combining the weight of the target information (reflecting the importance of the information) and the matching degree (reflecting the relevance of the user's interest), a multi-dimensional push decision model is constructed, which not only avoids pushing important information that the user is not interested in due to too high a weight, but also prevents pushing low-value content due to high matching degree but outdated information. The weight ratio of the weight and the matching degree can be adjusted to adapt to different scenario requirements (such as emphasizing the matching degree in e-commerce promotions and emphasizing the weight in news pushes).

[0032] Step S160: When the judgment result is yes, push the target information to the user.

[0033] According to the technical solution of this embodiment, through the dynamic weight mechanism, the feature-driven matching model and the multi-dimensional decision logic, while improving the push accuracy, timeliness and user experience, the efficient utilization of resources is realized. It not only solves the "one-size-fits-all" problem of the traditional push system, but also provides an extensible solution for personalized services and information distribution optimization, and is applicable to scenarios that require accurate information push such as e-commerce, social media, news media, and online education.

[0034] As Figure 2 shown, in an embodiment of the present invention, an information automatic push method is provided. Compared with the foregoing embodiments, in the information automatic push method of this embodiment, step S120 includes:

[0035] Step S210: Extract the number of information updates and the last update time of the data source within a preset time interval from the information update situation of the data source.

[0036] Step S220: Calculate the weight of the target information according to the number of information updates and the last update time of the data source within the preset time interval , where is the number of information updates of the data source within the time interval, It is a set of the number of information updates of a preset multiple data sources within a time interval. It is the number of information updates of any one of the multiple data sources within a time interval. It is a preset time decay factor. It is the current time. It is the last update time of the data source within the time interval.

[0037] In this embodiment, a dual dynamic control of the weight of the target information is realized. Compared with the static weight allocation, this design can more accurately reflect the timeliness of information and the activity of data sources, give priority to pushing high-update-frequency and recently updated high-quality content, and at the same time avoid the interference of extreme values through a smoothing function to ensure the rationality of weight allocation. For example, it avoids a certain data source monopolizing the push resources due to a single high-frequency update.

[0038] Among them, the time decay factor The calculation method can include:

[0039] (1) Obtain the historical time decay factor recorded last time .

[0040] (2) Count the number of information refreshes of the user within the time interval .

[0041] (3) Obtain the hot information within the time interval.

[0042] (4) Recalculate the time decay factor , where is a preset threshold of the number of refreshes, represents the target information, represents the set of hot information, represents any one of the hot information, represents the total number of hot information.

[0043] In this embodiment, by combining historical behavior data, user interaction frequency and hot information relevance, an intelligent adjustment of the time decay factor is realized, which can dynamically adjust the decay speed according to user behavior and current hotspots. For example, in the case of a breaking news event, the weight decay of hot information is slower, ensuring that users continuously pay attention to key events, and through automatic optimization of the formula, avoiding the complexity and lag of manual parameter adjustment, and improving the flexibility and robustness of the system.

[0044] According to the technical solution of this embodiment, through dynamic weight allocation, adaptive optimization of the time decay factor, and collaborative analysis of hot information and user behavior, the accuracy, timeliness, and system adaptability of information push are significantly improved. Its technical effect not only solves the problems of static weight allocation and insufficient timeliness in traditional methods, but also realizes a deep match between the pushed content and user needs by introducing dynamic feedback of user behavior and hot information. This design is particularly suitable for scenarios that need to balance global hotspots and personalized needs (such as news, e-commerce recommendations, social media), and can effectively improve user experience and system resource utilization.

[0045] In an embodiment of the present invention, an information automatic push method is provided. Compared with the foregoing embodiments, the information automatic push method of this embodiment, step S130, further includes:

[0046] (1) Identify the field to which the target information belongs.

[0047] (2) Obtain the corresponding knowledge base according to the field to which the target information belongs.

[0048] (3) Integrate the features of the target information with the knowledge base to update the features of the target information.

[0049] According to the technical solution of this embodiment, the original features of the target information may cause matching errors due to missing or ambiguous keywords, while the integration of the knowledge base can supplement implicit semantics and improve the expression ability of the features. At the same time, the knowledge bases for different fields can avoid cross-field interference (such as identifying "Apple" as a fruit or a technology company), improving the accuracy of the features.

[0050] In an embodiment of the present invention, an information automatic push method is provided. Compared with the foregoing embodiments, the information automatic push method of this embodiment, step S130 includes:

[0051] (1) Extract the dense vector of the target information as the spatial vector feature of the target information through the trained neural network, and extract the dense vector of the user's historical behavior data as the spatial vector feature of the user's historical behavior.

[0052] In this embodiment, by pre-training the neural network to extract the dense vectors of the target information and user behavior (such as the embedding vectors of text, images, and videos), the deep semantics and context relationships of multi-modal data can be captured. For example, the word vectors of text information can reflect semantics, and the convolutional features of images can capture visual content.

[0053] (2) Statistically analyze the discrete distribution of the target information as the probability distribution feature of the target information, and statistically analyze the discrete distribution of the user's historical behavior data as the probability distribution feature of the user's historical behavior.

[0054] In this embodiment, by statistically analyzing discrete distributions (such as the distributions of labels, categories, and behavior frequencies), the statistical attributes of information and user behavior are quantified, such as the preference probabilities of users for different categories or the label distribution of information.

[0055] As Figure 3 shown, step S140 includes:

[0056] Step S310, calculating the similarity degree between the target information and the historical behavior of the user according to the spatial vector features of the target information and the spatial vector features of the historical behavior of the user , where is the preset number of modalities, is the spatial vector feature of the -th modality in the spatial vector features of the target information, represents the historical behavior of the user, is the spatial vector feature of the -th modality in the spatial vector features of the historical behavior of the user, represents the target information, The function is the Jaccard function, which is used to calculate the similarity degree.

[0057] Step S320, calculating the difference degree between the target information and the historical behavior of the user according to the probability distribution features of the target information and the probability distribution features of the historical behavior of the user , where is the probability distribution feature of the -th modality in the probability distribution features of the target information, is the probability distribution feature of the -th modality in the probability distribution features of the historical behavior of the user, The function is used to calculate the Jensen-Shannon (JS) divergence.

[0058] Step S330, calculating the matching degree between the target information and the user according to the similarity degree and the difference degree between the target information and the historical behavior of the user , where is the preset smoothing factor, , are the preset first balance coefficient and second balance coefficient.

[0059] In this embodiment, the formula takes into account both semantic similarity and statistical difference, avoiding the one-sidedness of a single metric. For example, users may have different distribution preferences for information with similar semantics (such as preferring short videos over long texts). The weights of similarity and difference can be flexibly adjusted to meet the requirements of different scenarios (such as emphasizing similarity in e-commerce recommendations and balancing difference in news feeds to avoid information cocoons). The first balance coefficient and the second balance coefficient achieve interpretable matching degree control by quantifying similarity gain and difference loss, and their preset values need to be optimized in combination with business objectives and data distribution characteristics.

[0060] According to the technical solution of this embodiment, through multi-modal feature extraction, dynamic similarity and difference calculation, and parameterized matching degree formula, the accuracy, robustness, and scenario adaptability of information push are significantly improved. It not only solves the problems of insufficient single feature representation and poor dynamic scenario adaptability in traditional methods, but also realizes the complementarity of semantic and statistical characteristics through the combination of neural network and statistical model. This design is particularly suitable for scenarios with rich multi-modal content and complex and variable user behaviors (such as social media, e-commerce, news aggregation platforms), which can effectively improve the user experience, reduce push errors, and support flexible scenario-based parameter tuning.

[0061] As Figure 4 shown, in an embodiment of the present invention, an information automatic push method is provided. Compared with the foregoing embodiments, the information automatic push method of this embodiment, step S160 includes:

[0062] Step S410, converting non-text modal content in the target information into text modal.

[0063] Step S420, compressing the target information.

[0064] In this embodiment, the compressed target information , where is a preset decoding function, is a preset encoding function, is the original target information, is the similar information of the original target information. By superimposing the original target information with its similar information, the encoder can capture the core semantics and associated features, ensuring the integrity of the core semantics of the compressed target information.

[0065] In this embodiment, the target information is compressed by removing redundant information (such as repeated texts, low-value descriptions), simplifying expressions, or generating summaries, reducing the transmission and storage overhead.

[0066] Step S430, calculating the probability that each word in the target information is retained after compression, where the probability that the j-th word is retained after compression is , is the compressed target information.

[0067] Step S440, calculate the compression coefficient of the target information , where is the set of words included in the target information, is the length of the compressed target information.

[0068] In this embodiment, compression reduces the transmission bandwidth and storage cost, and is especially suitable for mobile devices or low-bandwidth scenarios. By quantifying the compression effect through the compression coefficient, it ensures that the pushed content retains the core information while being concise, and avoids information distortion caused by excessive compression. In this embodiment, the retention probability of each word after compression is evaluated through a probability model (such as a language model, an attention mechanism). High-probability words are usually semantic cores (such as "release" in "iPhone 15 release"). The compression process tends to retain high-probability words to ensure that the core information (such as events, entities, keywords) in the compressed text is not lost.

[0069] Step S450, determine whether the compression coefficient of the target information exceeds a preset threshold.

[0070] Step S460, when the judgment result is yes, execute the step of pushing the target information to the user.

[0071] Step S470, when the judgment result is no, return to the step of compressing the target information.

[0072] According to the technical solution of this embodiment, through multi-modal unified processing, dynamic compression and quality evaluation, and threshold-driven iterative optimization, the efficiency, quality and user experience of information pushing are significantly improved. Its technical effect not only solves the problems of difficult processing of multi-modal content and information distortion caused by compression in traditional methods, but also realizes the balance of resource optimization and content quality through a probability model and a dynamic threshold. This design is especially suitable for scenarios with large amounts of information and rich multi-modal content (such as social media, news aggregation platforms, e-commerce recommendations), can effectively reduce resource consumption, improve the accuracy of the pushed content, and support flexible scenario adaptation.

[0073] Such as Figure 5 shown, in an embodiment of the present invention, an information automatic pushing system is provided, including:

[0074] An information collection module 510, which collects the target information to be pushed through a preset data source and detects the information update situation of the data source.

[0075] In this embodiment, the type of the target information is not restricted. For example, it can be advertising information or course information in the field of green finance.

[0076] A weight setting module 520 sets weights for target information according to the information update situation of the data source.

[0077] In this embodiment, the weights of the target information are dynamically set according to the update frequency or importance of the data source. The system can give priority to pushing high-priority or frequently updated information (such as breaking news, emergency notifications), while suppressing the pushing of redundant or outdated information, solving the problem that the importance of information is not quantified in traditional pushing systems, ensuring that users can obtain key information in a timely manner and reducing information overload.

[0078] A feature extraction module 530 extracts the features of the target information from the target information and extracts the features of the user's historical behavior from the user's historical behavior data.

[0079] A matching degree calculation module 540 calculates the matching degree between the target information and the user according to the features of the target information and the features of the user's historical behavior.

[0080] In this embodiment, by extracting the features of the target information and the user behavior (such as keywords, interest tags, behavior patterns) and combining the matching degree calculation, the relevance between the potential needs of the user and the information can be more accurately identified. Compared with the rough pushing based on rules, this method reduces mis-pushing, improves the acceptance rate and click-through rate of the pushed content by users, and reduces the waste of resources for ineffective pushing.

[0081] A judgment module 550 judges whether to push the target information to the user according to the weight of the target information and the matching degree between the target information and the user.

[0082] In this embodiment, by combining the weight of the target information (reflecting the importance of the information) and the matching degree (reflecting the relevance of the user's interest), a multi-dimensional pushing decision model is constructed, which not only avoids pushing important information that the user is not interested in due to too high a weight, but also prevents pushing low-value content due to a high matching degree but outdated information. The weight ratio of the weight and the matching degree can be adjusted to adapt to different scenario requirements (such as emphasizing the matching degree in e-commerce promotions and emphasizing the weight in news pushing).

[0083] A pushing module 560 pushes the target information to the user when the judgment result is yes.

[0084] It should be noted that the information automatic pushing method provided by this application can also be applied to the pushing of advertising information or course information in the field of green finance, and can be used for the pushing of other courses, advertisements, and multimedia information.

[0085] According to the technical solution of this embodiment, through the dynamic weight mechanism, feature-driven matching model and multi-dimensional decision-making logic, while improving the push accuracy, timeliness and user experience, the efficient utilization of resources is achieved. It not only solves the "one-size-fits-all" problem of traditional push systems, but also provides an extensible solution for personalized services and information distribution optimization, and is applicable to scenarios such as e-commerce, social media, news media, and online education that require precise information push.

[0086] The basic principles of this application have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in this application are only examples and not limitations, and it cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of this application. In addition, the above-disclosed specific details are only for illustrative and easy-to-understand purposes, rather than limitations, and these details do not limit this application to necessarily adopt these specific details to implement.

[0087] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and do not intend to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open-ended words, meaning "including but not limited to", and can be used interchangeably with each other. The words "or" and "and" used herein refer to the word "and / or" and can be used interchangeably with each other, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to" and can be used interchangeably with each other.

[0088] It should also be noted that in the devices, equipment, and methods of this application, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent solutions of this application.

[0089] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects are very obvious to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

[0090] The above description has been given for purposes of illustration and description. In addition, this description does not intend to limit the embodiments of this application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions, and sub-combinations thereof.

Claims

1. An information automatic push method, comprising: Collecting target information to be pushed through a preset data source, and detecting the information update situation of the data source; Setting a weight for the target information according to the information update situation of the data source; Extracting the features of the target information from the target information, and extracting the features of the user's historical behavior from the user's historical behavior data; Calculating the matching degree between the target information and the user according to the features of the target information and the features of the user's historical behavior; Judging whether to push the target information to the user according to the weight of the target information and the matching degree between the target information and the user; When the judgment result is yes, pushing the target information to the user.

2. The information automatic push method according to claim 1, wherein, Setting a weight for the target information according to the information update situation of the data source, comprising: Extracting the information update times and the last update time of the data source within a preset time interval from the information update situation of the data source; Calculate the weight of the target information according to the number of information updates and the last update time of the data source within a preset time interval , where is the number of information updates of the data source within the time interval is the set of the number of information updates of a plurality of preset data sources within the time interval is the number of information updates of any one of the plurality of data sources within the time interval is a preset time decay factor is the current time is the last update time of the data source within the time interval 3. The information automatic push method according to claim 2, wherein, Before calculating the weight of the target information according to the information update times and the last update time of the data source within a preset time interval, the information automatic push method further comprises: Obtain the historical time decay factor of the most recent record ; Count the number of times the user's information is refreshed within the time interval ; Obtaining the hot information within the time interval; Recalculate the time decay factor , where is the threshold of the preset number of refreshes, represents the target information, represents the set of the hotspot information, represents any one of the hotspot information, represents the total number of the hotspot information.

4. The information automatic push method according to claim 1, wherein, Extracting the features of the target information from the target information, comprising: Identifying the field to which the target information belongs; Obtaining a corresponding knowledge base according to the field to which the target information belongs; Fusing the features of the target information with the knowledge base to update the features of the target information.

5. The information automatic push method according to claim 1, wherein, Extracting the features of the target information from the target information, and extracting the features of the user's historical behavior from the user's historical behavior data, comprising: Extracting a dense vector of the target information as the spatial vector feature of the target information through a trained neural network, and extracting a dense vector of the user's historical behavior data as the spatial vector feature of the user's historical behavior; Statistical discrete distribution of the target information as the probability distribution feature of the target information, and statistical discrete distribution of the user's historical behavior data as the probability distribution feature of the user's historical behavior.

6. The information automatic push method according to claim 5, wherein, Calculating the matching degree between the target information and the user according to the features of the target information and the features of the user's historical behavior, comprising: Calculate the similarity between the target information and the historical behavior of the user according to the spatial vector characteristics of the target information and the spatial vector characteristics of the historical behavior of the user , where is the preset number of modalities, is the spatial vector feature of the -th modality in the spatial vector characteristics of the target information, represents the historical behavior of the user, is the spatial vector feature of the -th modality in the spatial vector characteristics of the historical behavior of the user, represents the target information, is the Jaccard function for calculating the similarity; Calculate the degree of difference between the target information and the historical behavior of the user according to the probability distribution characteristics of the target information and the probability distribution characteristics of the historical behavior of the user , where is the probability distribution characteristic of the th modality in the probability distribution characteristics of the target information, is the probability distribution characteristic of the th modality in the probability distribution characteristics of the historical behavior of the user, and is used to calculate the JS divergence; Calculate the matching degree between the target information and the user according to the similarity degree and difference degree between the target information and the historical behavior of the user , where is a preset smoothing factor, , are the preset first balance coefficient and second balance coefficient respectively.

7. The information automatic push method according to claim 1, wherein, Pushing the target information to the user, comprising: Converting non-text modal content in the target information into text modal; Compressing the target information; Calculate the probability that each word in the target information is retained after compression, where the j-th word has a probability of being retained after compression of , is the target information after compression; Calculate the compression coefficient of the target information , where is the set of words included in the target information, is the length of the target information after compression; Determine the compression coefficient of the target information whether it exceeds a preset threshold; When the judgment result is yes, performing the step of pushing the target information to the user; When the judgment result is no, returning to the step of compressing the target information.

8. An information automatic push system, comprising: An information collection module, which collects target information to be pushed through a preset data source, and detects the information update situation of the data source; A weight setting module, which sets a weight for the target information according to the information update situation of the data source; A feature extraction module, which extracts the features of the target information from the target information, and extracts the features of the user's historical behavior from the user's historical behavior data; A matching degree calculation module calculates the matching degree between the target information and the user according to the characteristics of the target information and the characteristics of the user's historical behavior; A judgment module judges whether to push the target information to the user according to the weight of the target information and the matching degree between the target information and the user; A pushing module pushes the target information to the user when the judgment result is yes.

9. Application of the information automatic pushing method according to any one of claims 1 to 7 in the field of green finance.

Images