Cash flow data analysis method based on artificial intelligence

Abstract

The invention relates to the field of data analysis, in particular to a cash flow data analysis method based on artificial intelligence, and the method comprises the steps: generating labels according to cash flow data in a historical period, constructing a priori label library, analyzing the incidence relation between all kinds of labels, constructing a label pool for a user side, and periodically carrying out the correlation analysis of the label pool. Comprising the following steps: determining a label association coefficient according to an association relationship among various labels, verifying whether the validity of a label pool meets a standard or not, adaptively updating the label pool of a user side subsequently, generating a verification sub-label pool when the validity of the label pool does not meet the standard, and generating push information based on the verification sub-label pool. According to the method and the device, the accuracy of the labels in the label pool can be improved, so that the pushing accuracy of the pushing information is ensured.

Description

Technical Field

[0001] The present invention relates to the field of data analysis, and in particular to a cash flow data analysis method based on artificial intelligence. Background Art

[0002] Cash flow data includes key information such as the transaction object, transaction amount, and transaction goods during the user's transaction process. Collecting cash flow data under the conditions of big data acquisition permission has broad application value. For example, through artificial intelligence, the collected data can be analyzed to obtain user needs or set labels for users, so as to facilitate the subsequent targeted push of relevant recommended information to users.

[0003] China Patent Publication No.: CN109598540A, discloses a precise advertising push method and precise advertising push system, the precise advertising push method includes: parsing the associated transaction commodity information and transaction user account according to the commodity transaction data pre-stored on the blockchain platform; comparing the transaction commodity information with the pre-stored push information for similarity; if the similarity threshold is exceeded, the push information is pushed to the transaction user account associated with the transaction commodity information. The method and system of the present invention record the commodity transaction information according to the effectiveness of the commodity transaction smart contract on the blockchain platform and publish it to the blockchain platform, ensure the authenticity of the commodity transaction, and select the advertisement push object according to the commodity transaction information, which can achieve high efficiency of advertisement push.

[0004] However, there are still the following problems in the prior art:

[0005] In actual situations, when building user portraits or labels for the user side, the differences in behavioral habits of different users lead to inconsistent responses at the data level. In addition, due to the sporadic nature of user behavior, some labels with poor data representation may be introduced, resulting in inaccurate information push. Summary of the invention

[0006] To this end, the present invention provides a cash flow data analysis method based on artificial intelligence to overcome the problem in the prior art that when constructing user portraits or labels for the user side, the differences in behavior habits of different users lead to inconsistent responses at the data level, and due to the sporadic nature of user behavior, some labels with poor data representation are introduced, resulting in inaccurate information push.

[0007] To achieve the above objectives, the present invention provides a cash flow data analysis method based on artificial intelligence, comprising:

[0008] Obtain access permissions for cash flow data of several user terminals, generate labels based on the cash flow data in a historical period, and build a priori label library;

[0009] Determine the association between various tags based on the proportion of each tag in the prior tag library in all tags corresponding to the user terminal to which it belongs and the conditional probability that each tag belongs to the same user terminal;

[0010] Construct a tag pool for the user end, generate tags in real time based on the cash flow data updated by the user end and store them in the tag pool, and periodically perform association analysis on the tag pool, including determining the tag association coefficient based on the association relationship between various tags, so as to verify whether the validity of the tag pool meets the standards;

[0011] Based on the verification result of the validity of the tag pool, the tag pool of the user terminal is updated, including:

[0012] Call the tag to generate a verification sub-tag pool according to the association relationship between each tag, generate push information based on the verification sub-tag pool to perform verification push and collect the response data of the user end to the push information, and remove the tag in the corresponding verification sub-tag pool based on the response data;

[0013] Or, maintain the tags in the current tag pool;

[0014] Periodically generate push information based on the user's tag pool and push it to the user.

[0015] Furthermore, the process of generating labels based on cash flow data includes:

[0016] Pre-set several data features that need to be extracted from cash flow data;

[0017] Extracting several data features from the cash flow data, and generating corresponding labels according to each of the data features;

[0018] There is a one-to-one correspondence between each of the data features and the labels.

[0019] Further, determining whether there is a correlation between various tags includes:

[0020] Determine the average probability that each tag group belongs to the same user end in different periods;

[0021] Determine the proportion of each tag group in all tags corresponding to the user terminal to which it belongs;

[0022] If the association condition is met, it is determined that there is an association relationship between the corresponding category labels in the label group and the degree of association is determined;

[0023] The tag group includes two types of tags, the association condition is that the average probability is greater than a predetermined probability threshold and the proportion is greater than a predetermined proportion threshold, and the association degree is the average probability corresponding to the tag group.

[0024] Furthermore, the process of determining the label association coefficient according to the association relationship between various labels includes:

[0025] Determine the ratio of the number of tags with association relationships in the tag pool to the total number of tags;

[0026] Determine the mean value of the correlation between various tags that have correlation relationships;

[0027] The ratio of the proportion to a preset correlation proportion threshold is used as a first correlation feature;

[0028] The ratio of the correlation degree mean to a preset correlation degree threshold is used as a second correlation feature;

[0029] The sum of the first association feature and the second association feature is determined as a tag association coefficient.

[0030] Furthermore, the process of verifying whether the validity of the tag pool meets the standards includes:

[0031] If the tag association coefficient is greater than or equal to a preset tag association threshold, verifying that the validity of the tag pool meets the standard;

[0032] If the tag association coefficient is less than a preset tag association threshold, then it is verified that the validity of the tag pool does not meet the standard.

[0033] Further, updating the tag pool of the user terminal based on the verification result of the validity of the tag pool includes:

[0034] If the validity of the verification tag pool does not meet the standard, the tag is called to generate several verification sub-tag pools according to the association relationship between the tags, and push information is generated based on the verification sub-tag pool to perform verification push and collect the response data of the user end to the push information, and the tag in the corresponding verification sub-tag pool is removed based on the response data;

[0035] If the validity of the tag pool is verified to meet the criteria, the tags in the current tag pool are maintained.

[0036] Furthermore, the process of generating several verification sub-tag pools includes:

[0037] Identify irrelevant tags and potentially relevant tags;

[0038] After storing the unrelated tags and the potentially related tags, a verification sub-tag pool is generated;

[0039] The unrelated tag satisfies that there is no correlation relationship with the remaining tags, and the potential correlation tag satisfies that the correlation between the potential correlation tag and the remaining tags is greater than a preset potential correlation threshold.

[0040] Further, the process of generating push information based on the verification sub-tag pool for verification push includes:

[0041] Calling the unrelated tags and the potentially related tags in the verification sub-tag pool;

[0042] Generate corresponding push information based on unrelated tags and potentially related tags;

[0043] Push the generated push information to the user end;

[0044] Among them, the corresponding relationship between each tag and the required push information is preset.

[0045] Furthermore, the process of obtaining the user-side response data includes:

[0046] Determine the user's response data to the push information during multiple push verifications, including whether it is clicked and the browsing time;

[0047] Determine the number of valid clicks for each type of push information based on browsing time, and calculate the variance and mean of the number of valid clicks for each type of push information;

[0048] If the browsing time of a single click on the push information is greater than a predetermined browsing time threshold, it is determined to be a valid click.

[0049] Further, removing tags in the corresponding verification sub-tag pool based on the response data includes:

[0050] If the benchmark elimination condition is met, it is determined to eliminate all tags in the verification sub-tag pool;

[0051] If the benchmark elimination condition is not met, it is determined to eliminate the invalid verification tags in the verification sub-tag pool;

[0052] The benchmark elimination condition includes that the variance of the effective click count is greater than a predetermined variance threshold and the mean of the effective click count is less than a predetermined click count benchmark threshold;

[0053] The invalid verification tag satisfies that the number of valid clicks on the push information is less than a predetermined click reference threshold.

[0054] Compared with the prior art, the present invention generates tags based on cash flow data in a historical period, builds a priori tag library, analyzes the correlation between various tags, builds a tag pool for the user end, and periodically performs correlation analysis on the tag pool, including determining the tag correlation coefficient based on the correlation between various tags, verifying whether the validity of the tag pool meets the standard, and subsequently adaptively updating the tag pool of the user end. When the validity of the tag pool does not meet the standard, a verification sub-tag pool is generated, push information is generated based on the verification sub-tag pool, and tags in the sub-tag pool are removed according to the response data of the push information. The present invention can improve the accuracy of tags in the tag pool, thereby ensuring the accuracy of push information.

[0055] In particular, the present invention considers constructing a priori label library to determine the association relationship between various labels. In actual situations, the cash flow data of the user terminal has data characteristics, and corresponding labels can be generated. Due to individual differences, the cash flow data of different user terminals are different. However, for a single user terminal, the corresponding data characteristics between its cash flow data usually have potential associations. Therefore, the present invention considers the above characteristics and pre-constructs the association relationship between various labels through the attribution relationship of the labels in a specific priori label library, so as to provide data support for the subsequent verification of the user-specific label pool, so as to ensure the validity of the label pool, and then ensure the accuracy of subsequent information push.

[0056] In particular, the present invention considers periodically performing association analysis on the label pool. In actual situations, due to the unpredictability of individual behavior, when collecting cash flow data to determine labels, part of the cash flow data may be sporadic data, and the labels generated are poorly representative of the data on the user side. As the labels in the label pool accumulate, the labels generated by sporadic data are also accumulating, making the effectiveness of the label pool poor. Therefore, the present invention considers performing association analysis on the label pool and calculating the label association coefficient. The label association coefficient is calculated from data of two dimensions and comprehensively represents the association of the labels in the label pool. If the label association coefficient is low, too many labels with poor data representation are introduced, and the effectiveness of the label pool can be evaluated in a timely manner to ensure the accuracy of subsequent information push.

[0057] In particular, when the validity of the label pool does not meet the standards, the present invention specifically generates a verification sub-label pool, generates push information based on the verification sub-label pool for verification push, so as to selectively eliminate some labels. In actual situations, when the label pool does not meet the standards, there are many labels with poor data representativeness, so these labels are generated into a verification sub-label pool. At the same time, considering that there may be potential associations among these labels, push information is generated based on the entire verification sub-label pool, and verification push is performed. The labels in the verification sub-label pool are verified based on the response data, and then the label pool of the user end can be updated in time to ensure the validity of the label pool, and then the accuracy of subsequent information push is ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A step diagram of a cash flow data analysis method based on artificial intelligence according to an embodiment of the invention;

[0059] Figure 2 A logical determination diagram of whether there is an association relationship between various tags in an embodiment of the invention;

[0060] Figure 3 A logic decision diagram for verifying whether the validity of a tag pool meets the standard according to an embodiment of the invention;

[0061] Figure 4 This is a logical decision diagram for updating the tag pool of the user terminal based on the verification result of the validity of the tag pool according to an embodiment of the invention. DETAILED DESCRIPTION

[0062] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0063] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0064] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0065] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0066] See also Figures 1 to 4 As shown, Figure 1 This is a step diagram of a cash flow data analysis method based on artificial intelligence according to an embodiment of the invention. Figure 2 This is a logical determination diagram of whether there is an association relationship between various tags in the embodiment of the invention. Figure 3 A logic decision diagram for verifying whether the validity of a tag pool according to an embodiment of the invention meets the standard. Figure 4 In order to update the logic decision diagram of the tag pool of the user terminal based on the verification result of the validity of the tag pool according to the embodiment of the invention, the cash flow data analysis method based on artificial intelligence provided by the embodiment of the invention includes:

[0067] Step S1, obtaining access permissions for cash flow data of several user terminals, generating tags based on the cash flow data in a historical period, and constructing a priori tag library;

[0068] Step S2, determining the association relationship between various tags based on the proportion of various tags in the prior tag library in all tags corresponding to the user terminal to which they belong and the conditional probability that various tags belong to the same user terminal;

[0069] Step S3, constructing a tag pool for the user end, generating tags in real time based on the cash flow data updated by the user end and storing them in the tag pool, periodically performing association analysis on the tag pool, including determining the tag association coefficient according to the association relationship between various tags, so as to verify whether the validity of the tag pool meets the standard;

[0070] Step S4, based on the verification result of the validity of the tag pool, updates the tag pool of the user terminal, including:

[0071] Call the tag to generate a verification sub-tag pool according to the association relationship between each tag, generate push information based on the verification sub-tag pool to perform verification push and collect the response data of the user end to the push information, and remove the tag in the corresponding verification sub-tag pool based on the response data;

[0072] Or, maintain the tags in the current tag pool;

[0073] Step S5: periodically generate push information based on the tag pool of the user terminal and push it to the user terminal.

[0074] Specifically, the cash flow data includes the transaction objects and transaction commodities in the user transaction process. It can be understood that the cash flow data is multi-dimensional, in which both the transaction objects and transaction commodities can be used as data features.

[0075] Specifically, the process of generating labels based on cash flow data includes:

[0076] Pre-set several data features that need to be extracted from cash flow data;

[0077] Extracting several data features from the cash flow data, and generating corresponding labels according to each of the data features;

[0078] There is a one-to-one correspondence between each of the data features and the labels.

[0079] In some possible implementations, taking trading commodities as data features as an example, corresponding labels can be generated according to the categories of trading commodities. For example, trading commodities can be divided into furniture, daily necessities, books, hardware, etc. Of course, the categories can be further subdivided and then used as labels. Of course, it can also be any other form, such as classification based on the use field of the trading commodities, for example, not limited to field categories, mining categories, automotive categories, etc. Similarly, the categories can be further subdivided. It can be understood that the setting of labels is based on the needs of technical personnel in this field and can be changed, which will not be repeated here.

[0080] In some possible implementations, taking the transaction object as a data feature as an example, it can be divided according to the category of the transaction object, for example, official store, private store, personal transaction, etc. Similarly, the setting of the label is based on the needs of technical personnel in this field and can be changed, which will not be repeated here.

[0081] Specifically, determining whether there is a correlation between various tags includes:

[0082] Determine the average probability that each tag group belongs to the same user terminal in different periods. It can be understood that the average probability that the tag group belongs to the same user terminal can represent the probability that a single user terminal generates data features corresponding to the tags in the tag group in the data dimension, and then indirectly represent the possible association between the data features;

[0083] Determine the proportion of each tag group in all tags corresponding to the user terminal to which it belongs. It can be understood that the purpose of calculating the proportion is to filter out the tag group with a higher proportion to ensure the effectiveness of the tag group relative to the user terminal;

[0084] If the association condition is met, it is determined that there is an association relationship between the corresponding category labels in the label group and the degree of association is determined;

[0085] The tag group includes two types of tags, the association condition is that the average probability is greater than a predetermined probability threshold and the proportion is greater than a predetermined proportion threshold, and the association degree is the average probability corresponding to the tag group.

[0086] Specifically, the average probability that a tag group belongs to the same user terminal is the probability that various tags included in the tag group co-appear in the same user terminal, which will not be described in detail.

[0087] Specifically, the predetermined probability threshold and the predetermined proportion threshold are pre-set, and the cash flow data of several user terminals within the experimental period are collected in advance to determine the average probability that the tag group belongs to the same user terminal and the average proportion of each tag group of a single user terminal. The predetermined probability threshold is set to the product of the average probability and a first precision coefficient, and the predetermined proportion threshold is set to the product of the average proportion and a second precision coefficient. The first precision coefficient is selected in the interval [1.25, 1.5], and the second precision coefficient is selected in the interval [1.15, 1.3].

[0088] It is understandable that in order to ensure sample accuracy, the experimental cycle can be longer than the historical cycle corresponding to the construction of the prior label library. The historical cycle can be selected within one to two months, and the experimental cycle can be selected within three to five months.

[0089] Specifically, the present invention considers constructing a priori label library to determine the association relationship between various labels. In actual situations, the cash flow data of the user terminal has data characteristics, and corresponding labels can be generated. Due to individual differences, the cash flow data of different user terminals are different. However, for a single user terminal, the corresponding data characteristics between its cash flow data usually have potential associations. Therefore, the present invention considers the above characteristics and pre-constructs the association relationship between various labels through the attribution relationship of the labels in a specific priori label library, so as to provide data support for the subsequent verification of the user-specific label pool, so as to ensure the validity of the label pool, and then ensure the accuracy of subsequent information push.

[0090] Specifically, the process of determining the label correlation coefficient based on the correlation between various labels includes:

[0091] Determine the ratio of the number of tags with association relationships in the tag pool to the total number of tags;

[0092] Determine the mean value of the correlation between various tags that have correlation relationships;

[0093] The ratio of the proportion to a preset correlation proportion threshold is used as a first correlation feature;

[0094] The ratio of the correlation degree mean to a preset correlation degree threshold is used as a second correlation feature;

[0095] The sum of the first association feature and the second association feature is determined as a tag association coefficient.

[0096] Specifically, the preset association ratio threshold is pre-set, and a label pool for several user terminals is pre-built to solve the average ratio of the number of labels with association relationships in the label pool corresponding to each user terminal to the total number of labels, and the preset association ratio threshold is set to between 1.3 and 1.6 times the average ratio.

[0097] Specifically, the correlation threshold is preset, and the cash flow data of several user terminals within the experimental period are collected in advance, the tags with associated relationships are determined, the average correlation of each tag is solved, and the correlation threshold is set to between 1.25 and 1.5 times the average correlation.

[0098] Specifically, the process of verifying whether the validity of the tag pool meets the standards includes:

[0099] If the tag association coefficient is greater than or equal to a preset tag association threshold, verifying that the validity of the tag pool meets the standard;

[0100] If the tag association coefficient is less than a preset tag association threshold, then it is verified that the validity of the tag pool does not meet the standard.

[0101] Specifically, the label association threshold is selected in the interval [2.15, 2.3].

[0102] Specifically, the present invention considers periodically performing association analysis on the label pool. In actual situations, due to the unpredictability of individual behavior, when collecting cash flow data to determine labels, part of the cash flow data may be sporadic data, and the labels generated are poorly representative of the data on the user side. As the labels in the label pool accumulate, the labels generated by sporadic data are also accumulating, making the effectiveness of the label pool poor. Therefore, the present invention considers performing association analysis on the label pool and calculating the label association coefficient. The label association coefficient is calculated from data of two dimensions and comprehensively represents the association of the labels in the label pool. If the label association coefficient is low, too many labels with poor data representation are introduced, so that the effectiveness of the label pool can be evaluated in a timely manner to ensure the accuracy of subsequent information push.

[0103] Specifically, updating the tag pool of the user terminal based on the verification result of the validity of the tag pool includes:

[0104] If the validity of the verification tag pool does not meet the standard, the tag is called to generate several verification sub-tag pools according to the association relationship between the tags, and push information is generated based on the verification sub-tag pool to perform verification push and collect the response data of the user end to the push information, and the tag in the corresponding verification sub-tag pool is removed based on the response data;

[0105] If the validity of the tag pool is verified to meet the criteria, the tags in the current tag pool are maintained.

[0106] Specifically, the process of generating several verification sub-tag pools includes:

[0107] Identify irrelevant tags and potentially relevant tags;

[0108] After storing the unrelated tags and the potentially related tags, a verification sub-tag pool is generated;

[0109] The unrelated tag satisfies that there is no correlation relationship with the remaining tags, and the potential correlation tag satisfies that the correlation between the potential correlation tag and the remaining tags is greater than a preset potential correlation threshold.

[0110] Specifically, the potential association threshold is pre-set to 0.85 times the association threshold. It is understandable that the purpose of setting the potential association threshold is to determine tags that have a potential association but whose association with other tags does not reach the association threshold.

[0111] Specifically, the process of generating push information based on the verification sub-tag pool for verification push includes:

[0112] Calling the unrelated tags and the potentially related tags in the verification sub-tag pool;

[0113] Generate corresponding push information based on unrelated tags and potentially related tags;

[0114] Push the generated push information to the user end;

[0115] Among them, the corresponding relationship between each tag and the required push information is preset.

[0116] Specifically, in some possible implementations, the push information takes advertising as an example. For example, if the label is furniture, then furniture advertising information is pushed. It can be understood that the label and the push information are not in a one-to-one correspondence. A single label can correspond to multiple push information. Those skilled in the art can set it according to their needs, which will not be elaborated here.

[0117] Specifically, the process of obtaining the user-side response data includes:

[0118] Determine the user's response data to the push information during multiple push verifications, including whether it is clicked and the browsing time;

[0119] Determine the number of valid clicks for each type of push information based on browsing time, and calculate the variance and mean of the number of valid clicks for each type of push information;

[0120] If the browsing time of a single click on the push information is greater than a predetermined browsing time threshold, it is determined to be a valid click.

[0121] The browsing time can be determined based on the time spent on the corresponding push information page after clicking the push information, which will not be elaborated here.

[0122] Specifically, the browsing time threshold is intended to indicate that the user terminal browses the pushed information, and those skilled in the art may select it within the range of [2s, 5s].

[0123] Specifically, based on the response data, the tags in the corresponding verification sub-tag pool are removed,

[0124] If the benchmark elimination condition is met, it is determined to eliminate all tags in the verification sub-tag pool;

[0125] If the benchmark elimination condition is not met, it is determined to eliminate the invalid verification tags in the verification sub-tag pool;

[0126] The benchmark elimination condition includes that the variance of the effective click count is greater than a predetermined variance threshold and the mean of the effective click count is less than a predetermined click count benchmark threshold;

[0127] It is understandable that when the benchmark elimination conditions are met, the user's preference for various types of push information is relatively discrete, and the correlation between the labels in the verification sub-label pool is weak. They may all be labels generated by relatively occasional cash flow data on the user side, and therefore are eliminated accordingly.

[0128] In the implementation, the predetermined variance threshold is pre-set, wherein verification push is performed for user terminals in a valid tag pool for several tag pools, including generating push information based on the tags in the valid tag pool, recording response data, determining the mean variance of the number of valid clicks for each type of push information, and setting the predetermined variance threshold to between 1.25 and 1.5 times the mean variance of the number of valid clicks.

[0129] In practice, the click count benchmark threshold is determined based on the number of push verifications, and is usually set between 0.3 and 0.5 times the number of push verifications.

[0130] The invalid verification tag satisfies that the number of valid clicks on the push information is less than a predetermined click reference threshold.

[0131] It is understandable that if the number of valid clicks on the push information is less than the predetermined click reference threshold, it indicates that the user terminal has a weaker inclination towards this type of push information, and therefore the corresponding label is removed.

[0132] Specifically, when the validity of the label pool does not meet the standards, the present invention specifically generates a verification sub-label pool, generates push information based on the verification sub-label pool for verification push, so as to selectively remove some labels. In actual situations, when the label pool does not meet the standards, there are many labels with poor data representativeness, so these labels are generated into a verification sub-label pool. At the same time, considering that there may be potential associations among these labels, push information is generated based on the entire verification sub-label pool, and verification push is performed. The labels in the verification sub-label pool are verified based on the response data, and then the label pool of the user end can be updated in time to ensure the validity of the label pool, and then the accuracy of subsequent information push is ensured.

[0133] If the cash flow data analysis method based on artificial intelligence of the present invention is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention, and the aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0134] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A cash flow data analysis method based on artificial intelligence, characterized in that: include: Obtain access permissions for cash flow data of several user terminals, generate labels based on the cash flow data in a historical period, and build a priori label library; Determine the association between various tags based on the proportion of each tag in the prior tag library in all tags corresponding to the user terminal to which it belongs and the conditional probability that each tag belongs to the same user terminal; Construct a tag pool for the user end, generate tags in real time based on the cash flow data updated by the user end and store them in the tag pool, and periodically perform association analysis on the tag pool, including determining the tag association coefficient based on the association relationship between various tags, so as to verify whether the validity of the tag pool meets the standards; Based on the verification result of the validity of the tag pool, the tag pool of the user terminal is updated. include, Call the tag to generate a verification sub-tag pool according to the association relationship between each tag, generate push information based on the verification sub-tag pool to perform verification push and collect the response data of the user end to the push information, and remove the tag in the corresponding verification sub-tag pool based on the response data; Or, maintain the tags in the current tag pool; Periodically generate push information based on the user's tag pool and push it to the user.

2. The cash flow data analysis method based on artificial intelligence according to claim 1 is characterized in that: The process of generating labels based on cash flow data includes: Pre-set several data features that need to be extracted from cash flow data; Extracting several data features from the cash flow data, and generating corresponding labels according to each of the data features; There is a one-to-one correspondence between each of the data features and the labels.

3. The cash flow data analysis method based on artificial intelligence according to claim 2 is characterized in that: Determine whether there is a correlation between various tags, including: Determine the average probability that each tag group belongs to the same user end in different periods; Determine the proportion of each tag group in all tags corresponding to the user terminal to which it belongs; If the association condition is met, it is determined that there is an association relationship between the corresponding category labels in the label group and the degree of association is determined; The tag group includes two types of tags, the association condition is that the average probability is greater than a predetermined probability threshold and the proportion is greater than a predetermined proportion threshold, and the association degree is the average probability corresponding to the tag group.

4. The cash flow data analysis method based on artificial intelligence according to claim 3 is characterized in that: The process of determining the label association coefficient based on the association relationship between various labels includes: Determine the ratio of the number of tags with association relationships in the tag pool to the total number of tags; Determine the mean value of the correlation between various tags that have correlation relationships; The ratio of the proportion to a preset correlation proportion threshold is used as a first correlation feature; The ratio of the correlation degree mean to a preset correlation degree threshold is used as a second correlation feature; The sum of the first association feature and the second association feature is determined as a tag association coefficient.

5. The cash flow data analysis method based on artificial intelligence according to claim 1 is characterized in that: The process of verifying whether the validity of the tag pool meets the standards includes: If the tag association coefficient is greater than or equal to a preset tag association threshold, verifying that the validity of the tag pool meets the standard; If the tag association coefficient is less than a preset tag association threshold, then it is verified that the validity of the tag pool does not meet the standard.

6. The cash flow data analysis method based on artificial intelligence according to claim 1 is characterized in that: Updating the user terminal's tag pool based on the verification result of the tag pool validity includes: If the validity of the verification tag pool does not meet the standard, the tag is called to generate several verification sub-tag pools according to the association relationship between the tags, and push information is generated based on the verification sub-tag pool to perform verification push and collect the response data of the user end to the push information, and the tag in the corresponding verification sub-tag pool is removed based on the response data; If the validity of the tag pool is verified to meet the criteria, the tags in the current tag pool are maintained.

7. The cash flow data analysis method based on artificial intelligence according to claim 1 is characterized in that: The process of generating several verification sub-tag pools includes: Identify irrelevant tags and potentially relevant tags; After storing the unrelated tags and the potentially related tags, a verification sub-tag pool is generated; The unrelated tag satisfies that there is no correlation relationship with the remaining tags, and the potential correlation tag satisfies that the correlation between the potential correlation tag and the remaining tags is greater than a preset potential correlation threshold.

8. The cash flow data analysis method based on artificial intelligence according to claim 7 is characterized in that: The process of generating push information based on the verification sub-tag pool for verification push includes: Calling the unrelated tags and the potentially related tags in the verification sub-tag pool; Generate corresponding push information based on unrelated tags and potentially related tags; Push the generated push information to the user end; Among them, the corresponding relationship between each tag and the required push information is preset.

9. The cash flow data analysis method based on artificial intelligence according to claim 1 is characterized in that: The process of obtaining the user-side response data includes: Determine the user's response data to the push information during multiple push verifications, including whether it is clicked and the browsing time; Determine the number of valid clicks for each type of push information based on browsing time, and calculate the variance and mean of the number of valid clicks for each type of push information; If the browsing time of a single click on the push information is greater than a predetermined browsing time threshold, it is determined to be a valid click.

10. The cash flow data analysis method based on artificial intelligence according to claim 1 is characterized in that: Based on the response data, the tags in the corresponding verification sub-tag pool are removed, including: If the benchmark elimination condition is met, it is determined to eliminate all tags in the verification sub-tag pool; If the benchmark elimination condition is not met, it is determined to eliminate the invalid verification tags in the verification sub-tag pool; The benchmark elimination condition includes that the variance of the effective click count is greater than a predetermined variance threshold and the mean of the effective click count is less than a predetermined click count benchmark threshold; The invalid verification tag satisfies that the number of valid clicks on the push information is less than a predetermined click reference threshold.

Images