A Smart Matching Method and System Based on Association Rule Mining Algorithm

Abstract

The present application discloses an intelligent matching method and system based on an association rule mining algorithm, and the present application relates to the technical field of intelligent matching algorithms; an update analysis is performed on the electronic products updated in the intelligent recommendation matching platform through the received associated tags to generate a collection instruction, and the collection of the electronic products under each associated tag and their corresponding product data is controlled, thereby realizing an automated collection process; through feature word overlap analysis, browsing concentration analysis and browsing interval time analysis, the feature word concentration value, browsing concentration value and browsing frequency value are obtained, and a comprehensive analysis is performed on them to generate a preference evaluation index, thereby quantifying the degree of matching of each browsing tag to the user's preference; through natural language processing technology and TF-IDF algorithm, the keywords of the user's search terms can be accurately identified, and a comprehensive analysis is performed with the user's associated tags to obtain a matching value, and electronic products are pushed to the user based on the matching value, thereby realizing intelligent matching of electronic products.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent matching algorithms, and in particular to an intelligent matching method and system based on an association rule mining algorithm. Background Art

[0002] With the development and popularization of network technology, more and more people choose to obtain information and purchase electronic products through online shopping. The Internet provides a wealth of product information, user reviews and comparison functions, allowing consumers to conveniently understand the functional characteristics of electronic products at home or anywhere and make purchasing decisions. This convenience and the immediacy of information have significantly promoted people's online shopping behavior for electronic products.

[0003] At present, the market of electronic products is changing rapidly, and a large number of electronic products are updated and replaced, which requires users to spend more time and energy to find specific electronic products, which undoubtedly reduces the user experience;

[0004] In order to solve the above defects, an intelligent matching method based on association rule mining algorithm is provided. Summary of the invention

[0005] According to one aspect of the present application, there is provided an intelligent matching method based on an association rule mining algorithm, the method comprising the following steps:

[0006] R1: Collect user data; in addition, update analysis is performed on the updated electronic products through the received associated tags to determine whether the electronic products with associated tags need to be updated, and a collection instruction is generated;

[0007] R2: The association mining module mines association rules on user data to obtain the user's association tags and the preference evaluation index corresponding to the association tags, and sends them to R1 and R3;

[0008] R3: A matching value is obtained by comprehensively analyzing the search terms input by the user and the user's associated tags, and electronic products are pushed to the user based on the matching value; the specific matching steps are:

[0009] Step 1: When receiving the search term input by the user, the natural language processing technology is used to extract the keywords of the search term, and then the TF-IDF algorithm is used to calculate the keywords and the labels, feature words and product introductions of all electronic products in the database for comparison and analysis to obtain the importance score between each electronic product and the keyword. When the importance score is greater than the set score threshold, the electronic product is recorded as a preliminary product; thus, the preliminary products and the importance scores corresponding to the preliminary products can be obtained;

[0010] Step 2: Retrieve the associated tags corresponding to the user and the preference evaluation index corresponding to each associated tag, randomly select one of the preliminary selected products, and compare the preliminary selected product with the user's associated tags. When all the tags corresponding to the preliminary selected product belong to the associated tags, the total number of tags corresponding to the preliminary selected products is counted, and recorded as the number of fully overlapping tags Q1, and the preference evaluation index corresponding to the fully overlapping tags is averaged to obtain the full overlapping value recorded as AH1; when part of the tags corresponding to the preliminary selected product is the same as the associated tags, the total number of tags corresponding to the preliminary selected products is counted as Q2 and the number of tags that are the same as the associated tags, and recorded as the number of semi-overlapping tags Q3, and the preference evaluation index corresponding to the semi-overlapping tags is averaged to obtain the semi-overlapping value recorded as AH2; when all the tags corresponding to the preliminary selected products do not belong to the associated tags, the total number of tags corresponding to the preliminary selected products is counted, and recorded as the number of isolated tags Q4; use the set formula

[0011] Calculate and obtain the correlation value Qr between the primary selected product and the user's associated label, where r1, r2, and r3 are the set proportional coefficients respectively;

[0012] Step 3: Substitute the importance score Zf and the correlation value Qr of the preliminary selected product into the set formula ZQ=r4×Zf+r5×Qr to calculate the matching value ZQ of the preliminary selected product, where r4 and r5 are the set proportional coefficients respectively; sort the preliminary selected products in order according to their corresponding matching values ​​to obtain a recommendation list, and send it to the user end according to the recommendation list.

[0013] Optionally, the generation process of the acquisition instruction is specifically as follows:

[0014] Select the electronic products with associated tags from the updated electronic products and record them as associated products, and count the number of associated tags in the tags of each associated product and record them as Ni, where i = 1, 2, 3 ... I, I is a positive integer, I represents the total number of associated products, and i represents the serial number of any associated product; calculate the total number of tags corresponding to each associated product and record them as Bi; substitute Ni and Bi into the set formula Calculate and obtain the related value NBi of the associated product, where a1 and a2 are the set proportional coefficients respectively;

[0015] The correlation value of the associated product is compared with the set correlation interval to obtain the cumulative value of the associated product. When the cumulative value of the associated product is greater than the set cumulative threshold, a collection instruction is generated to control the collection of electronic products under each associated label and their corresponding product data, where the product data includes the labels involved in the electronic products, the feature words involved under each label and the product introduction.

[0016] Optionally, the calculation process of the cumulative value of the associated products is:

[0017] Compare and analyze the correlation values ​​of the associated products with the set correlation intervals to classify the associated products corresponding to the correlation values ​​into high correlation, medium correlation and low correlation, and count the cumulative times of high correlation, medium correlation and low correlation respectively, and record them as G1, G2, and G3 respectively; sum the correlation values ​​corresponding to high correlation, medium correlation and low correlation respectively to obtain high correlation value, medium correlation value and low correlation value, and record them as G4, G5, and G6 respectively;

[0018] Substitute G1, G2, G3, G4, G5, G6 into the set formula

[0019] The cumulative value Ga of the associated products is calculated, where b1, b2, b3, a3, a4, and a5 are respectively set proportional coefficients, and a3>a4>a5>0.

[0020] Optionally, the specific steps of mining association rules for user data are:

[0021] 401: Obtain user data, where the data includes the number of user browsing times, the browsing start time and the browsing end time corresponding to each browsing, and the label and feature word corresponding to the electronic product;

[0022] 402: Classify the browsed electronic products according to their corresponding tags to obtain a number of browsed tags, the electronic products under each browsed tag, and the browsing start time and browsing end time of each electronic product; and perform feature word overlap analysis based on this to obtain a feature word concentration value;

[0023] 403: Calculate the time difference between the browsing start time and the browsing end time of each electronic product under the browsing tag to obtain the browsing duration, and perform browsing concentration analysis based on the time difference to obtain the browsing concentration value;

[0024] 404: extracting the browsing start time and browsing end time of each electronic product under each browsing tag, marking them on the time horizontal axis to obtain a time trace graph, and performing browsing frequency analysis based on the time trace graph to obtain a browsing frequency value;

[0025] 405: The characteristic word concentration value Ad, the browsing concentration value Ld and the browsing frequency value Hd of the browsing tag are calculated by the set formula AH=f1×Ad+f2×Ld+f3×Hd to obtain the preference evaluation index AH, wherein f1, f2, f3 are respectively the set proportional coefficients; thereby, the preference evaluation index corresponding to each browsing tag can be obtained, and when the preference evaluation index is greater than the set preference threshold, the browsing tag corresponding to the preference evaluation index is recorded as an associated tag; thereby, the user's associated tag and the preference evaluation index corresponding to the associated tag can be obtained, and they are sent to the intelligent matching module and the database respectively.

[0026] Optionally, the specific process of feature word overlap analysis is:

[0027] Select any one of the browsing tags and its corresponding electronic products, extract the feature words corresponding to the browsing tags of the electronic products, when there is overlap of feature words under two browsing tags, the number of overlaps corresponding to the feature words increases once, thus the number of overlaps of each feature word is recorded as Aj, where J = 1, 2, 3...J, J is a positive integer, J represents the total number of different feature words, and j represents the serial number of any feature word; using the set formula Calculate and get the characteristic word concentration value Ad, where d1 and d2 are the set proportional coefficients respectively. is the average number of overlaps of each feature word.

[0028] Optionally, the specific process of browsing concentration analysis is:

[0029] The browsing time is calculated by calculating the time difference between the browsing start time and the browsing end time of each electronic product under the browsing tag, and the browsing time is compared with the set time interval to divide the browsing tags corresponding to the browsing time into high-concentration browsing, medium-concentration browsing and low-concentration browsing, and the cumulative number of high-concentration browsing, medium-concentration browsing and low-concentration browsing is counted respectively, and they are recorded as L1, L2, and L3 respectively; the browsing time corresponding to high-concentration browsing, medium-concentration browsing and low-concentration browsing is summed up to obtain high-concentration value, medium-concentration value and low-concentration value, and they are recorded as L4, L5, and L6 respectively;

[0030] Substitute L1, L2, L3, L4, L5, L6 into the set formula

[0031]

[0032] The browsing concentration value Ld of each browsing tag is calculated, wherein d3, d4, and d5 are respectively set proportional coefficients, and d3>d4>d5>0.

[0033] Optionally, the specific process of browsing frequency analysis is:

[0034] Extract the browsing start time and browsing end time of each electronic product under each browsing tag, and mark them on the time horizontal axis to obtain a time trace graph; calculate the browsing interval time of two adjacent electronic products, and mark them in sequence according to their positions on the time trace graph, thereby obtaining the number on each browsing interval time; use the number as the horizontal axis and the browsing interval time as the vertical axis to obtain a browsing interval time change curve graph; draw a tangent line at the interval point, use data fitting to obtain a tangent line expression, and perform a derivative operation on the tangent line expression to obtain a tangent line derivative at the interval point;

[0035] The browsing reduction degree is calculated by summing up the tangent derivatives greater than zero and recorded as H1, and the browsing increase degree is calculated by summing up the tangent derivatives less than zero and taking the absolute value and recorded as H2; the average of each browsing interval is calculated to get the interval mean and recorded as H3; the browsing reduction degree H1, browsing increase degree H2 and interval mean H3 are substituted into the set formula The browsing frequency value Hd is calculated, where d6, d7, and d8 are respectively set proportional coefficients.

[0036] According to one aspect of the present application, a smart matching system based on an association rule mining algorithm is provided, the system comprising: a smart recommendation matching platform and a user terminal; a built-in database, an association mining module and a smart matching module of the smart recommendation matching platform; the database collects user data; in addition, an update analysis is performed on the updated electronic products through the received association tags to determine whether the electronic products with the association tags need to be updated, and a collection instruction is generated;

[0037] The association mining module mines association rules on user data to obtain the user's association tags and the preference evaluation index corresponding to the association tags, and sends them to the database and intelligent matching module;

[0038] The intelligent matching module obtains a matching value by comprehensively analyzing the search terms entered by the user and the user's associated tags, and pushes electronic products to the user based on the matching value.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] (1) The updated electronic products in the intelligent recommendation and matching platform are updated and analyzed through the received associated tags to generate a collection index, and the electronic products under each associated tag and their corresponding product data are collected, thereby realizing an automated collection process, ensuring that the intelligent recommendation and matching platform can always adapt to the upgrading of electronic products, and improving the timeliness and practicality of the intelligent recommendation and matching platform;

[0041] (2) Through feature word overlap analysis, browsing concentration analysis and browsing frequency analysis, the feature word concentration value, browsing concentration value and browsing frequency value are obtained, and they are comprehensively analyzed to generate a preference evaluation index, thereby quantifying the degree of matching of each browsing tag to the user's preference; the user's browsing behavior can be accurately mapped to the relevant tags. This personalized tag classification can help the system understand the user's interests and hobbies more accurately, thereby providing more accurate electronic product recommendation matching;

[0042] (3) Through natural language processing technology and TF-IDF algorithm, it can accurately identify the keywords of user search terms and conduct a comprehensive analysis with the user's associated tags to obtain matching values. Electronic products can be pushed to users based on the matching values. This not only realizes real-time personalized recommendations, but also continuously optimizes the algorithm and recommendation results through user feedback data. The data-driven optimization mechanism can maintain the flexibility and response speed of the system, ensuring the continuous improvement of recommendation strategies and the continuous improvement of user experience. It not only saves users' search time, but also improves users' experience on the platform and increases user satisfaction and loyalty. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0044] Figure 1 is a flowchart of the method of the present invention;

[0045] Figure 2 It is a system module block diagram of the present invention. DETAILED DESCRIPTION

[0046] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0047] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0048] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0049] It should be noted that the relevant information (including but not limited to user data, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties. Figure 1-2 As shown, the embodiment of the present application provides an intelligent matching method based on an association rule mining algorithm, which is implemented based on an intelligent matching system based on an association rule mining algorithm, including: an intelligent recommendation matching platform and a user end; a built-in database, an association mining module and an intelligent matching module of the intelligent recommendation matching platform;

[0050] The method comprises the following steps:

[0051] R1: The database collects user data by communicating with the user end. In addition, the database also updates and analyzes the updated electronic products in the intelligent recommendation matching platform through the received associated tags to determine whether the electronic products with associated tags need to be updated; specifically:

[0052] Select the electronic products with associated tags from the updated electronic products and record them as associated products, and count the number of associated tags in the tags of each associated product and record them as Ni, where i = 1, 2, 3 ... I, I is a positive integer, I represents the total number of associated products, and i represents the serial number of any associated product; calculate the total number of tags corresponding to each associated product and record them as Bi; it can be seen that Ni ≤ Bi; using the set formula Calculate and obtain the correlation value NBi of the associated product, where a1 and a2 are the set proportional coefficients respectively. It can be seen from the formula that when Ni is larger and Bi is smaller, it means that the proportion of the associated label in the label of the associated product is larger, and the corresponding correlation value is larger. Thus, the correlation value of each associated product can be obtained;

[0053] Compare and analyze the correlation value of the associated product with the set correlation interval. When the correlation value is greater than the upper limit of the set correlation interval, it means that there is a relatively large correlation between the associated product corresponding to the correlation value and the associated tag value of the user, and a high correlation is accumulated; when the correlation value is in the set correlation interval, a medium correlation is accumulated; when the correlation value is less than the lower limit of the set correlation interval, a low correlation is accumulated; the cumulative number of high correlation, medium correlation and low correlation is counted respectively, and they are recorded as G1, G2, and G3 respectively; the correlation values ​​corresponding to high correlation, medium correlation and low correlation are summed up to obtain high correlation value, medium correlation value and low correlation value, and they are recorded as G4, G5, and G6 respectively; use the set formula The cumulative value Ga of the associated products is calculated, where b1, b2, b3, a3, a4, and a5 are respectively the set proportional coefficients, and a3>a4>a5>0; it can be seen from the formula that the greater the correlation between the associated tags of the updated electronic products and the user (that is, the greater the correlation value, the greater the number of highly correlated ones), it means that the updated electronic products need to be collected;

[0054] When the cumulative value of the associated products is greater than the set cumulative threshold, a collection instruction is generated to control the collection of electronic products under each associated tag and their corresponding product data, where the product data includes the tags involved in the electronic products, the feature words involved under each tag, and the product introduction;

[0055] The updated electronic products in the intelligent recommendation and matching platform are updated and analyzed through the received associated tags to generate collection indicators, and the collection of electronic products under each associated tag and their corresponding product data is controlled, thereby realizing an automated collection process, ensuring that the intelligent recommendation and matching platform can always adapt to the upgrading of electronic products, and improving the timeliness and practicality of the intelligent recommendation and matching platform.

[0056] R2: The association mining module obtains the user's associated tags by mining association rules on user data; specifically: obtaining user data, where the data includes the number of user browsing times and the browsing start time and end time corresponding to each browsing, as well as the tags and feature words corresponding to the electronic products; it should be noted that each electronic product corresponds to several tags, each tag includes several feature words; each electronic product involves multiple tags, each tag is associated with multiple feature words, and multiple tags and multiple feature words are used to limit the electronic products at the same time, so as to quickly locate the electronic products and lay the foundation for matching with the appropriate user end;

[0057] The browsed electronic products are classified according to their corresponding tags. The specific classification method is: those with the same tag are classified into one category, and the same tag is used as the browsing tag, and the electronic products browsed by the user are classified into each browsing tag; thus, several browsing tags, electronic products under each browsing tag, and the browsing start time and browsing end time of each electronic product can be obtained; any one of the browsing tags and its corresponding electronic products are selected, and the feature words corresponding to the browsing tags of several electronic products are extracted. When there is overlap of feature words under two browsing tags, the overlap number corresponding to the feature word increases once, and thus the overlap number of each feature word is recorded as Aj, where J=1, 2, 3...J, J is a positive integer, J represents the total number of different feature words, and j represents the serial number of any feature word; using the set formula Calculate and get the characteristic word concentration value Ad, where d1 and d2 are the set proportional coefficients respectively. is the average value of the overlap times of each feature word; it should be noted that when the feature word concentration value in the browsing tag is larger, it means that the relevance of each electronic product in the browsing tag is greater, and the browsing tag is more in line with the user's preference;

[0058] The browsing time is calculated by taking the time difference between the browsing start time and the browsing end time of each electronic product under the browsing tag, and compared with the set time interval. When the browsing time is greater than the upper limit of the set time interval, a high-concentration browsing is accumulated; when the browsing time is within the set time interval, a medium-concentration browsing is accumulated; when the browsing time is less than the upper limit of the set time interval, a low-concentration browsing is accumulated; the cumulative number of high-concentration browsing, medium-concentration browsing and low-concentration browsing is counted respectively, and they are recorded as L1, L2, and L3 respectively; the browsing time corresponding to high-concentration browsing, medium-concentration browsing and low-concentration browsing is summed up to obtain the high-concentration value, medium-concentration value and low-concentration value, and they are recorded as L4, L5, and L6 respectively; using the set formula

[0059]

[0060] Calculate and obtain the browsing concentration value Ld of each browsing tag, where d3, d4, and d5 are respectively set proportional coefficients, and d3>d4>d5>0;

[0061] Extract the browsing start time and browsing end time of each electronic product under each browsing tag, and mark them on the time horizontal axis to obtain a time trace graph; calculate the browsing interval time of two adjacent electronic products, and mark them in sequence according to their positions on the time trace graph, thereby obtaining the number on each browsing interval time; construct a two-dimensional rectangular coordinate system with the number as the horizontal coordinate and the browsing interval time as the vertical coordinate, input the browsing interval time into the coordinate axis according to its corresponding number, and record the position of the browsing interval time on the coordinate axis as the interval point, and use a smooth curve to connect the interval points in sequence to obtain a browsing interval time change curve graph; draw a tangent line at the interval point, and use data fitting to obtain the tangent line Expression, the tangent expression is differentiated to obtain the tangent derivative at the interval point; it should be noted that when the tangent derivative is greater than zero, it means that the browsing interval at the interval point shows an increasing trend, and when the tangent derivative is less than zero, it means that the browsing interval at the interval point shows a decreasing trend. Therefore, when there are more tangent derivatives less than zero, it means that the browsing tag is browsed more frequently, and the browsing tag is more in line with the user's browsing preference; the tangent derivatives greater than zero are summed up to obtain the browsing reduction degree, which is recorded as H1, and the tangent derivatives less than zero are summed up and the absolute value is taken to obtain the browsing increase degree, which is recorded as H2; the average of each browsing interval is calculated to obtain the interval average, which is recorded as H3; using the set formula The browsing frequency value Hd is calculated, where d6, d7, and d8 are the set proportional coefficients respectively;

[0062] The characteristic word concentration value Ad, the browsing concentration value Ld and the browsing frequency value Hd of the browsing tag are calculated by the set formula AH=f1×Ad+f2×Ld+f3×Hd to obtain the preference evaluation index AH, where f1, f2, and f3 are respectively set proportional coefficients; thereby, the preference evaluation index corresponding to each browsing tag can be obtained, and it is compared and analyzed with the preference threshold. When the preference evaluation index is greater than the set preference threshold, the browsing tag corresponding to the preference evaluation index is recorded as an associated tag; thereby, the user's associated tag and the preference evaluation index corresponding to the associated tag can be obtained, and they are sent to the intelligent matching module and the database respectively;

[0063] Through feature word overlap analysis, browsing concentration analysis and browsing interval time analysis, the feature word concentration value, browsing concentration value and browsing frequency value are obtained, and they are comprehensively analyzed to generate a preference evaluation index, thereby quantifying the degree of matching of each browsing tag to the user's preference; it can accurately map the user's browsing behavior to the relevant tags. This personalized tag classification can help the system understand the user's interests and hobbies more accurately, thereby providing more accurate electronic product recommendation matching. R3: The intelligent matching module comprehensively analyzes the search terms entered by the user and the user's associated tags to obtain matching values, and pushes electronic products to the user based on the matching values, thereby realizing intelligent matching of electronic products and quickly helping users find specific electronic products; specifically:

[0064] Step 1: When receiving the search term input by the user, the natural language processing (NLP) technology is used to extract the keywords of the search term, and the TF-IDF algorithm (term frequency-inverse document frequency algorithm) is used to calculate the keywords and the labels, feature words and product introductions of all electronic products in the database for comparison analysis to obtain the importance score between each electronic product and the keyword, and the importance score is compared with the set score threshold for analysis. When the importance score is greater than the set score threshold, the electronic product is recorded as a preliminary product; thus, the preliminary products and the importance scores corresponding to the preliminary products are recorded as Zf;

[0065] Step 2: Retrieve the associated tags corresponding to the user and the preference evaluation index corresponding to each associated tag, randomly select one of the preliminary selected products, and compare the preliminary selected product with the user's associated tags. When all the tags corresponding to the preliminary selected product belong to the associated tags, the total number of tags corresponding to the preliminary selected products is counted, and recorded as the number of fully overlapping tags Q1, and the preference evaluation index corresponding to the fully overlapping tags is averaged to obtain the full overlapping value recorded as AH1; when part of the tags corresponding to the preliminary selected product is the same as the associated tags, the total number of tags corresponding to the preliminary selected products is counted as Q2 and the number of tags that are the same as the associated tags, and recorded as the number of semi-overlapping tags Q3, and the preference evaluation index corresponding to the semi-overlapping tags is averaged to obtain the semi-overlapping value recorded as AH2; when all the tags corresponding to the preliminary selected products do not belong to the associated tags, the total number of tags corresponding to the preliminary selected products is counted, and recorded as the number of isolated tags Q4; use the set formula

[0066] Calculate and obtain the correlation value Qr between the primary selected product and the user's associated label, where r1, r2, and r3 are the set proportional coefficients respectively;

[0067] Step 3: Calculate the importance score Zf and the correlation value Qr of the preliminary selected products by the set formula ZQ=r4×Zf+r5×Qr to obtain the matching value ZQ of the preliminary selected products, where r4 and r5 are the set proportional coefficients respectively; sort the preliminary selected products according to their corresponding matching values ​​to obtain a recommendation list, and send it to the user end according to the recommendation list;

[0068] Through natural language processing technology and TF-IDF algorithm, it is possible to accurately identify the keywords of user search terms, and conduct a comprehensive analysis with the user's associated tags to obtain matching values. According to the matching values, electronic products are pushed to users, which not only realizes real-time personalized recommendations, but also continuously optimizes the algorithm and recommendation results through user feedback data. The data-driven optimization mechanism can maintain the flexibility and response speed of the system, ensure the continuous improvement of recommendation strategies and the continuous improvement of user experience; it not only saves users' search time, but also improves users' experience on the platform, and increases user satisfaction and loyalty. It should also be noted that the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device that includes a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, commodity or device that includes the elements.

[0069] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. An intelligent matching method based on association rule mining algorithm, characterized in that: The following steps are involved: Database R1: collects user data; in addition, updates and analyzes the updated electronic products through the received associated tags to determine whether the electronic products with associated tags need to be updated, and generates collection instructions; Association mining module R2: It mines association rules on user data to obtain the user's association tags and the preference evaluation index corresponding to the association tags, and sends them to R1 and R3; Intelligent matching module R3: by comprehensively analyzing the search terms input by the user and the user's associated tags to obtain matching values, electronic products are pushed to the user based on the matching values; the specific matching steps are: Step 1: When receiving the search term input by the user, the natural language processing technology is used to extract the keywords of the search term, and then the TF-IDF algorithm is used to calculate the keywords and the labels, feature words and product descriptions of all electronic products in the database for comparison and analysis to obtain the importance score between each electronic product and the keyword, and the electronic products with importance scores greater than the set score threshold are recorded as preliminary selected products; thus, the preliminary selected products and the importance scores corresponding to the preliminary selected products can be obtained; Step 2: retrieve the associated tags corresponding to the user and the preference evaluation index corresponding to each associated tag, randomly select one of the preliminary selected products, and compare the preliminary selected product with the user's associated tags. When all the tags corresponding to the preliminary selected products belong to the associated tags, count the total number of tags corresponding to the preliminary selected products and record it as the number of fully overlapping tags Q1. Calculate the average of the preference evaluation indexes corresponding to the fully overlapping tags to obtain the fully overlapping value, which is recorded as AH1. When part of the labels corresponding to the initial selected products are the same as the associated labels, the total number of labels corresponding to the initial selected products is counted and recorded as Q2, and the number of labels that are the same as the associated labels is recorded as the number of semi-overlapping labels Q3. The preference evaluation index corresponding to the semi-overlapping labels is averaged to obtain the semi-overlapping value, which is recorded as AH2. When all the labels corresponding to the initial selected products do not belong to the associated labels, the total number of labels corresponding to the initial selected products is counted and recorded as the number of isolated labels Q4. Using the set formula Calculate and obtain the correlation value Qr between the primary selected product and the user's associated label, where r1, r2, and r3 are the set proportional coefficients respectively; Step 3: normalize the importance scores and associated values ​​of the preliminary selected products and obtain their numerical values, analyze the numerical values ​​to obtain the matching values ​​of the preliminary selected products, sort the preliminary selected products according to their corresponding matching values ​​to obtain a recommendation list, and send the recommendation list to the user end; The specific steps for mining association rules for user data are as follows: 401: Obtain user data, where the data includes the number of user browsing times, the browsing start time and the browsing end time corresponding to each browsing, and the label and feature word corresponding to the electronic product; 402: Classify the browsed electronic products according to their corresponding tags to obtain a number of browsed tags, the electronic products under each browsed tag, and the browsing start time and browsing end time of each electronic product; and perform feature word overlap analysis based on this to obtain a feature word concentration value; 403: Calculate the time difference between the browsing start time and the browsing end time of each electronic product under the browsing tag to obtain the browsing duration, and perform browsing concentration analysis based on the time difference to obtain the browsing concentration value; 404: extracting the browsing start time and browsing end time of each electronic product under each browsing tag, marking them on the time horizontal axis to obtain a time trace graph, and performing browsing frequency analysis based on the time trace graph to obtain a browsing frequency value; 405: The characteristic word concentration value, the browsing concentration value and the browsing frequency value of the browsing tag are normalized and their values ​​are taken, and the preference evaluation index is obtained by numerical analysis, thereby obtaining the preference evaluation index corresponding to each browsing tag. When the preference evaluation index is greater than the set preference threshold, the browsing tag corresponding to the preference evaluation index is recorded as an associated tag; thereby, the user's associated tag and the preference evaluation index corresponding to the associated tag are obtained, and they are sent to the intelligent matching module and the database respectively.

2. The intelligent matching method based on association rule mining algorithm according to claim 1 is characterized in that: The specific process of generating the acquisition instruction is as follows: Select electronic products with associated tags from the updated electronic products and record them as associated products, and count the number of associated tags in the tags of each associated product and record them as Ni, where i=1,2,3...I, I is a positive integer, I represents the total number of associated products, and i represents the serial number of any associated product; calculate the total number of tags corresponding to each associated product and record them as Bi; perform formula calculation and analysis on Ni and Bi to obtain the correlation value of the associated product; The correlation value of the associated product is compared with the set correlation interval to obtain the cumulative value of the associated product. When the cumulative value of the associated product is greater than the set cumulative threshold, a collection instruction is generated to control the collection of electronic products under each associated label and their corresponding product data, where the product data includes the labels involved in the electronic products, the feature words involved under each label and the product introduction.

3. The intelligent matching method based on association rule mining algorithm according to claim 2 is characterized in that: The calculation process of the cumulative value of associated products is: Comparing and analyzing the correlation values ​​of the associated products with the set correlation intervals to classify the associated products corresponding to the correlation values ​​into high correlation, medium correlation and low correlation, respectively counting the cumulative times of high correlation, medium correlation and low correlation, and summing up the correlation values ​​corresponding to the high correlation, medium correlation and low correlation to obtain the high correlation value, medium correlation value and low correlation value; The cumulative number of highly correlated values, the cumulative number of moderately correlated values, the cumulative number of lowly correlated values, the highly correlated values, the moderately correlated values ​​and the lowly correlated values ​​are normalized and their numerical values ​​are taken, and the cumulative value of the associated products is obtained by numerical analysis.

4. The intelligent matching method based on association rule mining algorithm according to claim 1 is characterized in that: The specific process of feature word overlap analysis is as follows: Select any one of the browsing tags and its corresponding electronic products, extract the feature words corresponding to the browsing tags of the electronic products, when there is overlap of feature words under two browsing tags, the number of overlaps corresponding to the feature words increases once, thus the number of overlaps of each feature word is recorded as Aj, where J=1, 2, 3...J, J is a positive integer, J represents the total number of different feature words, and j represents the serial number of any feature word; using the set formula Calculate and get the characteristic word concentration value Ad, where d1 and d2 are the set proportional coefficients respectively. is the average number of overlaps of each feature word.

5. The intelligent matching method based on association rule mining algorithm according to claim 4 is characterized in that: The specific process of browsing concentration analysis is as follows: The browsing start time and the browsing end time of each electronic product under the browsing tag are calculated by the time difference to obtain the browsing time, and the time difference is compared with the set time interval to analyze and classify the browsing tags corresponding to the browsing time into high-concentration browsing, medium-concentration browsing and low-concentration browsing, and the cumulative number of high-concentration browsing, medium-concentration browsing and low-concentration browsing is counted respectively, and the browsing time corresponding to high-concentration browsing, medium-concentration browsing and low-concentration browsing is summed up to obtain the high-concentration value, medium-concentration value and low-concentration value respectively; The cumulative number of high-concentration browsing, the cumulative number of medium-concentration browsing, the cumulative number of low-concentration browsing, the high concentration value, the medium concentration value and the low concentration value are calculated and analyzed by formulating to obtain the browsing concentration value of each browsing tag.

6. The intelligent matching method based on association rule mining algorithm according to claim 5 is characterized in that: The specific process of browsing frequency analysis is as follows: Extract the browsing start time and browsing end time of each electronic product under each browsing tag, and mark them on the time horizontal axis to obtain a time trace graph; calculate the browsing interval time of two adjacent electronic products, and mark them in sequence according to their positions on the time trace graph, thereby obtaining the number on each browsing interval time; use the number as the horizontal axis and the browsing interval time as the vertical axis to obtain a browsing interval time change curve graph; draw a tangent line at the interval point, use data fitting to obtain a tangent line expression, and perform a derivative operation on the tangent line expression to obtain a tangent line derivative at the interval point; The tangent derivatives greater than zero are summed up to obtain the browsing reduction, and the tangent derivatives less than zero are summed up and the absolute value is taken to obtain the browsing increase; the average of each browsing interval is calculated to obtain the interval mean; The browsing frequency value is obtained by formulating and analyzing the browsing decrease, browsing increase and interval mean.

7. An intelligent matching system based on association rule mining algorithm, characterized by An intelligent matching method based on an association rule mining algorithm as described in any one of claims 1 to 6, the system comprising: an intelligent recommendation matching platform and a user end, wherein the intelligent recommendation matching platform has a built-in database, an association mining module and an intelligent matching module; The database collects user data, and also performs update analysis on the updated electronic products through the received associated tags to determine whether the electronic products with the associated tags need to be updated, and generates a collection instruction; The association mining module mines association rules on user data to obtain the user's association tags and the preference evaluation index corresponding to the association tags, and sends them to the database and intelligent matching module; The intelligent matching module obtains a matching value by comprehensively analyzing the search terms entered by the user and the user's associated tags, and pushes electronic products to the user based on the matching value.

Images