A cloud-based agricultural input information management system

Abstract

This invention relates to the field of data analysis, and more particularly to a cloud-based agricultural input information management system, comprising: a profile analysis unit for determining whether to perform target analysis on target users based on historical search status; a data analysis unit for determining the confirmation method of recommendation coefficient based on the coverage balance value of each term cluster; a cluster update unit for determining whether each term cluster should undergo association update analysis based on periodic status; an information push unit for determining recommended information based on recommendation coefficient under a first preset recommendation condition and recording the term clusters of related users as term clusters of target users under a second preset recommendation condition; and a cloud storage unit for storing the search terms and term clusters corresponding to each target user. This invention effectively improves the accuracy of agricultural input information recommendation management.

Description

Technical Field

[0001] This invention relates to the field of data analysis, and in particular to an agricultural input information management system based on a cloud platform. Background Technology

[0002] As the internet continues to extend into agriculture and rural areas, various agricultural input e-commerce platforms have become important channels for purchasing production materials. Therefore, accurate product recommendations are a core element in improving the service quality and user satisfaction of agricultural input e-commerce platforms. In-depth analysis based on users' actual search behavior is a key step in optimizing recommendation systems and achieving personalized services. However, existing technologies typically rely on single static tags for coarse recommendations. This approach struggles to dynamically capture the changing true intentions of users, easily leading to recommendations that do not match actual user needs, resulting in poor agricultural input product recommendations. Therefore, how to fully mine user data with limited historical data and achieve real-time adjustments to recommendation strategies for target users is a key technical challenge that urgently needs to be solved by those skilled in the art.

[0003] Chinese Patent Publication No. CN108876562A discloses a product recommendation method and apparatus for an agricultural input e-commerce platform, comprising: obtaining at least one first associated tag group with corresponding seasonal and geographical intervals based on the user's geographic information and access time; obtaining at least one corresponding first information tag group based on the usage tag, category tag, and crop tag in the first associated tag group; filtering a corresponding product information list from a product library based on the first information tag group; sorting the product information list according to a preset product sorting mechanism to obtain a product recommendation ranking, and providing feedback to the user. It is evident that while the above technical solution discloses assigning different types of tags to agricultural input products and filtering and pushing products through tag groups, it fails to consider the platform's need to provide differentiated recommendation strategies based on user characteristics during the actual recommendation process and fails to optimize and adjust the corresponding recommendation strategies in a timely manner based on the actual recommendation results, which can easily lead to poor recommendation accuracy for agricultural input information. Summary of the Invention

[0004] To address this, the present invention provides a cloud-based agricultural input information management system to overcome the shortcomings of existing technologies that fail to consider providing differentiated recommendation strategies based on user characteristics during the actual recommendation process and fail to optimize and adjust the corresponding recommendation strategies in a timely manner based on the actual recommendation results, which easily leads to poor recommendation accuracy of agricultural input information.

[0005] To achieve the above objectives, the present invention provides an agricultural input information management system based on a cloud platform, comprising:

[0006] The profile analysis unit is used to determine the target user's historical search status based on the diversity and number of search terms, and to determine whether to conduct target analysis on the target user based on the historical search status.

[0007] The data analysis unit, which is connected to the portrait analysis unit, is used to determine the recommendation coefficient based on the comparison result of the coverage balance value and the coverage balance value threshold of each term cluster. The process is to determine the recommendation coefficient corresponding to each term cluster based on the coverage ratio, or to determine the recommendation coefficient corresponding to each term cluster based on the search frequency characterization value.

[0008] The cluster update unit is connected to the portrait analysis unit and the data analysis unit respectively. It is used to determine the periodic status based on the periodic search fit and the recommendation usage rate, and to determine whether to perform association update analysis on the term cluster based on the periodic status.

[0009] The information push unit is connected to the profile analysis unit, the data analysis unit and the cluster update unit respectively. It is used to obtain agricultural input products according to the recommendation coefficient, and to recommend the obtained agricultural input products in a loop and to record the word clusters of associated users as the word clusters of target users.

[0010] The cloud storage unit is connected to the portrait analysis unit, data analysis unit, cluster update unit and information push unit respectively, and is used to store the search terms and term clusters corresponding to each target user.

[0011] Furthermore, the profile analysis unit responds to preset analysis conditions and performs target analysis on target users whose historical search status is that the search term diversity is greater than the search term diversity threshold and the number of search terms is greater than the preset number of search terms;

[0012] The preset analysis condition is that the current time is within a loop analysis node.

[0013] Furthermore, the user profile analysis unit performs target analysis on the target user, including:

[0014] For each search term, related search terms are determined in descending order of usage frequency, and the search term and its corresponding related search terms are recorded as a term cluster;

[0015] The associated search terms are determined based on the similarity of the terms' corresponding tag names and the relevance of the terms to the search results.

[0016] Furthermore, the data analysis unit responds to preset filtering conditions to detect the coverage of each term cluster and obtain the coverage balance value;

[0017] If the coverage balance value is greater than the coverage balance value threshold, the recommendation coefficient corresponding to each term cluster is determined based on the coverage ratio.

[0018] If the coverage balance value is less than or equal to the coverage balance value threshold, the recommendation coefficient corresponding to each term cluster is determined based on the search frequency characterization value.

[0019] The preset filtering conditions are used to complete the target analysis for the target users.

[0020] Furthermore, the data analysis unit determines the coverage balance threshold based on the number of term clusters, and the coverage balance threshold is positively correlated with the number of term clusters.

[0021] Furthermore, the cluster update unit responds to preset update conditions to perform update analysis for each term cluster, including:

[0022] For target users whose periodic search relevance is less than the preset periodic search relevance and whose recommended usage rate is less than the preset recommended usage rate, a correlation update analysis is performed.

[0023] For target users whose periodic search fit is greater than or equal to the preset periodic search fit or whose recommended usage rate is greater than or equal to the preset recommended usage rate, it is determined that no optimization or update is needed.

[0024] The preset update condition is that the current time is in an update node.

[0025] Furthermore, the cluster update unit performs relational update analysis, including:

[0026] Based on the descending order of similarity in the search tree diagram, a preset number of associated users are obtained, and the priority recommendation information of each associated user is entered.

[0027] Furthermore, the cluster update unit determines the search tree diagram similarity based on the anchored similarity reference value and the extended similarity reference value;

[0028] The search tree diagram similarity is positively correlated with both the anchored similarity reference value and the extended similarity reference value.

[0029] Furthermore, in response to the first preset recommendation condition, the information push unit sequentially obtains the corresponding agricultural input products from the cluster of terms with recommendation coefficients greater than the preset recommendation coefficient in descending order, and then cyclically recommends the obtained agricultural input products.

[0030] The first preset recommendation condition is the completion of the determination process for setting the recommendation coefficient.

[0031] Furthermore, in response to the second preset recommendation conditions, the information push unit simultaneously records the term clusters of associated users as the term clusters of the target user;

[0032] The second preset recommendation condition is that the correlation update analysis is completed.

[0033] Compared with the prior art, the beneficial effects of the present invention are as follows: the technical solution of the present invention determines related search terms based on the similarity of term tags and the relevance of term searches, reducing the possibility that a single influencing factor will lead to poor accuracy of related search terms. Furthermore, it determines related search terms for each search term in descending order of hot usage, and the hot usage reflects the timeliness of the target user's needs corresponding to the search terms. By effectively filtering related search terms through hot usage, it avoids the problem of wasting resources due to invalid analysis of a large amount of data, thereby reducing the availability of subsequent term clusters and improving the accuracy of subsequent data analysis.

[0034] Furthermore, in the technical solution of this invention, the recommendation coefficient corresponding to each term cluster is determined based on the comparison result between the coverage balance value and the coverage balance value threshold, or based on the search frequency characterization value. The coverage balance value reflects the distribution balance of search terms among term clusters, and different recommendation coefficient setting processes are selected accordingly. The distribution balance of search terms further reflects the difference in the contribution value of term clusters to information recommendation selection, avoiding the problem of inaccurate recommendation coefficient setting caused by a single recommendation coefficient confirmation method, and improving the accuracy of subsequent product recommendations.

[0035] Furthermore, in this invention, the periodic status of the target user is determined based on the periodic search fit and the recommendation usage rate. The periodic status is then used to determine whether an association update analysis is needed. The periodic search fit and recommendation usage rate reflect the matching degree between the target user and the system recommendations. Considering the strong regional and seasonal characteristics of agricultural inputs in actual application scenarios, cluster update analysis is needed based on the matching degree between the target user and the system recommendations during actual use. This avoids the problem in existing technologies where fixed update node settings cannot meet the cluster update needs in actual scenarios, leading to poor cluster optimization results. In the association update analysis, this invention uses different filtering methods for associated users based on the number of associated users, effectively improving the filtering accuracy of associated users and thus improving the accuracy of agricultural input platform recommendations.

[0036] Furthermore, in the technical solution of this invention, when making relevant pushes based on the recommendation coefficient, the agricultural input products corresponding to the cluster of terms with recommendation coefficients greater than the preset recommendation coefficient are obtained in descending order. This avoids indiscriminate product recommendations that lead to unsatisfactory recommendation results, thereby reducing recommendation effectiveness and improving the accuracy of agricultural input platform recommendations. Attached Figure Description

[0037] Figure 1 This is a unit connection diagram of the cloud-based agricultural input information management system of the present invention;

[0038] Figure 2 This is a flowchart illustrating the process of determining whether to perform target analysis based on historical search status in this invention.

[0039] Figure 3 This is a flowchart illustrating the method for determining the recommendation coefficient based on the coverage balance value of term clusters in this invention. Detailed Implementation

[0040] To make the objectives and advantages of this invention clearer, the invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0041] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0042] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0043] Please see Figures 1 to 3 As shown, the present invention provides an agricultural input information management system based on a cloud platform, comprising:

[0044] The profile analysis unit is used to determine the target user's historical search status based on the diversity and number of search terms, and to determine whether to conduct target analysis on the target user based on the historical search status.

[0045] The data analysis unit, which is connected to the portrait analysis unit, is used to determine the recommendation coefficient based on the comparison result of the coverage balance value and the coverage balance value threshold of each term cluster. The process is to determine the recommendation coefficient corresponding to each term cluster based on the coverage ratio, or to determine the recommendation coefficient corresponding to each term cluster based on the search frequency characterization value.

[0046] The cluster update unit is connected to the portrait analysis unit and the data analysis unit respectively. It is used to determine the periodic status based on the periodic search fit and the recommendation usage rate, and to determine whether to perform association update analysis on the term cluster based on the periodic status.

[0047] The information push unit is connected to the profile analysis unit, the data analysis unit and the cluster update unit respectively. It is used to obtain agricultural input products according to the recommendation coefficient, and to recommend the obtained agricultural input products in a loop and to record the word clusters of associated users as the word clusters of target users.

[0048] The cloud storage unit is connected to the portrait analysis unit, data analysis unit, cluster update unit and information push unit respectively, and is used to store the search terms and term clusters corresponding to each target user.

[0049] The application scenario of this invention is to optimize the recommendation and management of agricultural input information on agricultural input platforms. This invention sets a continuous cyclical monitoring cycle, wherein the duration of a single monitoring cycle can be adaptively set by the user according to actual application needs. This invention provides a value for the duration of a single monitoring cycle, which is 1 month.

[0050] This invention also utilizes several historical records. Each historical record includes at least the search term diversity, number of search terms, weight adjustment, relevance coefficient, time interval, recommendation coefficient, coverage ratio, search frequency characterization value, number of term clusters, coverage balance value, periodic search fit, recommendation usage rate, number of associated users, anchored similarity reference value, and extended similarity reference value for a single historical process. Each historical record is also recorded with a corresponding qualification mark, which indicates whether the historical record meets the user's needs. The use of self-defined indicators (e.g., obtaining user satisfaction by sending a satisfaction survey questionnaire) to determine whether the historical record meets the user's needs is a concept already known to those skilled in the art and will not be elaborated here.

[0051] Specifically, the profile analysis unit responds to preset analysis conditions and performs target analysis on target users whose historical search status is that the search term diversity is greater than the search term diversity threshold and the number of search terms is greater than the preset number of search terms;

[0052] The preset analysis condition is that the current time is within a loop analysis node.

[0053] Historical search states where the search term diversity is greater than the search term diversity threshold and the number of search terms is greater than the preset number of search terms are recorded as the first preset historical search state. Historical search states where the search term diversity is less than or equal to the search term diversity threshold or the number of search terms is less than or equal to the preset number of search terms are recorded as the second preset historical search state. It is understood that users in the second preset historical search state are not relevant and do not require target analysis.

[0054] The cyclical analysis node is the moment corresponding to the end of the most recent monitoring period. The target user is the user who has completed registration on the agricultural input platform. For a single target user, the total number of search terms entered by the target user on the agricultural input platform in the most recent monitoring period is recorded as the number of search terms for the target user. The term tags corresponding to the search terms are extracted. The search term diversity = 1 - the number of duplicate term tags / the total number of term tags.

[0055] During the process of target users searching for agricultural input products through an agricultural input platform, the name of the agricultural input product that the target user wants to search for is the search term. In this invention, both agricultural input products and search terms are agricultural input product names. Each agricultural input product name is pre-labeled with corresponding tags by the operators. The tag includes, but is not limited to, related crops, pests and diseases, product type, product form, and price range. Each agricultural input product name corresponds to at least one tag. For example, for rice stem borer pesticide, the corresponding tags are related crop "rice", pest "stem borer", product type "insecticide", product form "liquid", and price range "20-50 yuan". For example, for wheat seeds, the corresponding tags are related crop "wheat", product type "crop seeds", product form "solid", and price range "50-90 yuan". The operators are the management personnel of the agricultural input platform.

[0056] Users can adaptively set the search term diversity threshold and preset search term number based on their actual application needs. Understandably, the higher the user's requirements for the data analysis efficiency of the platform's information recommendation optimization process, the larger the search term diversity threshold and the preset search term number will be. One method is to extract the search term diversity and search term number corresponding to the historical records that meet the user's needs, remove outliers from the search term diversity and search term number, and record the average values ​​of the search term diversity and search term number after removing outliers as the search term diversity threshold and the preset search term number, respectively. The methods for removing outliers include, but are not limited to, the 3σ criterion or the IQR method.

[0057] Specifically, the user profile analysis unit performs target analysis on the target user, including:

[0058] For each search term, related search terms are determined in descending order of usage frequency, and the search term and its corresponding related search terms are recorded as a term cluster;

[0059] The associated search terms are determined based on the similarity of the terms' corresponding tag names and the relevance of the terms to the search results.

[0060] For any search term, the corresponding hotspot usage is determined by detecting the number of platform searches corresponding to that search term between the two most recent cyclic analysis nodes, which is recorded as the hotspot usage of that search term. The number of platform searches refers to the number of times a user enters and submits the search term in the search box of the agricultural materials platform. When determining related search terms for a single search term, the search term is recorded as the target search term. The correlation coefficient between the target search term and other search terms corresponding to the target user is detected. Search terms with a correlation coefficient greater than a preset correlation coefficient are recorded as related search terms corresponding to the target search term. It is worth noting that if a search term has already been included in the term cluster, it is not necessary to determine the related search terms for that search term.

[0061] The correlation coefficient is determined based on the similarity of the terms' tags to the search terms and the relevance of the search terms. The correlation coefficient is calculated as: α1 × term tag similarity + α2 × term search relevance. Here, α1 is the first weight coefficient, α2 is the second weight coefficient, and α1 + α2 = 1. Users can set the values ​​of α1 and α2 directly based on their domain experience, or use statistical methods, such as regression analysis or principal component analysis, to determine the contribution of term tag similarity and term search relevance to the correlation coefficient, thereby determining the corresponding weight coefficient values. The greater the contribution, the larger the weight coefficient value. The weights can also be adjusted through training based on historical records (such as machine learning). This invention provides one set of values: α1 = 0.6, α2 = 0.4.

[0062] The preset relevance coefficient value can be adaptively set by the user according to the actual application needs. It can be understood that the higher the user's requirements for the relevance of each search term within the term cluster, the larger the preset relevance coefficient value will be. The relevance coefficient corresponding to the historical records that meet the user's needs is extracted, outliers in the relevance coefficient are removed, and the average value of the relevance coefficient after removing outliers is recorded as the preset relevance coefficient value.

[0063] For each search term, the corresponding term tags are obtained by constructing a knowledge graph in the agricultural input field. For any two search terms, the similarity of their corresponding term tags is calculated as the number of identical term tags / the average number of term tags. The average number of term tags is the average number of term tags corresponding to the two search terms.

[0064] Any two search terms are considered a search combination. For any search combination, the corresponding term search relevance is determined as follows: Term search relevance = Number of search intervals between two search terms that are less than a preset search interval / Number of search combinations. The calculation method for the number of search combinations is easily understood by those skilled in the art and will not be elaborated here. The value of the preset time interval can be adaptively set by the user according to actual application needs. It is understood that the higher the user's requirement for the relevance of search terms within the term cluster, the smaller the value of the preset time interval. One value is provided, and the time intervals corresponding to the historical records that meet the user's needs are extracted. Outliers in the time intervals are removed, and the average value of the time intervals after removing outliers is recorded as the value of the preset time interval.

[0065] Specifically, the data analysis unit responds to preset filtering conditions to detect the coverage of each term cluster and obtain the coverage balance value;

[0066] If the coverage balance value is greater than the coverage balance value threshold, the recommendation coefficient corresponding to each term cluster is determined based on the coverage ratio.

[0067] If the coverage balance value is less than or equal to the coverage balance value threshold, the recommendation coefficient corresponding to each term cluster is determined based on the search frequency characterization value.

[0068] The preset filtering conditions are used to complete the target analysis for the target users.

[0069] The coverage balance value is determined as follows: for a single term cluster, the number of terms corresponding to it is recorded as the coverage of that term cluster, and the coverage balance value is... ;in, The number of term clusters for the target user. For the first The coverage of a cluster of terms. This represents the average coverage of the term cluster.

[0070] When determining the recommendation coefficient for each term cluster based on the search frequency representation value, the recommendation coefficient = baseline recommendation coefficient × coverage percentage / preset coverage percentage; extract the coverage of each term cluster in the most recent monitoring period, and for a single term cluster, its corresponding coverage percentage = coverage of that term cluster / coverage of all term clusters corresponding to the target user.

[0071] When determining the recommendation coefficient for each term cluster based on the search frequency characterization value, the recommendation coefficient = baseline recommendation coefficient × search frequency characterization value / preset search frequency characterization value; wherein, the search frequency characterization value is determined by extracting the search count of search terms within each term cluster in the most recent monitoring period. For a single term cluster, its corresponding search frequency characterization value is the average of the search counts of search terms within that term cluster. It can be understood that the search terms include target search terms and related search terms.

[0072] The value of the baseline recommendation coefficient can be adaptively set by the user according to the actual application needs. It can be understood that the higher the user's requirement for the accuracy of information recommendation, the larger the value of the baseline recommendation coefficient. One value is provided by extracting the recommendation coefficient corresponding to the historical records that meet the user's needs, removing outliers of the recommendation coefficient, and recording the average value of the recommendation coefficient after removing outliers as the value of the baseline recommendation coefficient. In this invention, the baseline recommendation coefficient is 8.5.

[0073] Users can adaptively set the preset coverage percentage and preset search frequency representation values ​​according to their actual application needs. It is understandable that the greater the user's tolerance for the impact of the preset coverage percentage and preset search frequency representation values ​​on the recommendation coefficient, the larger the preset coverage percentage and preset search frequency representation values ​​will be. One set of values ​​is provided, and the coverage percentage and search frequency representation values ​​corresponding to the historical records that meet the user's needs are extracted respectively. Outliers in the coverage percentage and search frequency representation values ​​are removed, and the average of the coverage percentage and search frequency representation values ​​after removing outliers is recorded as the preset coverage percentage and preset search frequency representation values, respectively.

[0074] Specifically, the data analysis unit determines the coverage balance threshold based on the number of term clusters, and the coverage balance threshold is positively correlated with the number of term clusters.

[0075] Coverage balance threshold = baseline coverage balance value × (number of term clusters / preset number of term clusters); for a single target user, the number of term clusters in the most recent monitoring period is recorded as the term cluster number for that target user. Determining the coverage balance threshold based on the number of term clusters improves the applicability of the coverage balance threshold, and dynamic adjustment of the coverage balance threshold ensures accurate switching of the recommendation coefficient confirmation method.

[0076] The preset number of term clusters can be set by the user according to the actual application needs. It can be understood that the greater the user's tolerance for the impact of the number of term clusters on the coverage balance threshold, the larger the preset number of term clusters will be. One value is provided, the number of term clusters corresponding to the historical records that meet the user's needs is extracted, outliers in the number of term clusters are removed, and the average value of the number of term clusters after removing outliers is recorded as the preset number of term clusters.

[0077] Users can adaptively set the baseline coverage balance value according to their actual application needs. It can be understood that the higher the user's requirement for the accuracy of information recommendation, the larger the baseline coverage balance value will be. One value is provided, the coverage balance value corresponding to the historical records that meet the user's needs is extracted, outliers in the coverage balance value are removed, and the average value of the coverage balance value after removing outliers is recorded as the baseline coverage balance value.

[0078] Specifically, the cluster update unit responds to preset update conditions to perform update analysis for each term cluster, including:

[0079] For target users whose periodic search relevance is less than the preset periodic search relevance and whose recommended usage rate is less than the preset recommended usage rate, a correlation update analysis is performed.

[0080] For target users whose periodic search fit is greater than or equal to the preset periodic search fit or whose recommended usage rate is greater than or equal to the preset recommended usage rate, it is determined that no optimization or update is needed.

[0081] The preset update condition is that the current time is in an update node.

[0082] The method for confirming the periodic search fit is to obtain the recommended term tags and search term tags between two update nodes, and to record the parts of the recommended term tags and search term tags that are the same as those of the search term tags as overlapping term tags. The term tags corresponding to the recommended information are called recommended term tags, and the term tags corresponding to the search terms are called search term tags. The periodic search fit = number of overlapping term tags / number of recommended term tags; the time corresponding to the end of the most recent monitoring period is the update node.

[0083] The recommended usage rate is confirmed by extracting the number of times the target user clicked on the recommended information within the most recent monitoring period for a single target user; the recommended information is the information pushed to the target user by the agricultural input platform.

[0084] Users can adaptively set the preset periodic search relevance and preset recommendation usage rate values ​​according to their actual application needs. It is understandable that the lower the user's tolerance for the impact of periodic search relevance and recommendation usage rate on the update analysis of each term, the higher the preset periodic search relevance and the higher the preset recommendation usage rate. One set of values ​​is provided, and the periodic search relevance and recommendation usage rate corresponding to the historical records that meet the user's needs are extracted respectively. Outliers in the periodic search relevance and recommendation usage rate are removed, and the average values ​​of the periodic search relevance and recommendation usage rate after removing outliers are recorded as the preset periodic search relevance and preset recommendation usage rate values, respectively.

[0085] Specifically, the cluster update unit performs correlation update analysis, including:

[0086] Based on the descending order of similarity in the search tree diagram, a preset number of associated users are obtained, and the priority recommendation information of each associated user is entered.

[0087] The preset number of associated users can be set by the user according to the actual application needs. It can be understood that the higher the user’s requirements for the accuracy of the association update analysis, the larger the preset number of associated users will be. One value is provided, the number of associated users corresponding to the historical records that meet the user’s needs is extracted, outliers in the number of associated users are removed, and the average value of the number of associated users after removing outliers is recorded as the preset number of associated users.

[0088] Specifically, the cluster update unit determines the search tree diagram similarity based on the anchored similarity reference value and the extended similarity reference value;

[0089] The search tree diagram similarity is positively correlated with both the anchored similarity reference value and the extended similarity reference value.

[0090] The method for confirming the anchored similarity reference value is as follows: for the target user and any associated user, obtain the target search terms of the target user and any associated user in the most recent monitoring period. The anchored similarity reference value = the number of identical target search terms / the total number of target search terms; the total number of target search terms = the number of target search terms of the target user + the number of target search terms of the associated user.

[0091] The method for confirming the extended similarity reference value is as follows: for the target user and any of their corresponding associated users, obtain the term clusters with the same target search terms for both the target user and any of their corresponding associated users. For any term cluster with the same target search terms, the corresponding sub-similarity = number of identical associated search terms / total number of associated search terms. The average of the sub-similarity is recorded as the extended similarity reference value between the target user and any of their associated users. It can be understood that the total number of associated search terms = number of associated search terms of the target user + number of associated search terms of the associated user.

[0092] Search tree diagram similarity = Anchor similarity reference value / Preset anchor similarity reference value + Extended similarity reference value / Preset extended similarity reference value; The values ​​of the preset anchor similarity reference value and the preset extended similarity reference value can be adaptively set by the user according to the actual application needs. It can be understood that the smaller the user's tolerance for the impact of the anchor similarity reference value and the extended similarity reference value on the search tree diagram similarity, the larger the value of the preset anchor similarity reference value and the larger the value of the preset extended similarity reference value will be. One value is provided, and the anchor similarity reference value and the extended similarity reference value corresponding to the historical records that meet the user's needs are extracted respectively. Outliers in the anchor similarity reference value and the extended similarity reference value are removed. The average value of the anchor similarity reference value and the extended similarity reference value after removing outliers is recorded as the value of the preset anchor similarity reference value and the preset extended similarity reference value, respectively.

[0093] Specifically, in response to the first preset recommendation condition, the information push unit obtains the corresponding recommendation information for the cluster of terms with recommendation coefficients greater than the preset recommendation coefficient in descending order, and recommends the recommendation information to the target user.

[0094] The first preset recommendation condition is the completion of the recommendation coefficient setting process. The completion of the recommendation coefficient setting process means that the data analysis unit has determined, based on the comparison result of the coverage balance value and the coverage balance value threshold, whether to determine the recommendation coefficient corresponding to each term cluster based on the coverage ratio or based on the search frequency characterization value, and the recommendation coefficient calculation is completed.

[0095] For any term cluster with a recommendation coefficient greater than a preset recommendation coefficient, the agricultural input products corresponding to this term cluster are the agricultural input products corresponding to each search term within the term cluster. It can be understood that the search terms are the names of agricultural input products, so each search term corresponds to an agricultural input product. In this invention, the agricultural input products corresponding to the term clusters with recommendation coefficients greater than the preset recommendation coefficient are obtained in descending order and recorded as recommended agricultural input products. After the target user logs into the agricultural input platform, the platform randomly recommends a fixed number of recommended agricultural input products, and the agricultural input products recommended each time are different until all agricultural input products have been recommended.

[0096] Specifically, in response to the second preset recommendation conditions, the information push unit supplements the target user's term cluster with the term cluster of the associated user;

[0097] The second preset recommendation condition is that the association update analysis is completed. The association update is completed when the cluster update unit has obtained the preset number of associated users by performing the association update analysis.

[0098] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles 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 all fall within the scope of protection of the present invention.

Claims

1. A cloud platform-based agricultural information management system, characterized in that, The application comprises the following: An image analysis unit is configured to determine a historical search state of a target user according to search word diversity and the number of search words, and determine whether to perform target analysis on the target user according to the historical search state; A data analysis unit connected to the image analysis unit is configured to determine a recommendation coefficient setting process based on coverage ratio to determine the recommendation coefficient corresponding to each term cluster or based on search frequency representation value to determine the recommendation coefficient corresponding to each term cluster according to the comparison result of the coverage balance value of each term cluster and the coverage balance value threshold; A cluster update unit connected to the image analysis unit and the data analysis unit is configured to determine a periodic state according to periodic search fit degree and recommendation utilization rate, and determine whether to perform associated update analysis on the term cluster according to the periodic state; An information pushing unit connected to the image analysis unit, the data analysis unit and the cluster update unit is configured to obtain agricultural products according to the recommendation coefficient, and perform cyclic recommendation on the obtained agricultural products and record the term cluster of the associated user as the term cluster of the target user; A cloud storage unit connected to the image analysis unit, the data analysis unit, the cluster update unit and the information pushing unit is configured to store the search words and the term cluster corresponding to each target user; Wherein, for a single word cluster, the number of corresponding words is recorded as the coverage of the word cluster, and the coverage balance value ; wherein, is the number of word clusters of the target user, is the coverage of the th word cluster, is the average value of the coverage corresponding to the word cluster. 2.The cloud platform-based agricultural information management system according to claim 1, characterized in that, The image analysis unit performs target analysis on the target user with the historical search state of the search word diversity greater than the search word diversity threshold and the number of search words greater than the preset number of search words in response to the preset analysis condition. The preset analysis condition is that the current time is at a cyclic analysis node. 3.The cloud platform-based agricultural information management system according to claim 2, characterized in that, The target analysis of the image analysis unit on the target user comprises: Determining associated search words for each search word in descending order of hot spot usage degree, and recording the search word and the corresponding associated search word as a term cluster; The associated search word is determined according to the term label similarity and the term search relevance corresponding to the search word. 4.The cloud platform-based agricultural information management system according to claim 3, characterized in that, The data analysis unit detects the coverage of each term cluster and obtains the coverage balance value in response to the preset screening condition; If the coverage balance value is greater than the coverage balance value threshold, the recommendation coefficient corresponding to each term cluster is determined based on the coverage ratio; If the coverage balance value is less than or equal to the coverage balance value threshold, the recommendation coefficient corresponding to each term cluster is determined based on the search frequency representation value; The preset screening condition is that the target analysis on the target user is completed. 5.The cloud platform-based agricultural information management system according to claim 4, characterized in that, The data analysis unit determines the coverage balance value threshold based on the number of term clusters, and the coverage balance value threshold and the number of term clusters are in a positive correlation. 6.The cloud platform-based agricultural information management system according to claim 4, characterized in that, The cluster update unit performs update analysis on each term cluster in response to the preset update condition, comprising: For the target user with the periodic state of the periodic search fit degree less than the preset periodic search fit degree and the recommendation utilization rate less than the preset recommendation utilization rate, it is determined to perform associated update analysis; For the target user with the periodic state of the periodic search fit degree greater than or equal to the preset periodic search fit degree or the recommendation utilization rate greater than or equal to the preset recommendation utilization rate, it is determined not to perform optimization update; The preset update condition is that the current time is at an update node. 7.The cloud platform-based agricultural information management system according to claim 6, characterized in that, The cluster update unit performs associated update analysis, comprising: The associated users of the preset associated user quantity are screened according to the descending order of the search tree diagram similarity, and the priority recommendation information of each associated user is recorded. 8.The cloud platform-based agricultural information management system according to claim 7, characterized in that, The cluster updating unit determines the search tree diagram similarity according to the anchor similar reference value and the extended similar reference value; The search tree diagram similarity and the anchor similar reference value and the extended similar reference value are in positive correlation. 9.The cloud platform-based agricultural information management system according to claim 4, characterized in that, The information pushing unit acquires the agricultural products corresponding to the term cluster with the recommendation coefficient greater than the preset recommendation coefficient in the order from large to small according to the first preset recommendation condition, and performs cyclic recommendation on the acquired agricultural products. The first preset recommendation condition is the judgment completion of the recommendation coefficient setting process. 10.The cloud platform-based agricultural information management system according to claim 7, characterized in that, The information pushing unit records the term cluster of the associated user as the term cluster of the target user at the same time according to the second preset recommendation condition. The second preset recommendation condition is the completion of the associated updating analysis.

Images