Service information pushing method and system, electronic equipment and medium

Abstract

The embodiment of the invention relates to the technical field of agricultural informatization and big data application, and provides a service information pushing method and system, electronic equipment and a medium, and the method comprises the steps: obtaining agricultural data; constructing a farmer portrait according to the agricultural data; matching to-be-pushed target agricultural service information according to the farmer portrait; and pushing the target agricultural service information to a corresponding farmer user side according to a preset information pushing rule. Therefore, existing massive objective data mining and analysis are fused, different types of peasant portraits are constructed, peasant demands are accurately grasped, personalized information and services are pushed to peasants in time, the information acquisition efficiency of the peasants and the timeliness and satisfaction degree of obtaining the services are improved, and a support is provided for improving the agricultural production efficiency and the peasant life quality.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural informatization and big data applications, and particularly relates to a service information push method, system, electronic device and medium. Background Art

[0002] With the rapid development of modern agriculture, farmers have diverse needs, and it is very difficult to find the required services from a vast amount of information. Currently, most information push systems are vertical services, and the information and services within the systems are very limited. However, the amount of information farmers need is much larger than the existing information within the systems. Moreover, existing agricultural services, training courses, etc. are mostly organized regularly, and functional departments arrange them centrally according to work practices, hot technologies, etc., and cannot meet the personalized needs of farmers in real time.

[0003] The existing method of analyzing farmers' interest preferences based on data such as the reach rate, complaint rate, message click-through rate, graphic and text click-through rate, browsing, liking, and collection of information pushed to farmers in the system, and corresponding pushing of information does not consider the needs of different types of farmers for service information from the perspective of actual business when analyzing farmers' behaviors and preferences, resulting in inaccurate information pushing. Therefore, how to push personalized information and services to farmers in a timely manner has become an urgent problem to be solved. Summary of the Invention

[0004] The present invention provides a service information push method, system, electronic device and medium, which are used to solve the defect that in the prior art, the needs of different types of farmers for service information are not considered from the perspective of actual business, resulting in inaccurate information pushing, and to realize timely and accurate pushing of personalized service information, and improve the acquisition efficiency of service information by farmers, the timeliness and satisfaction of obtaining services.

[0005] The present invention provides a service information push method, including: Obtain agricultural data; Construct a farmer portrait according to the agricultural data; Match the target agricultural service information to be pushed according to the farmer portrait; Push the target agricultural service information to the corresponding farmer client according to the preset information push rules.

[0006] In a possible implementation manner, the method further includes: Collect feedback data of farmers on the target service information through the farmer client; Update the farmer portrait and optimize the preset information push rules based on the feedback data.

[0007] In a possible implementation manner, the method further includes: Extract farmer feature information from the agricultural data, and construct a farmer portrait based on the farmer feature information, where the farmer portrait includes a basic information portrait, a production feature portrait, an economic feature portrait, a risk feature portrait, and a behavior feature portrait.

[0008] In a possible implementation, the method further includes: Extract farmer basic information, production feature information, economic feature information, risk feature information, and behavior feature information from the agricultural data; Construct a basic information portrait based on the farmer basic information, construct a production feature portrait based on the production feature information, construct an economic feature portrait based on the economic feature information, construct a risk feature portrait based on the risk feature information, and construct a behavior feature portrait based on the behavior feature information; Aggregate the basic information portrait, production feature portrait, economic feature portrait, risk feature portrait, and behavior feature portrait to obtain a classified set of farmer portraits.

[0009] In a possible implementation, the method further includes: Real-time collect different types of agricultural service information and multiple pieces of agricultural service information corresponding to each type; Generate corresponding information tags according to the title and content of each piece of agricultural service information; Match based on the information tags and the classified set of farmer portraits to obtain the target agricultural service information to be pushed corresponding to each classified set of farmer portraits.

[0010] In a possible implementation, the method further includes: The preset information push rules include a manual push mode, an automatic push mode, and push restriction conditions; If the target agricultural service information requires collecting farmer feedback information, determine to perform a manual push; Push the target agricultural service information to the corresponding farmer client according to the manual push setting requirements; If the target agricultural service information does not require collecting farmer feedback information, determine to perform an automatic push; Push the target agricultural service information to the corresponding farmer client according to the automatic push setting requirements, where the automatic push setting requirements include information push quantity requirements, time requirements, frequency requirements, and priority push requirements.

[0011] The present invention also provides a service information push system, including the following modules: A multi-source data collection module for obtaining agricultural data; A farmer portrait construction module for constructing a farmer portrait according to the agricultural data; A service information and portrait matching module, configured to match target agricultural service information to be pushed according to the farmer portrait; A service information pushing module, configured to push the target agricultural service information to the corresponding farmer client according to a preset information pushing rule.

[0012] In a possible implementation manner, the service information pushing system further includes: An information pushing effect evaluation module, configured to collect feedback data of farmers on the target service information through the farmer client; update the farmer portrait and optimize the preset information pushing rule based on the feedback data.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the service information pushing method described in any one of the above is implemented.

[0014] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the service information pushing method described in any one of the above is implemented.

[0015] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, the service information pushing method described in any one of the above is implemented.

[0016] The service information pushing method, system, electronic device, and medium provided by the present invention obtain agricultural data; construct a farmer portrait according to the agricultural data; match target agricultural service information to be pushed according to the farmer portrait; and push the target agricultural service information to the corresponding farmer client according to a preset information pushing rule. Compared with the defect in the prior art that the needs of different types of farmers for service information are not considered from the perspective of actual business, resulting in inaccurate information pushing, in this solution, a large amount of existing objective data is integrated for mining and analysis, different types of farmer portraits are constructed, the needs of farmers are accurately grasped, and then personalized information and services are timely pushed to farmers, improving the information acquisition efficiency of farmers, the timeliness and satisfaction of obtaining services, and providing support for improving agricultural production efficiency and the quality of farmers' lives. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 One of the schematic flowcharts of the service information push method provided by the present invention.

[0019] Figure 2 Another schematic flowchart of the service information push method provided by the present invention.

[0020] Figure 3 The third schematic flowchart of the service information push method provided by the present invention.

[0021] Figure 4 Schematic diagram of the matching between farmer portraits and service information provided by the present invention.

[0022] Figure 5 Schematic diagram of the structure of the service information push system provided by the present invention.

[0023] Figure 6 Schematic diagram of the structure of the electronic device provided by the present invention. Detailed implementation manners

[0024] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0025] For the convenience of understanding the embodiments of the present invention, further explanatory descriptions will be given below with specific embodiments in conjunction with the accompanying drawings. The embodiments do not constitute a limitation to the embodiments of the present invention.

[0026] Figure 1 One of the schematic flowcharts of the service information push method provided by the present invention. As Figure 1 shown, the method includes the following: S11. Obtain agricultural data.

[0027] In the embodiments of the present invention, first, data is collected from multiple agricultural data sources. The data sources involved include: agricultural production systems, agricultural product trading platforms, land resource management systems, meteorological data platforms, financial institution credit systems, agricultural policy release platforms, agricultural supplies sales platforms, production and sales docking platforms, skill training platforms, etc.

[0028] Among them, the agricultural production system: used to collect data on farmers' planting, breeding, irrigation, fertilization, pest control, etc., such as the varieties of crops planted, the planting area, the number of livestock raised, the agricultural machinery and agricultural supplies used, etc.

[0029] Agricultural product trading platform: used to obtain sales data of agricultural products, including sales channels, prices, trading volumes, sales times, etc.

[0030] Land resource management system: used to collect information such as farmers' land area, land fertility, land type, etc.

[0031] Meteorological data platform: used to provide meteorological data in the area where farmers are located, such as temperature, precipitation, sunlight, historical information and early warning information of disastrous weather, etc.

[0032] Credit system of financial institutions: used to record information such as farmers' credit records, repayment situations, asset mortgages, etc., as well as credit and insurance product promotion activities and preferential information.

[0033] Agricultural policy release platform: used to obtain national and local agricultural policy information.

[0034] Agricultural materials sales platform: used to obtain information such as brands and prices of agricultural materials products such as seeds, pesticides, fertilizers, agricultural machinery, etc. sold online, supplier information, customer consumption records, and agricultural materials preferential activities.

[0035] Production and sales docking platform: used to obtain information on the supply of agricultural and livestock products, sales information, publicity information on production and sales activities, docking records, etc.

[0036] Skills training platform: used to obtain agricultural technology training courses, notices of farmers' training activities, historical training records, satisfaction survey information, etc.

[0037] The data collection methods include: API interface calls, web crawlers, sensor data transmission, etc. According to the characteristics of different data sources and the data update frequency, data is collected regularly or in real time and stored in the data storage unit.

[0038] S12. Construct a farmer portrait based on the agricultural data.

[0039] Extract farmers' characteristic information from the agricultural data and construct a farmer portrait based on the farmers' characteristic information. Among them, the farmer portrait includes a basic information portrait, a production characteristic portrait, an economic characteristic portrait, a risk characteristic portrait, and a behavior characteristic portrait.

[0040] Specifically, extract farmers' basic information, production characteristic information, economic characteristic information, risk characteristic information, and behavior characteristic information from the agricultural data; construct a basic information portrait based on the farmers' basic information, a production characteristic portrait based on the production characteristic information, an economic characteristic portrait based on the economic characteristic information, a risk characteristic portrait based on the risk characteristic information, and a behavior characteristic portrait based on the behavior characteristic information; aggregate the basic information portrait, the production characteristic portrait, the economic characteristic portrait, the risk characteristic portrait, and the behavior characteristic portrait to obtain a farmer portrait classification set.

[0041] S13. Match the target agricultural service information to be pushed according to the farmer portrait.

[0042] Collect different types of agricultural service information in real time and multiple pieces of agricultural service information corresponding to each type; generate corresponding information tags according to the title and content of each piece of agricultural service information; match based on the information tags and the farmer portrait classification set to obtain the target agricultural service information to be pushed corresponding to each farmer portrait classification set.

[0043] S14. Push the target agricultural service information to the corresponding farmer client according to the preset information push rules.

[0044] If the target agricultural service information requires collecting farmer feedback information, determine to perform manual push; push the target agricultural service information to the corresponding farmer client according to the manual push setting requirements.

[0045] If the target agricultural service information does not require collecting farmer feedback information, determine to perform automatic push; push the target agricultural service information to the corresponding farmer client according to the automatic push setting requirements, where the automatic push setting requirements include information push quantity requirements, time requirements, frequency requirements, and priority push requirements.

[0046] Furthermore, it is also possible to collect the feedback data of farmers on the target service information through the farmer client; update the farmer portrait and optimize the preset information push rules based on the feedback data.

[0047] Specifically, collect the feedback of farmers on the service information, including whether they have signed up for the pushed service, whether they have read the pushed information, the time and duration of reading the service information, the records of participating in the pushed service activities, etc., record the service usage effect, satisfaction evaluation, etc. into the system, further update the farmer portrait and the group label set, and realize the continuous improvement of service push. Analyze based on the collected feedback data, master the information push arrival rate, click-through rate, conversion rate, service satisfaction analysis, etc., provide optimization suggestions for the push strategy of farmer services, and improve the push effect and service satisfaction.

[0048] The service information push method provided by the present invention obtains agricultural data, constructs a farmer profile based on the agricultural data, matches target agricultural service information to be pushed according to the farmer profile, and pushes the target agricultural service information to the corresponding farmer user terminal according to a preset information push rule. Compared with the defect in the prior art that the needs of different types of farmers for service information are not considered from the perspective of actual business, resulting in inaccurate information push, this method integrates the mining and analysis of existing massive objective data, constructs farmer profiles of different types, accurately grasps the needs of farmers, and then timely pushes personalized information and services to farmers, improving the efficiency of farmers' access to information, the timeliness and satisfaction of obtaining services, and providing support for improving agricultural production efficiency and farmers' living quality.

[0049] Figure 2 is the second flow diagram of the service information push method provided by the present invention, as Figure 2 shown, this method includes the following: S21. Obtain agricultural data.

[0050] In the embodiment of the present invention, it will be described in detail in combination with Figure 3 First, data is collected from multiple agricultural data sources, and the data sources involved include: agricultural production systems, agricultural product trading platforms, land resource management systems, meteorological data platforms, credit systems of financial institutions, agricultural policy release platforms, agricultural supplies sales platforms, production and sales docking platforms, skill training platforms, etc.

[0051] Among them, the agricultural production system is used to collect data on farmers' planting, breeding, irrigation, fertilization, pest control, etc., such as the varieties of crops planted, the planting area, the number of livestock raised, the agricultural machinery and agricultural supplies varieties used, etc.

[0052] The agricultural product trading platform is used to obtain the sales data of agricultural products, including sales channels, prices, trading volumes, sales times, etc.

[0053] The land resource management system is used to collect information such as the land area, land fertility, and land type of farmers.

[0054] The meteorological data platform is used to provide meteorological data in the area where farmers are located, such as temperature, precipitation, sunlight, historical information and warning information of disastrous weather, etc.

[0055] The credit system of financial institutions is used to record information such as farmers' credit records, repayment situations, asset mortgages, etc., as well as credit and insurance product promotion activities and preferential information.

[0056] The agricultural policy release platform is used to obtain national and local agricultural policy information.

[0057] Agricultural supplies sales platform: used to obtain information such as brands, prices, supplier information, customer consumption records, and agricultural supplies preferential activities of agricultural supplies products such as seeds, pesticides, fertilizers, and agricultural machinery sold online.

[0058] Production and sales docking platform: used to obtain information on agricultural and livestock product supply, sales information, publicity information on production and sales activities, docking records, etc.

[0059] Skills training platform: used to obtain agricultural technology training courses, notices of farmers' training activities, historical training records, satisfaction survey information, etc.

[0060] The data collection methods include: API interface calls, web crawlers, sensor data transmission, etc. According to the characteristics of different data sources and the data update frequency, data is collected regularly or in real-time and stored in the data storage unit.

[0061] Furthermore, the collected agricultural-related data can be efficiently stored and managed. Relational databases (such as MySQL) or non-relational databases are used to store structured and unstructured data. Classified storage is carried out according to data sources, types, and uses for easy management and query.

[0062] The collected data is cleaned, associated, standardized, etc. Data cleaning means removing noise data, duplicate data, and error data, such as clearing abnormal sensor readings, incorrect transaction records, etc. The data is standardized, converting data with different units and formats into a unified standard. For example, different land area units are unified, and different agricultural product prices are unified into the same currency unit. Missing data is filled. According to the distribution characteristics of the data and business logic, methods such as mean filling, median filling, or prediction filling based on machine learning algorithms are used to ensure the integrity of the data.

[0063] After data preprocessing, farmers' information from different data sources is associated and aggregated through the farmer's ID number as the unique identifier to form the characteristic source data of the farmer household and enter the farmer characteristic data resource library. After data preprocessing, the multi-source agricultural service information collected in real-time is put into the agricultural service information data resource library in real-time.

[0064] S22. Extract farmers' basic information, production characteristic information, economic characteristic information, risk characteristic information, and behavior characteristic information from the agricultural data.

[0065] Extract information related to farmers' characteristics from the farmer characteristic data resource library to construct a multi-dimensional farmer portrait, including but not limited to basic information portrait, production characteristic portrait, economic characteristic portrait, risk characteristic portrait, behavior characteristic portrait, etc.

[0066] S23. Construct a basic information portrait based on the basic information of the farmers, construct a production feature portrait based on the production feature information, construct an economic feature portrait based on the economic feature information, construct a risk feature portrait based on the risk feature information, and construct a behavior feature portrait based on the behavior feature information.

[0067] Basic information portrait: Gender, age, education level, occupation type (ordinary farmer, large-scale grower, family farm operator, leader of farmer cooperatives, professional agricultural technician, etc.), other identities, types of crops / breeds raised, whether participating in medical insurance, whether enjoying subsistence allowances, etc.

[0068] Production feature portrait: Such as planting and breeding structure (types and proportions of planted crops and raised animals), production scale (land area, number of livestock), production technology level (situation of new technologies used, advanced agricultural machinery and tools), etc.

[0069] Among them, for the planting and breeding structure score: Collect data on the types and quantities of crops planted and animals raised by farmers, calculate the proportions of various crops and animals in the total production, and then calculate the planting and breeding structure score according to the preset proportional weights.

[0070] Production scale score: Statistically analyze data such as the land area and number of livestock of farmers, and form a production scale score according to the preset scale standards, such as a land area greater than X mu is large-scale, less than Y mu is small-scale, etc.

[0071] Production technology level score: Record information such as the types of new technologies used by farmers, the quantity and usage frequency of advanced agricultural machinery and tools, and score the production technology level of farmers according to the technical level index system, such as the number of new technology applications exceeding M items, the proportion of advanced agricultural machinery and tools exceeding N% is high-tech level, etc.

[0072] Set weights for the scores of each dimension of the production feature portrait, and this weight can be adjusted in real time as needed. Finally, sum the weighted scores of each dimension to obtain the production feature portrait score of the farmer household.

[0073] According to the production feature portrait score, farmer households can be classified into large-scale, intensive, and extensive types.

[0074] Economic feature portrait: Such as income level (calculated based on agricultural product sales and subsidy income), debt level (based on credit records), asset status (land, agricultural machinery and tools, livestock in stock, etc.).

[0075] Among them, for the income level score: Integrate the agricultural product sales data and subsidy income data of farmers, calculate their annual total income, set income level thresholds, such as an income higher than P yuan is high income, lower than Q yuan is low income, etc., to determine the income level score of farmers.

[0076] Debt level score: Based on the farmer’s credit record, the total debt, debt structure and other information are counted, combined with the farmer’s debt repayment ability indicators, such as a debt-to-income ratio exceeding R% indicating high debt risk, to assess the farmer’s debt level score.

[0077] Asset status scoring: Check the quantity and value of farmers' land, farm machinery, livestock and other assets, create an asset list, and score the farmers' asset status based on the total asset value and asset allocation. For example, if the total asset value exceeds S million yuan, it is considered high assets.

[0078] The weights are set for the scores of each dimension of the economic characteristic portrait, and the weights can be adjusted in real time as needed. Finally, the weighted sum of the scores of each dimension is used to obtain the economic characteristic portrait score of the farmer.

[0079] Based on the economic characteristic portrait scores, farmers can be divided into profitable, loss-making and balanced types.

[0080] Risk profile: Assess the natural risks faced by farmers based on meteorological data and pest and disease data, assess market risks based on market price fluctuations, and obtain farmers' credit risks from external credit scoring systems.

[0081] Among them, the natural risk score: obtain meteorological data and pest and disease data in the farmers' area, analyze the natural disaster risks and pest and disease risk levels they face, and build a natural risk assessment model. For example, if the number of consecutive heavy rain days exceeds T days and the frequency of pest and disease occurrence exceeds U times / year, it is a high natural risk, and determine the farmers' current natural risk score.

[0082] Market risk score: Collect data on price fluctuations in the agricultural product market, combine it with farmers' production and sales situation, and use market risk assessment methods, such as price volatility exceeding V% for high market risk, to assess the market risk score faced by farmers.

[0083] Credit risk score: Access to an external credit scoring system to obtain a calculated credit score. This score refers to the farmer’s credit record, repayment ability and other information, and is obtained using a credit risk assessment model. After standardized processing by this module, the farmer’s credit risk score is obtained.

[0084] Based on the scores of various risk characteristic portraits, farmers can be divided into high-risk, medium-risk and low-risk types.

[0085] Behavioral characteristic portrait: By analyzing farmers' trading behavior and information acquisition behavior, we can analyze their decision-making habits and preferences, such as whether they prefer online or offline transactions, and what kind of agricultural information they pay more attention to.

[0086] Among them, the online transaction behavior score: Record data such as the farmer's transaction method, transaction frequency, transaction amount, etc., and judge whether they tend to online or offline transactions. For example, if the proportion of online transaction amount exceeds X%, it is an online transaction type, etc., and determine the score of the online transaction behavior.

[0087] Online information acquisition behavior score: Statistically analyze information such as the channels through which farmers obtain agricultural information, the types and frequencies of information they pay attention to, etc. According to the information acquisition behavior indicators, such as obtaining agricultural technology information through the online platform at least Y times a week is an online learning type, etc., classify the information acquisition behavior types of farmers, and determine the score of the online information acquisition behavior.

[0088] Set weights for the scores of each dimension of the behavioral characteristic portrait, and this weight can be adjusted in real time according to needs. Finally, sum the weighted scores of each dimension to obtain the score of the farmer's behavioral characteristic portrait.

[0089] According to the score of the behavioral characteristic portrait, farmers can be divided into online type, offline type, and online-offline integrated type.

[0090] S24. Aggregate the basic information portrait, production characteristic portrait, economic characteristic portrait, risk characteristic portrait, and behavioral characteristic portrait to obtain a farmer portrait classification set.

[0091] Conduct clustering analysis on all farmer information, and define the generated clusters as new farmer groups. Clustering methods include but are not limited to K-means, hierarchical clustering, DBSCAN, spectral clustering, BIRCH, etc. Compare each new farmer group with the existing portrait group. If the overlap degree is greater than 70%, then the portrait group is redefined as the union of the new farmer group and the portrait group, and finally aggregate the portraits to generate a new portrait classification set.

[0092] The aggregation steps of the farmer portrait include: label vectorization representation, similarity matrix construction, label matching and merging.

[0093] First, label vectorization representation: Perform label vectorization representation on the existing portrait group label set V1 and the new farmer group label set V2. Label representation methods include but are not limited to pre-trained language models such as word embedding, BERT, Word2Vec, etc. ti ∈ V1 and tj ∈ V2 are the vectorizations of two labels. Convert each label ti into a high-dimensional vector representation vi = Embed(ti), ∀ti ∈ V1 ∪ V2 Construct a similarity matrix: Use cosine similarity to construct a similarity matrix to judge the similarity between labels Sij = Sij ∈ [0, 1] is the similarity of ti and tj.

[0094] Then, construct a dual-graph structure: Generate a graph structure G1 and G2 respectively from the label set V1 of the existing portrait group and the label set V2 of the new farmer group. The nodes are labels, and the edge weights are the similarities between the nodes. The higher the weight, the more similar the labels on the nodes. Let the nodes of V1 and V2 be {v1, v2, …, vn1} and {vn1+1, …, vn1+n2} respectively, and the edge weights are the similarities between the labels. The edge weight of the constructed graph is Sij, that is: Weight(vi, vj)=Sij, ∀i∈[1, n1], j∈[n1+1, n1+n2]. This not only makes the labels in each set associated with each other but also allows the similarities between cross-set labels to be integrated.

[0095] Node update of the graph structure: The representation of the nodes in the graph will be updated through the representations of their neighbors. The vector representation of the label ti at the (k + 1)-th layer of the neural network is: where, is the vector representation of the label ti at the k-th layer, is the neighbor set of the label ti, is the weight between the node i and the neighbor node j, and W(k) is the weight matrix at the k-th layer, is the Sigmoid activation function. is the learning rate, is the gradient of the loss function with respect to the node representation of.

[0096] Finally, label matching and merging: During the training process of the neural network, use the optimization objective function to drive the merging of labels, ensuring that labels with relatively high similarities are clustered into one category and avoiding excessive merging between labels of the same category. The loss function representing label difference is: where, is a regular expression, is the final clustering center of the label ti. The representations of the labels will be gradually merged as the graph structure is updated. Finally, V1 and V2 are merged to form a new farmer portrait label set V3, retaining the independence and diversity of all labels in the original farmer group set.

[0097] is the label cluster obtained by clustering for each label .

[0098] Furthermore, the initial automatic construction of the farmer portrait group label set can be achieved through statistical features based on TF-IDF and TextRank, machine learning methods based on classifiers and feature engineering, deep learning methods based on LDA, RNN, LSTM, KeyBERT, etc., or rules such as regular expressions and conditional random fields to generate portrait group labels.

[0099] Furthermore, the farmer portraits can also be classified and managed, including farmer portrait name management and farmer portrait group label management.

[0100] Among them, for farmer portrait group label management: according to different farmer portrait classifications, keywords are extracted from the corresponding types of farmer-related information as the label set. Different farmer portrait groups correspond to different information label keyword sets. First, these label sets are automatically extracted based on all the information keywords of this type of farmer household in the platform. Then, business experts can edit the label keyword set, add, delete, or modify keywords to form the farmer portrait group label set for this type.

[0101] S25. Real-time collect different types of agricultural service information and multiple pieces of information corresponding to each type.

[0102] The agricultural service information in the platform is aggregated from multiple platforms. The platform regularly collects various agricultural service information in real time from relevant platforms such as the agricultural product trading platform, land resource management system, meteorological data platform, financial institution credit system, agricultural policy release platform, agricultural materials sales platform, production and marketing docking platform, and skills training platform, such as agricultural technology training courses, loan products, agricultural materials preferential activities, agricultural machinery sales information, etc., and stores them in the agricultural service information data resource library after preprocessing.

[0103] S26. Generate corresponding information labels according to the title and content of each piece of agricultural service information.

[0104] Extract agricultural service information data from the agricultural service information data resource library. Each piece of agricultural service information automatically extracts one or more keyword labels according to the title and content, and is arranged in descending order of the information generation date. Each service information includes a generation date label, an activity date label, whether it is an online activity, whether registration is required, a keyword label, a recommended farmer portrait label set, a label indicating whether it has expired, and a label indicating whether it has been pushed.

[0105] The administrator can also manually mark a piece of service information as important.

[0106] If the deadline in the information is greater than the current date, the system automatically marks it as expired; each service information is defaulted to not pushed and is marked as pushed after being pushed. Expired and pushed information is automatically placed in the historical information management module for display and management, and the rest is placed in the to-be-pushed information management module.

[0107] S27. Match based on the information tags and the classified set of farmer portraits to obtain the target agricultural service information to be pushed corresponding to each classified set of farmer portraits.

[0108] As Figure 4 shown, the set of farmer portrait tags is matched with the set of service information tags, and the keyword set of each service information is matched with the tag set of each farmer portrait type to form key-value pairs of (service information id, farmer portrait type id, matching degree). According to the matching degree, sort from high to low, and include the top 20% or the top 6 farmer portrait type tags with the highest matching degree into the set of farmer portrait tags recommended for pushing according to the rules.

[0109] The matching rule between service information and farmers is a service information package set in advance based on farmer portraits. Farmer portraits include basic information portraits, production feature portraits, economic feature portraits, risk feature portraits, behavior feature portraits, etc. The package setting rule of service information is: the service information with a high matching degree with the farmer portrait group is pushed first; the service information with important tags is pushed first; the information with a nearby activity date is pushed first.

[0110] For agricultural technology training courses, loan products, agricultural supplies preferential activities, agricultural machinery sales information, loan product information, etc., extract multiple keywords according to the information source and information content, automatically set keyword tags for the information, and match different farmer portrait groups according to the information keywords.

[0111] Pre-set tag sets for different farmer portrait groups, and these tag sets are obtained by extracting all information keywords of this type of farmer households on the platform. The set of farmer portrait group tags can be generated by statistical features based on TF-IDF and TextRank, or by machine learning methods based on classifiers and feature engineering, or by deep learning methods based on LDA, RNN, LSTM, KeyBERT, etc., or by rules such as regular expressions and conditional random fields.

[0112] Calculating the matching degree between service information and farmer portrait groups mainly considers two aspects. One is the number of overlapping keywords in the keyword set of the service information and the keyword set of the farmer portrait group information tags, and the other is the proportion of the number of identical keywords in the keyword set to the total number of keywords in the set. The larger the value of the second aspect factor, the better.

[0113] Calculation of the matching degree: Define the keyword set of service information and the tag set of the farmer portrait group as S and P respectively, and define the matching degree as: Among them, It represents the proportion of overlapping keywords in the total number of keywords, i.e., the Jaccard similarity, which is used to measure the degree of overlap of sets. It represents the ratio of the number of overlapping keywords to the size of the smaller set, emphasizing the absolute number of overlapping keywords. It is the contribution weight of two factors to the matching degree, and the platform default setting is 0.5.

[0114] Generate the matching degree between each service information and the farmer portrait according to the above rules.

[0115] When automatically pushing information, if at least 80% of the information keywords match the set of group labels of the basic information portrait, and 70% match the set of group labels of any one of the farmer portrait types such as the production feature portrait, economic feature portrait, risk feature portrait, and behavior feature portrait, then the information can be pushed to the corresponding one or more portrait groups.

[0116] S28. Push the target agricultural service information to the corresponding farmer user terminal according to the preset information push rules.

[0117] The platform sets push limits for service information, with corresponding upper limits on the number of information pushes per day / month / year.

[0118] The information push methods are divided into two types: manual push and automatic push.

[0119] Manual push module, manual push method: The administrator manually selects service information, push objects, sets the push time and push channels, and the information will be pushed to the farmer group or individual with the corresponding label according to the set requirements.

[0120] Automatic push module, automatic push method: The system automatically pushes service information to the farmer group in the set of farmer portrait labels recommended for pushing. The administrator can set the automatic push time, push frequency, the number of messages pushed per day, and the rules for preferentially pushing information, such as at most 2 messages can be automatically pushed per day, normal push on holidays, and matching degree takes precedence over the information generation date, etc.

[0121] Push rule setting module, push limit: To improve the farmer experience and sensitivity to push information, set a maximum of 2 messages to be automatically pushed per day and a maximum of 3 service information to be manually pushed.

[0122] According to the farmer portrait, customize personalized service information push packages for different types of farmers. The packages are divided into two categories according to the farmer feedback form: one is that feedback registration information is required after reaching the farmer, and the other is information for reading and learning only. The information for reading and learning includes: policy information, disaster warning information, agricultural materials preferential activity information, loan product information, etc.

[0123] For those who need to provide feedback on registration information after reaching farmers, a manual push method is generally used. After setting rules such as the push cycle and push time for the automatic push method, the system automatically matches information with farmers and performs pushes according to the rules.

[0124] Ultimately, for large-scale farmers growing corn, service information such as high-end agricultural technology training for corn planting, promotion of advanced agricultural machinery, and agricultural product brand building is pushed; for cattle-raising farmers with low credit risk and online presence, information on livestock loans, online training courses, and online beef cattle docking is pushed; for users with high natural risks, weather warnings, pest and disease warnings and response plans, agricultural insurance information, etc. are pushed. Moreover, multiple information push channels can be adopted, such as mobile phone text messages, mobile application notifications, WeChat messages, social media, etc., to push service information to farmers in a timely manner, while recording the push information and farmers' feedback information.

[0125] Furthermore, it is also possible to collect feedback data from farmers on target service information through the farmer user terminal; update the farmer portrait and optimize the preset information push rules based on the feedback data.

[0126] Specifically, collect farmers' feedback on service information, including whether they have signed up for the pushed service, whether they have read the pushed information, the time and duration of reading the service information, records of participating in the pushed service activities, etc., record the service usage effect, satisfaction evaluation, etc. into the system, further update the farmer portrait and the set of group labels, and achieve continuous improvement of service push. Analyze based on the collected feedback data, master content such as information push arrival rate, click-through rate, conversion rate, and service satisfaction analysis, provide optimization suggestions for the push strategy of farmers' services, and improve the push effect and service satisfaction.

[0127] This service information push method can collect and integrate multi-source information related to farmers in real time, break down system barriers, and provide rich data resources for farmer portraits and service information. Two methods, namely farmer group classification and farmer information keyword clustering, are used in different feature fields to generate farmer group portraits, and then the portraits are re-clustered to correspondingly generate different types of farmer portrait group label sets, which can more accurately explore the business needs of different types of farmer groups. According to the matching degree between the service information and the farmer portrait label set, service information is pushed to farmers in both manual and automatic ways, which better meets the service information acquisition needs of various farmer groups and the actual information push management business needs of functional departments.

[0128] Next, the service information push system provided by the present invention is described. The service information push system described below can be mutually referred to the service information push method described above.

[0129] Figure 5It is a schematic structural diagram of the service information push system provided by the present invention, specifically including: a multi-source data collection module, a data storage and management module, a data preprocessing module, a farmer portrait construction module, a service information push module, and an information push effect evaluation module.

[0130] Specifically, the multi-source data collection module is used to collect data from multiple agricultural data sources; the data storage and management module is used to store and manage the collected data; the data preprocessing module is used to clean, standardize, and handle missing values of the collected data; the farmer portrait construction module is used to construct farmer portraits from the preprocessed data; the agricultural service information push module is used to push agricultural service information according to the farmer portraits; the information push effect evaluation module is used to evaluate the push effect of the information.

[0131] Specifically, the data sources involved in the multi-source data collection module include: agricultural production systems, agricultural product trading platforms, land resource management systems, meteorological data platforms, financial institution credit systems, agricultural policy release platforms, agricultural input sales platforms, production and sales docking platforms, skill training platforms, etc.

[0132] Among them, the agricultural production system: is used to collect data on farmers' planting, breeding, irrigation, fertilization, pest control, etc., such as the varieties of crops planted, the planting area, the number of livestock raised, the agricultural machinery and agricultural input varieties used, etc.

[0133] The agricultural product trading platform: is used to obtain the sales data of agricultural products, including sales channels, prices, trading volumes, sales times, etc.

[0134] The land resource management system: is used to collect information on farmers' land area, soil fertility, land type, etc.

[0135] The meteorological data platform: is used to provide meteorological data in the area where farmers are located, such as temperature, precipitation, sunlight, historical information and early warning information of disastrous weather, etc.

[0136] The credit system of financial institutions: is used to record information such as farmers' credit records, repayment situations, asset mortgages, credit, insurance product promotion activities and preferential information.

[0137] The agricultural policy release platform: is used to obtain national and local agricultural policy information, such as subsidy policies, industrial support policies, etc.

[0138] The agricultural input sales platform: is used to obtain information on the brands and prices of agricultural inputs such as seeds, pesticides, fertilizers, and agricultural machinery sold online, supplier information, customer consumption records, agricultural input preferential activities, etc.

[0139] The production and sales docking platform: is used to obtain information on the supply of agricultural and livestock products, sales information, publicity information on production and sales activities, docking records, etc.

[0140] Skills Training Platform: Used to obtain agricultural technology training courses, notices of farmer training activities, historical training records, satisfaction survey information, etc.

[0141] Data collection methods include: API interface calls, web crawlers, sensor data transmission, etc. According to the characteristics of different data sources and the data update frequency, data is collected regularly or in real time and stored in the data storage unit.

[0142] The data storage and management module realizes the efficient storage and management of the collected agricultural-related data, including sub-modules such as database management, metadata management, data backup and recovery, data security control, data quality control, data sharing and integration, etc. This module uses a relational database (such as MySQL) or a non-relational database to store structured and unstructured data. It is classified and stored according to the data source, type, and usage for easy management and query.

[0143] The data preprocessing module is used to clean, associate, standardize, etc. the collected data. Data cleaning means removing noise data, duplicate data, and error data, such as clearing abnormal sensor readings, incorrect transaction records, etc. Standardize the data, convert data with different units and formats into a unified standard, for example, unify different land area units and unify different agricultural product prices into the same currency unit. Fill in the missing data, and according to the data distribution characteristics and business logic, use methods such as mean filling, median filling, or prediction filling based on machine learning algorithms to ensure the integrity of the data.

[0144] After data preprocessing, farmer information from different data sources is associated and aggregated through the farmer's ID number as the unique identifier to form the characteristic source data of the farmer household and enter the farmer characteristic data resource library. After data preprocessing, the multi-source agricultural service information collected in real time is put into the agricultural service information data resource library in real time.

[0145] Farmer Portrait Construction Module: Includes farmer portrait classification generation module, clustering portrait generation module, farmer portrait aggregation module, and farmer portrait classification management module.

[0146] Among them, the farmer portrait classification generation module: Extracts information related to farmer characteristics from the farmer characteristic data resource library to construct a multi-dimensional farmer portrait, including but not limited to basic information portrait, production characteristic portrait, economic characteristic portrait, risk characteristic portrait, behavior characteristic portrait, etc.

[0147] Basic Information Portrait: Gender, age, education level, occupation type (ordinary farmer, large-scale grower, family farm operator, leader of farmer cooperatives, professional agricultural technician, etc.), other identities, planting / breeding varieties, whether to participate in medical insurance, whether to enjoy subsistence allowances, etc.

[0148] Production characteristic portrait: such as planting and breeding structure (types and proportions of planted crops and raised animals), production scale (land area, breeding quantity), production technology level (new technologies used, situation of advanced agricultural machinery), etc.

[0149] Among them, the planting and breeding structure score: collect data on the types and quantities of crops planted and animals raised by farmers, calculate the proportions of various crops and animals in the total production, and then calculate the planting and breeding structure score according to the preset proportional weights.

[0150] Production scale score: count data such as the land area and breeding quantity of farmers, and form a production scale score according to the preset scale standards, such as a large scale when the land area is greater than X mu and a small scale when it is less than Y mu, etc.

[0151] Production technology level score: record information such as the types of new technologies used by farmers, the quantity and usage frequency of advanced agricultural machinery, and score the production technology level of farmers according to the technical level index system, such as a high-tech level when the number of new technology applications exceeds M items and the proportion of advanced agricultural machinery exceeds N%.

[0152] Set weights for the scores of each dimension of the production characteristic portrait, and this weight can be adjusted in real time as needed. Finally, sum the weighted scores of each dimension to obtain the production characteristic portrait score of the farmer household.

[0153] According to the production characteristic portrait score, farmer households can be classified into large-scale, intensive, and extensive types.

[0154] Economic characteristic portrait: such as income level (calculated based on agricultural product sales and subsidy income), debt level (based on credit records), asset status (land, agricultural machinery, livestock on hand, etc.).

[0155] Among them, the income level score: integrate the agricultural product sales data and subsidy income data of farmers, calculate their annual total income, set income level thresholds, such as a high income when the income is higher than P yuan and a low income when it is lower than Q yuan, etc., to determine the income level score of farmers.

[0156] Debt level score: Based on the credit records of farmers, count information such as their total debt and debt structure, and evaluate the debt level score of farmers in combination with the debt repayment ability indicators, such as a high debt risk when the debt-income ratio exceeds R%.

[0157] Asset status score: Conduct an inventory of the quantity and value of assets such as the land, agricultural machinery, and livestock on hand of farmers, establish an asset list, and score the asset status of farmers according to the total asset value and asset allocation situation, such as a high asset when the total asset value exceeds S million yuan.

[0158] Weigh the scores of each dimension of the economic characteristics portrait, and this weight can be adjusted in real time according to needs. Finally, sum the weighted scores of each dimension to obtain the score of the farmer's economic characteristics portrait.

[0159] Based on the score of the economic characteristics portrait, farmers can be classified into profit-making type, loss-making type, and balanced type.

[0160] Risk characteristics portrait: Evaluate the natural risks faced by farmers based on meteorological data and pest and disease data, and evaluate market risks based on market price fluctuations. Obtain the credit risks of farmers from an external credit scoring system.

[0161] Among them, the natural risk score: Obtain the meteorological data and pest and disease data of the farmer's location, analyze the degree of natural disaster risks and pest and disease risks they face, construct a natural risk assessment model, such as continuous heavy rain days exceeding T days, pest and disease occurrence frequency exceeding U times / year as high natural risks, etc., and determine the current natural risk score of the farmer.

[0162] Market risk score: Collect data on the fluctuations of agricultural product market prices, combine with the production and sales situation of farmers, and use market risk assessment methods, such as price volatility exceeding V% as high market risk, etc., to evaluate the market risk score faced by farmers.

[0163] Credit risk score: Connect to an external credit scoring system to obtain the calculated credit score. This score refers to information such as the farmer's credit record and repayment ability, and is obtained using a credit risk assessment model. After being standardized by this module, the farmer's credit risk score is obtained.

[0164] Based on the scores of various risk characteristics portraits, farmers can be classified into high-risk type, medium-risk type, and low-risk type.

[0165] Behavior characteristics portrait: Analyze the farmer's trading behavior and information acquisition behavior to analyze their decision-making habits and preferences, such as whether they tend to trade online or offline, and what kind of agricultural information they are more concerned about.

[0166] Among them, the online trading behavior score: Record data such as the farmer's trading method, trading frequency, and trading amount, and judge whether they tend to trade online or offline. For example, if the proportion of online trading amount exceeds X% as the online trading type, etc., and determine the score of the online trading behavior.

[0167] Online information acquisition behavior score: Statistically analyze information such as the channels through which farmers obtain agricultural information, the types and frequencies of information they are concerned about, and based on information acquisition behavior indicators, such as obtaining agricultural technology information through an online platform at least Y times a week as the online learning type, etc., classify the information acquisition behavior types of farmers and determine the score of the online information acquisition behavior.

[0168] Weights are set for the scores of each dimension of the wind behavior characteristic portrait, and these weights can be adjusted in real time according to needs. Finally, the scores of each dimension are weighted and summed to obtain the score of the farmer's behavior characteristic portrait.

[0169] According to the score of the behavior characteristic portrait, farmers can be classified into online type, offline type, and online-offline integrated type.

[0170] Cluster portrait generation module: Cluster analysis is used to generate portraits. All farmer information is clustered, and the generated clusters are defined as new farmer groups. Clustering methods include but are not limited to K-means, hierarchical clustering, DBSCAN, spectral clustering, BIRCH, etc.

[0171] Farmer portrait aggregation module: Each new farmer group obtained by the cluster portrait generation module is compared with the existing portrait group. If the overlap degree is greater than 70%, the portrait group is redefined as the union of the new farmer group and the portrait group. Finally, the portraits are aggregated to generate a new portrait classification set.

[0172] The aggregation steps of the farmer portrait include: label vectorization representation, similarity matrix construction, label matching and merging.

[0173] First, label vectorization representation: The label sets V1 of the existing portrait group and V2 of the new farmer group are vectorized. Label representation methods include but are not limited to pre-trained language models such as word embedding, BERT, Word2Vec, etc. ti ∈ V1 and tj ∈ V2 are the vectorizations of two labels. Each label ti is converted into a high-dimensional vector representation vi = Embed(ti), ∀ti ∈ V1 ∪ V2 Construct a similarity matrix: Use cosine similarity to construct a similarity matrix to judge the similarity between labels Sij = Sij ∈ [0,1] is the similarity between ti and tj.

[0174] Then, construct a bipartite graph structure: The label set V1 of the existing portrait group and the label set V2 of the new farmer group generate a graph structure G1 and G2 respectively, where the nodes are labels and the edge weights are the similarities between the nodes. The higher the weight, the more similar the labels on the nodes. Let the nodes of V1 and V2 be {v1, v2, …, vn1} and {vn1+1, …, vn1+n2} respectively, and the edge weights are the similarities between the labels. The edge weight of the constructed graph is Sij, that is: Weight(vi, vj) = Sij, ∀i ∈ [1, n1], j ∈ [n1+1, n1+n2]. This not only makes the labels in each set related to each other, but also allows the similarities between labels across sets to be integrated.

[0175] Node Update of Graph Structure: The representation of nodes in the graph is updated through the representations of their neighbors. The vector representation of label ti at the (k + 1)-th layer of the neural network is: where, is the vector representation of label ti at the k-th layer, is the set of neighbors of label ti, is the weight between node i and neighbor node j, and W(k) is the weight matrix at the k-th layer, is the Sigmoid activation function. is the learning rate, is the gradient of the loss function with respect to the node representation of.

[0176] Finally, Label Matching and Merging: During the training process of the neural network, an optimization objective function is used to drive the merging of labels, ensuring that labels with high similarity are clustered into one category and avoiding excessive merging among labels of the same category. The loss function for label dissimilarity is expressed as: where, is a regular expression, is the final clustering center of label ti. The representations of labels will be gradually merged as the graph structure is updated. Finally, V1 and V2 are merged to form a new set of farmer portrait labels V3, retaining the independence and diversity of all labels in the original farmer group set.

[0177] is the label cluster obtained by clustering each label of.

[0178] Portrait Group Label Generation Module: The initial automatic construction of the farmer portrait group label set can be generated through statistical features based on TF-IDF and TextRank, or machine learning methods based on classifiers and feature engineering, or deep learning methods based on LDA, RNN, LSTM, KeyBERT, etc., or rules such as regular expressions and conditional random fields.

[0179] Farmer Portrait Classification Management Module: The management of farmer portrait classification includes farmer portrait name management and farmer portrait group label management.

[0180] Among them, the management of farmer portrait group tags: According to different farmer portrait classifications, extract keywords from the relevant information of farmers of the corresponding type as the tag set. Different farmer portrait groups correspond to different information tag keyword sets. First, these tag sets are automatically extracted from all the information keywords of this type of farmer household on the platform. Then, business experts can edit the tag keyword set, add, delete, or modify keywords to form the tag set for this type of farmer portrait group.

[0181] Agricultural service information push module: The agricultural service information on the platform is aggregated from multiple platforms. The platform regularly collects various agricultural service information in real time from relevant platforms such as the agricultural product trading platform, land resource management system, meteorological data platform, financial institution credit system, agricultural policy release platform, agricultural materials sales platform, production and marketing docking platform, and skills training platform, such as agricultural technology training courses, loan products, agricultural materials preferential activities, agricultural machinery sales information, etc., and stores them in the agricultural service information data resource library after preprocessing.

[0182] The agricultural service information push module includes an information tag automatic generation module, a manual push module, a to-be-pushed information management module, a historical information management module, an automatic push module, and a push rule setting module.

[0183] Among them, the tag automatic generation module: Extract agricultural service information data from the agricultural service information data resource library. Each piece of agricultural service information automatically extracts one or more keyword tags according to the title and content, and is arranged in descending order according to the information generation date. Each service information includes a generation date tag, an activity date tag, whether it is an online activity, whether registration is required, keyword tags, a recommended farmer portrait tag set to be pushed, a tag indicating whether it has expired, and a tag indicating whether it has been pushed.

[0184] The administrator can also manually mark a certain service information as an important tag.

[0185] If the deadline in the information is greater than the current date, the system automatically marks it as expired; each service information is defaulted to not pushed and is marked as pushed after being pushed. The expired and pushed information is automatically put into the historical information management module for display and management, and the others are put into the to-be-pushed information management module.

[0186] Generation of the recommended farmer portrait tag set to be pushed: The farmer portrait tag set is matched with the tag set of the service information. The keyword set of each service information is matched with the tag set of each farmer portrait type to form key-value pairs of (service information id, farmer portrait type id, matching degree). According to the matching degree, they are sorted from high to low, and the top 20% or the top 6 farmer portrait types with the highest matching degree are marked and included in the recommended farmer portrait tag set to be pushed according to the rules.

[0187] The information push methods are divided into two types: manual push and automatic push.

[0188] Manual push module, manual push method: the administrator manually selects the service information, push objects, sets the push time and push channels, and the information will be pushed to the farmer groups or individuals with the corresponding labels according to the set requirements.

[0189] Automatic push module, automatic push method: the system automatically pushes service information to the farmer groups in the recommended farmer portrait tag set. The administrator can set the automatic push time, push frequency, number of pushes per day, and priority rules for pushing information. For example, a maximum of 2 messages will be automatically pushed per day, holidays will be pushed as usual, and matching degree takes precedence over information generation date, etc.

[0190] Push rule setting module, push limit: In order to improve farmers' experience and sensitivity to push information, set the automatic push of up to 2 messages per day and the manual push of up to 3 service messages.

[0191] Based on the farmer portraits, we customize personalized service information push packages for different types of farmers. The packages are divided into two categories according to the farmer feedback form: one is the registration information that needs to be fed back after reaching the farmers, and the other is information for reading and learning only. Information includes: policy information, disaster warning information, agricultural material preferential activities information, loan product information, etc.

[0192] For those who need to feedback registration information after reaching farmers, manual push is generally used. For automatic push, the system automatically matches information and farmers and pushes according to the rules after setting push cycle, push time and other rules.

[0193] Service information and portrait matching module: The matching rules between service information and farmers are based on the service information package set in advance based on the farmer portrait. The farmer portrait includes basic information portrait, production characteristic portrait, economic characteristic portrait, risk characteristic portrait, behavioral characteristic portrait, etc. The package setting rules for service information are as follows: service information with a high degree of match between the service information and the farmer portrait group will be pushed first; service information with important tags will be pushed first; information with an activity date approaching will be pushed first.

[0194] For agricultural technology training courses, loan products, agricultural inputs promotions, agricultural machinery sales information, loan product information, etc., multiple keywords are extracted based on the source and content of information, and keyword tags are automatically set for the information, so as to match different farmer portrait groups according to the information keywords.

[0195] Different groups of farmer portraits preset label sets, which are obtained by extracting all information keywords of this type of farmer households on the platform. The label sets of farmer portrait groups can be generated by statistical features based on TF-IDF and TextRank, or by machine learning methods based on classifiers and feature engineering, or by deep learning methods based on LDA, RNN, LSTM, KeyBERT, etc., or by rules such as regular expressions and conditional random fields.

[0196] When calculating the matching degree between service information and farmer portrait groups, two main factors are considered. One is the number of overlapping keywords in the keyword set of service information and the keyword set of farmer portrait group information labels. The other is the proportion of the number of identical keywords in the keyword set to the total number of keywords in the set. The larger the value of the second factor, the better.

[0197] Calculation of the matching degree: The keyword set of service information and the label set of farmer portrait groups are defined as S and P respectively, and the matching degree is defined as: Among them, represents the proportion of overlapping keywords to the total number of keywords, that is, the Jaccard similarity, which is used to measure the degree of overlap of sets, and represents the proportion of the number of overlapping keywords to the size of the smaller set, emphasizing the absolute number of overlapping keywords. is the contribution weight of the two factors to the matching degree, and the platform default setting is 0.5.

[0198] Generate the matching degree of each service information and farmer household portrait according to the above rules.

[0199] When automatically pushing information, if at least 80% of the information keywords match the label set of the basic information portrait group, and 70% match the label set of any one of the farmer portrait groups such as production characteristics portrait, economic characteristics portrait, risk characteristics portrait, and behavior characteristics portrait, then the information can be pushed to the corresponding one or more portrait groups.

[0200] The platform sets a push limit for service information, with corresponding upper limits on the number of information pushes per day / month / year.

[0201] Ultimately, for large-scale farmers who grow corn, we can push service information such as high-end agricultural technology training for corn planting, promotion of advanced agricultural machinery and tools, and brand building of agricultural products; for low-credit-risk, online farmers who raise cattle, we can push animal husbandry loan product information, online training course information, and online beef cattle docking information; for users with high natural risks, we can push weather warnings, pest warning information and response plans, agricultural insurance information, etc. In addition, we can use a variety of information push channels, such as mobile phone text messages, mobile phone application notifications, WeChat messages, social media, etc., to push service information to farmers in a timely manner, and record the push information and farmers' feedback information at the same time.

[0202] The information push effect evaluation module includes a feedback information collection module and an information push effect analysis module. The feedback information collection module is responsible for collecting farmers' feedback on service information, including whether they have signed up for the pushed service, whether they have read the pushed information, the time and duration of reading the service information, records of participating in the pushed service activities, etc. It will record the service usage effect and satisfaction evaluation in the system, further update the farmer portrait and group label collection, and achieve continuous improvement of service push.

[0203] The information push effect analysis module conducts analysis based on the collected feedback data, grasps the information push reach, click-through rate, conversion rate, service satisfaction analysis, etc., provides optimization suggestions for the push strategy of farmers' services, and improves the push effect and service satisfaction.

[0204] This service information push system can collect and integrate multi-source information related to farmers in real time, breaking down system barriers and providing rich data resources for farmer portraits and service information. Farmer group portraits are generated by using two methods: farmer group classification formed by different feature fields and farmer information keyword clustering. The portraits are then re-clustered to generate different types of farmer portrait group label sets, which can more accurately tap into the business needs of different types of farmer groups. Service information is pushed to farmers based on the degree of match between service information and farmer portrait label sets, and in both manual and automatic ways, it is more in line with the service information acquisition needs of various farmer groups and the business needs of functional departments for actual information push management.

[0205] Figure 6The schematic diagram of the physical structure of an electronic device is exemplified. The electronic device may include: a processor 810, a communications interface 820, a memory 830, and a communication bus 840. Among them, the processor 810, the communications interface 820, and the memory 830 complete mutual communication through the communication bus 840. The processor 810 can call the logical instructions in the memory 830 to execute the service information pushing method, and the method includes: obtaining agricultural data; constructing a farmer portrait according to the agricultural data; matching target agricultural service information to be pushed according to the farmer portrait; and pushing the target agricultural service information to the corresponding farmer client according to the preset information pushing rules.

[0206] In addition, when the logical instructions in the above-mentioned memory 830 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0207] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the service information pushing method provided by the above-mentioned various methods. The method includes: obtaining agricultural data; constructing a farmer portrait according to the agricultural data; matching target agricultural service information to be pushed according to the farmer portrait; and pushing the target agricultural service information to the corresponding farmer client according to the preset information pushing rules.

[0208] On yet another hand, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the service information pushing method provided by the above-mentioned various methods. The method includes: obtaining agricultural data; constructing a farmer portrait according to the agricultural data; matching target agricultural service information to be pushed according to the farmer portrait; and pushing the target agricultural service information to the corresponding farmer client according to the preset information pushing rules.

[0209] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0210] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0211] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments or equivalently replace some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A service information push method, characterized in that: include: Access to agricultural data; constructing farmer profiles based on said agricultural data; matching the target agricultural service information to be pushed according to the farmer portrait; According to the preset information push rules, the target agricultural service information is pushed to the corresponding farmer user terminal.

2. The method according to claim 1, characterized in that The method further comprises: Collecting feedback data of farmers on the target service information through the farmer user terminal; The farmer portrait is updated based on the feedback data and the preset information push rules are optimized.

3. The method according to claim 1, characterized in that The step of constructing a farmer portrait based on the agricultural data includes: Farmer characteristic information is extracted from the agricultural data, and a farmer portrait is constructed based on the farmer characteristic information, wherein the farmer portrait includes a basic information portrait, a production characteristic portrait, an economic characteristic portrait, a risk characteristic portrait and a behavior characteristic portrait.

4. The method according to claim 3, characterized in that The extracting farmer characteristic information from the agricultural data and constructing farmer portraits based on the farmer characteristic information includes: Extracting farmers' basic information, production characteristic information, economic characteristic information, risk characteristic information and behavior characteristic information from the agricultural data; constructing a basic information portrait based on the farmer's basic information, constructing a production characteristic portrait based on the production characteristic information, constructing an economic characteristic portrait based on the economic characteristic information, constructing a risk characteristic portrait based on the risk characteristic information, and constructing a behavioral characteristic portrait based on the behavioral characteristic information; The basic information portrait, production characteristic portrait, economic characteristic portrait, risk characteristic portrait and behavioral characteristic portrait are aggregated to obtain a classification set of farmer portraits.

5. The method according to claim 3, characterized in that: The matching of the target agricultural service information to be pushed according to the farmer portrait includes: Collect different types of agricultural service information and multiple pieces of agricultural service information corresponding to each type in real time; Generate corresponding information tags according to the title and content of each piece of agricultural service information; Based on matching the information tag and the farmer portrait classification set, the target agricultural service information to be pushed corresponding to each farmer portrait classification set is obtained.

6. The method according to claim 5, characterized in that The preset information push rules include manual push mode, automatic push mode and push restriction conditions; The method of pushing the target agricultural service information to the corresponding farmer user terminal according to the preset information pushing rules includes: If the target agricultural service information needs to collect farmers' feedback information, it is determined to be pushed manually; According to the manual push setting requirements, the target agricultural service information is pushed to the corresponding farmer user terminal; If the target agricultural service information does not need to collect farmer feedback information, then automatically push it; According to the automatic push setting requirements, the target agricultural service information is pushed to the corresponding farmer user terminal, wherein the automatic push setting requirements include information push quantity requirements, time requirements, frequency requirements and priority push requirements.

7. A service information push system, characterized in that: include: Multi-source data acquisition module for obtaining agricultural data; A farmer portrait construction module, used to construct a farmer portrait based on the agricultural data; A service information and portrait matching module, used to match the target agricultural service information to be pushed according to the farmer portrait; The service information push module is used to push the target agricultural service information to the corresponding farmer user terminal according to the preset information push rules.

8. The system according to claim 7, characterized in that The system further comprises: The information push effect evaluation module is used to collect farmers' feedback data on the target service information through the farmer user terminal; update the farmer portrait based on the feedback data and optimize the preset information push rules.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the service information pushing method according to any one of claims 1 to 6 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the service information pushing method according to any one of claims 1 to 6 is implemented.

Images