Intelligent recommendation system for agricultural products based on climate and consumption behavior

Abstract

The present application relates to the technical field of electronic commerce recommendation, in particular to an intelligent agricultural product recommendation system based on climate and consumption behavior, which comprises the following steps: collecting user behavior, commodity category and time information, obtaining climate data through association positioning, extracting behavior and climate cross-field, constructing behavior timeline and climate link, identifying behavior and climate change consistent section, screening interest deviation direction, and combining inventory and transaction performance to sort and output recommended results. The present application introduces temperature, humidity, precipitation and solar term data of the geographical location where the user is located by fusing time and category information in user behavior, forms the cross-field of behavior and environment, and enhances the recognition of consumption situation change; the continuous behavior timeline is constructed and associated with regional climate indicators, and the behavior link with correlation is extracted; the consumption trend change segment is identified through the screening of behavior and climate change direction, and the goods are sorted and recommended in combination with interest deviation, inventory and transaction performance.

Description

Technical Field

[0001] This invention relates to the field of e-commerce recommendation technology, and in particular to an intelligent recommendation system for agricultural products based on climate and consumer behavior. Background Technology

[0002] The field of e-commerce recommendation technology involves matching and recommending products to users through user behavior analysis and product attribute modeling. Its core includes steps such as user data collection, feature extraction, similarity calculation, recommendation generation, and feedback updates. Commonly used methods include collaborative filtering and content recommendation to build personalized recommendation systems, improving recommendation accuracy and user satisfaction.

[0003] Traditional intelligent agricultural product recommendation systems rely on user consumption behavior and agricultural product sales data, using rule-based matching or similarity algorithms for recommendations. The main steps include organizing historical user data, classifying and statistically analyzing agricultural products, analyzing consumer preferences, and pushing agricultural products based on rules. These systems largely depend on static data and lack processing for external dynamic factors such as climate, making it difficult to provide real-time recommendations in response to adjustments in consumer behavior caused by climate change.

[0004] Traditional systems rely on static consumption records and category tags for rule-based recommendations, neglecting the impact of environmental variables on consumer behavior. Under atypical weather conditions, such as hot and humid or cold and dry weather, the system cannot identify changes in user preferences, leading to ineffective recommendations. Behavior processing is based on isolated events, lacking the ability to model the continuity and trends of behavior, resulting in delayed responses to shifts in interest. Data analysis is not linked to real-time climate characteristics and lacks a cross-modeling mechanism between behavior and external context, causing recommended content to be disconnected from actual user needs and making it difficult to support personalized recommendations in climate-sensitive consumption scenarios. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing an intelligent agricultural product recommendation system based on climate and consumer behavior.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent agricultural product recommendation system based on climate and consumer behavior, the system comprising, The behavioral climate data acquisition module acquires user browsing, shopping, and ordering behaviors, extracts product category and time fields, associates and reads regional temperature, humidity and solar terms, retrieves agricultural product category and climate adaptation fields, filters the intersection of behavior and climate, and obtains the data content corresponding to climate behavior. The corresponding mapping construction module extracts continuous behavior time to form a timeline based on user ID, product category and operation time in the data content corresponding to the climate behavior, associates geographic ID, temperature and humidity, connects behavior and climate with category and time, removes discontinuous segments, and obtains behavior-climate link structure. The consumption trend judgment module extracts changes in behavioral direction based on user paths in the behavioral climate link structure, compares behavioral directions at different time periods, combines temperature and humidity changes, filters out behaviors with consistent directions, extracts agricultural product paths in hot and humid and cold and dry areas, and obtains information on climate behavior linkage segments. The interest direction filtering module relies on user and category data in the climate behavior linkage section information to extract browsing behavior within a time range, associate it with geographical and climate text, screen categories and climate text, extract content with inconsistent directions and classify it to obtain a list of interest direction offset targets. The product order output module obtains products with climate tags based on the user's interest direction offset target list category, extracts the warehouse quantity text and transaction performance field, calls the content of the two fields to divide the candidate product order, adds recommendation descriptions to them and compiles them into the output template, and obtains the climate and consumption behavior linkage recommendation results.

[0007] As a further aspect of the present invention, the climate behavior corresponding data content includes a unique user identifier, product category, behavior occurrence time, geographical location, corresponding temperature value, relative humidity value, precipitation level, and solar term type. The behavior-climate link structure specifically includes continuous behavior time nodes, geographical location number, time-category mapping relationship, and behavior-climate cross-field set. The climate behavior linkage change segment information includes behavior direction change type, climate indicator change direction, path time period identifier, regional climate conditions, and agricultural product association type. The interest direction offset target list specifically includes user browsing preference categories, behavior trend characteristics, regional climate text tags, direction offset markers, and interest change categories. The climate and consumption behavior linkage recommendation results include matched product categories, recommendation priority sequence, inventory quantity information, sales performance indicators, and recommendation prompt text.

[0008] As a further aspect of the present invention, the behavioral climate acquisition module includes: The behavior extraction submodule acquires the user's operational behaviors during browsing, adding to cart, and placing orders. It collects product category and corresponding operation time information, calls device location content to locate the specific location, determines the region and date and time based on the location results and time content, and obtains spatiotemporal behavior location results. The regional climate association submodule collects the temperature, humidity, precipitation and solar term information of the corresponding region at the time point based on the region and time content in the spatiotemporal behavior positioning results, and sequentially associates each climate content with the product category content, extracts the corresponding items that are related to the operation time and climate information, and obtains the product climate association results. The climate suitability screening submodule calls each category and climate content involved in the commodity climate association results, compares the suitable temperature, humidity and precipitation ranges corresponding to each category in the agricultural product information, screens out the corresponding items that have intersection between the two types of data content, removes the remaining items, and obtains the data content corresponding to climate behavior.

[0009] As a further aspect of the present invention, the corresponding mapping construction module includes: The timeline extraction submodule obtains the user ID, product category and operation time from the data content corresponding to the climate behavior. It categorizes the operation time sequence content of the same user by product category, filters out broken segments in the time field and retains continuous segments, arranges them by time to form an operation trajectory sequence, and obtains a continuous behavior time chain. The climate association submodule, based on the time and product category content in the continuous behavior time chain, calls the temperature and humidity information under the corresponding geographic number, performs the corresponding linking operation according to time, and clears the misaligned items in the operation category and climate to obtain category climate connection pairs. The segment filtering submodule calls the time content and climate data combination in the category climate connection pair, determines whether there are missing items or incomplete connection content in the continuity, removes the missing parts from the operation trajectory, and only retains the temperature, humidity and behavior synchronization content to obtain the behavior climate link structure.

[0010] As a further aspect of the present invention, the consumption trend judgment module includes: The behavior direction extraction submodule obtains the continuous behavior path of each user in the behavior climate link structure of the agricultural product category, extracts the change description between adjacent behaviors from the operation direction field, judges the difference of the operation direction of the same user in different time periods according to the time sequence, and obtains the behavior change direction group after filtering out content with no direction change. The climate trend comparison submodule extracts the temperature and humidity data for the corresponding time period based on the time period in the behavior change direction group, makes a directional judgment on the temperature and humidity between two time nodes, and then compares the directional judgment result with the behavior change direction to obtain a set of linked and consistent segments. The regional path filtering submodule calls the time and climate direction correspondence information in the linked consistent segment set, divides the geographical region into hot and humid regions and cold and dry regions, extracts the corresponding segment content according to the product category, removes data segments that are irrelevant to the region and category, and obtains the climate behavior linkage change segment.

[0011] As a further aspect of the present invention, the interest direction filtering module includes: The browsing behavior extraction submodule obtains the user ID and product category from the climate behavior linkage change segment information, extracts the user's browsing behavior content within the corresponding time range, retains the operation start and end time and browsing category fields, clears data that does not correspond to the time period, and obtains the browsing behavior group within the time period. The text direction screening submodule extracts the climate type text content of the browsing area based on the product category field and operation time in the browsing behavior group within the time period, performs a direction comparison operation between the browsing category and the climate text, identifies browsing items with different directional content, and obtains the behavior climate direction difference segment. The trend offset classification submodule calls the product category field and user ID content in the behavior climate direction difference segment, classifies similar browsing items according to user operation trends, assigns the direction change path to a unified category label, organizes the classification content by user ID, and obtains the interest direction offset target list.

[0012] As a further aspect of the present invention, the product sequence output module includes: The candidate product extraction submodule obtains the product category content in the target list of interest direction offset, extracts all product entries with corresponding climate tags, filters out items with missing climate tags and items that are inconsistent with the category, and retains the candidate set to obtain the product group corresponding to the climate category. The sequential division processing submodule extracts the quantity text and transaction performance field content from the warehousing field according to the content of each product in the product group corresponding to the climate category. After performing numerical conversion processing on the two fields respectively, it performs superimposed sorting operation, divides the products into order according to the merged result, and obtains the product sorting order information. The recommended content output submodule calls the product number and category content in the product sorting order information, extracts the corresponding recommendation statement template content according to the sorting order, fills the recommendation statement into the corresponding product text output frame, and obtains the climate and consumption behavior linkage recommendation result by combining the recommendation content of each product.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by integrating time and category information from user browsing, adding to cart, and placing orders, and introducing temperature, humidity, precipitation, and solar term data corresponding to geographical location, a cross-description of behavior and environment is formed, enhancing the ability to perceive changes in consumption scenarios. By constructing a continuous behavior timeline and associating it with climate indicators of the same region, relevant behavioral links are retained, supporting behavioral evolution analysis. By comparing and filtering changes in behavioral direction with climate trends, relevant consumption change segments are identified, and product ranking is completed by combining interest shifts with inventory and transaction performance, improving the fit between recommended content and climate conditions and behavioral trends. Attached Figure Description

[0014] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the acquisition process of the behavioral climate data acquisition module of the present invention. Figure 3 This is a flowchart illustrating the process of obtaining the corresponding mapping construction module in this invention; Figure 4 This is a flowchart illustrating the acquisition process of the consumer trend judgment module of the present invention. Figure 5 This is a flowchart illustrating the acquisition process of the interest direction filtering module of the present invention. Figure 6 This is a flowchart illustrating the acquisition process of the product sequence output module of this invention. Detailed Implementation

[0015] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0016] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0017] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0018] Please see Figure 1 This invention provides a technical solution: an intelligent agricultural product recommendation system based on climate and consumer behavior, the system comprising: The behavior and climate data collection module acquires user behavior information during browsing, adding to cart, and placing orders. It extracts product category and operation time data, associates device location information to obtain temperature, humidity, precipitation, and solar term descriptions of the geographical area, calls the category and suitable climate fields in agricultural product information, filters the fields between the same behavior and climate, removes non-overlapping data and retains the cross-field results to obtain the data content corresponding to the climate behavior. The corresponding mapping construction module extracts continuous behavior time content to form timeline information based on user ID, product category and operation time in the data content corresponding to climate behavior. It associates temperature and humidity content under the same geographic ID, connects behavior and climate information according to time and category fields, removes missing fields or discontinuous segments and retains the content where operation behavior and climate content intersect, and obtains behavior-climate link structure. The consumption trend judgment module extracts change descriptions from the operation direction field based on the continuous behavior path of each user in the behavior climate link structure of agricultural product categories, compares the behavior activities in different time periods, extracts the change information of temperature and humidity in the same time period, filters out data segments whose behavior direction changes are consistent with the direction of climate change, and extracts the data paths corresponding to agricultural product types in hot and humid regions and dry and cold regions respectively, to obtain information on climate behavior linkage change segments. The interest direction filtering module relies on user and category data in the climate behavior linkage change segment information to extract the browsing behavior content of users within the corresponding time range, and associate it with the climate type text of their region. It performs a screening of the correspondence between behavior categories and climate text, extracts data content with different directions, classifies it according to behavior trends, and obtains a list of interest direction offset targets. The product order output module obtains product entries with corresponding climate tags based on the category content in the target list offset by the user's interest direction. It extracts the quantity text and transaction performance field of the product from the storage field, calls the content of the two fields to divide all candidate products into order, adds recommendation text descriptions to the divided product sets and compiles them into the output template framework, and obtains the climate and consumption behavior linkage recommendation results.

[0019] The data content corresponding to climate behavior includes a unique user identifier, product category, time of occurrence, geographical location, corresponding temperature value, relative humidity value, precipitation level, and solar term type. The behavior-climate link structure specifically includes continuous time nodes of behavior, geographical location number, time-category mapping relationship, and behavior-climate cross-field set. The information on climate behavior linkage change segments includes behavior direction change type, climate indicator change direction, path time period identifier, regional climate conditions, and agricultural product association type. The interest direction offset target list specifically includes user browsing preference categories, behavior trend characteristics, regional climate text tags, direction offset markers, and interest change categories. The climate and consumption behavior linkage recommendation results include matching product categories, recommendation priority sequence, inventory quantity information, sales performance indicators, and recommendation prompt text.

[0020] Please see Figure 2 The behavioral climate data acquisition module includes: The behavior extraction submodule acquires the user's operational behaviors during browsing, adding to cart, and placing orders. It collects product category and corresponding operation time information, calls device location content to locate the specific location, determines the region and date and time based on the location results and time content, and obtains spatiotemporal behavior location results. By retrieving the interaction footprint of user U2026011601 on the agricultural e-commerce platform in real time through a high-concurrency log server, the system accurately captured the specific operations performed by this user on the "Gannan Navel Orange" category within a specific time period from 10:15:20 to 10:45:30 on January 16, 2026. During the execution, the system recorded in detail three click-through behaviors on the product details page, two decisions to add oranges of different sizes to the shopping cart, and one order placement behavior to complete the payment. It also simultaneously captured the precise dwell time on the page for each operation; for example, the browsing stage took an average of 120 seconds, while the adding to the cart stage took only 15 seconds. The system then invoked the GPS interface deeply integrated into the mobile terminal to obtain detailed coordinate data of longitude 114.942 and latitude 25.851. By parsing the administrative division vector layer, it located the specific residential building location in Zhanggong District, Ganzhou City, Jiangxi Province, and combined with the system clock, categorized it as a winter morning time in East China. The system uses a three-dimensional overlay mapping of geospatial coordinates, administrative division attributes, and absolute time axis to exclude abnormal offline records with no location permissions or ambiguous location information, and obtains spatiotemporal behavior positioning results that include specific location coordinates, operation time series, product category attributes, and behavior intensity level.

[0021] The regional climate association submodule collects the temperature, humidity, precipitation and solar term information of the corresponding region at the time point based on the corresponding regional and time content in the spatiotemporal behavior positioning results. It then associates various climate contents with the product category content in sequence, extracts corresponding items that are related to the operation time and climate information, and obtains the product climate association results. Based on the geographic location code and the time scale of the behavior in the spatiotemporal positioning results, the system connects to the National Meteorological Monitoring Center via a secure application programming interface to collect micro-environmental parameters of the coordinate point at the corresponding time in real time. During execution, it obtains specific climate indicators returned by the observation station, such as a temperature of 8 degrees Celsius, relative humidity of 65%, hourly precipitation of 0 mm, and the current solar term being Minor Cold. Using "Gannan Navel Orange" as the core index item, a dynamic correlation mapping matrix is ​​established with the aforementioned meteorological factors. The system performs refined alignment, matching each user's operation time point with the minute-level observation window of the meteorological station, removing outdated meteorological data with observation time exceeding the set 3600-second safety threshold to ensure data timeliness. The system extracts correlation items that highly overlap with climate information such as temperature and humidity on the logical axis, thereby determining the user's specific consumption background characteristics in a low-temperature and humid environment. By analyzing the influence weight of temperature on the sensory evaluation of fresh agricultural products, it obtains commodity climate correlation results reflecting a strong physical correlation between commodity category demand and the immediate environmental climate.

[0022] The climate suitability screening submodule calls up each category and climate content involved in the commodity climate association results, compares the suitable temperature, humidity and precipitation ranges corresponding to each category in the agricultural product information, screens out the corresponding items with overlap between the two types of data content, removes the remaining items, and obtains the data content corresponding to climate behavior.

[0023] The system retrieves product categories from the product climate association results and compares them with real-time temperature, humidity, and precipitation data, then compares them against preset threshold ranges for growth, storage, transportation, and sensory suitability for consumption in the agricultural product knowledge graph. For example, the system extracts the suitable temperature range for Gannan navel oranges in winter as 5 to 15 degrees Celsius, the suitable humidity range as 50% to 70%, and the extreme weather precipitation threshold of 10 mm. A rigorous scalar comparison is performed, determining that the current temperature of 8 degrees Celsius and humidity of 65% both fall precisely within their corresponding closed ranges, and the 0 mm precipitation data is far below the precipitation threshold that would cause logistical disruptions or decreased consumer willingness. During execution, the system performs a logical AND operation between the real-time climate parameters and the suitable range boundary values. If the temperature parameter meets the range requirements, it is marked as valid; if the precipitation parameter exceeds the threshold, it is marked as interference and removed. By performing a deep intersection retrieval on the two types of data, the system automatically identifies and filters out redundant climate items such as air pressure, wind speed, and cloud cover that are in unsuitable ranges or have no significant impact on consumption decisions. Only the intersection of key climate factors that meet the suitability logic is retained, ensuring that subsequent analysis is based on a real and effective climate-driven background and obtaining corresponding data content that reflects the actual supporting role of climate in consumption decisions.

[0024] Please see Figure 3 The corresponding mapping building modules include: The timeline extraction submodule retrieves the user ID, product category, and operation time from the data content corresponding to climate behavior. It then categorizes the operation time sequence under the same user by product category, filters out broken segments in the time field and retains continuous segments, arranges them by time to form an operation trajectory sequence, and obtains a continuous behavior time chain. The system retrieves the user ID U2026011601, product category Gannan navel oranges, and all operation timestamps from the data corresponding to the user's behavior. It then logically reassembles the user's sporadic behaviors throughout the day in chronological order and calculates the intervals between adjacent operations to assess the continuity of the behavior. The system sets a continuity benchmark of 1800 seconds. By sliding a scan across the time span of each behavior point, it finds that the interval between the operations at 10:15 AM and 10:20 AM is 290 seconds, which is considered consistent with the shopping decision and belongs to the same behavioral segment. However, the interval between the operations at 10:20 AM and 11:30 AM reaches 4235 seconds, far exceeding the preset benchmark, and is judged as a behavioral mental interruption. During execution, the system physically filters out all isolated time fields corresponding to the identified breaks, retaining only continuous behavioral sequences with causal logical relationships within the benchmark range. By reconnecting the retained continuous segments in ascending order of time, a sequence of operation trajectories without breaks is formed, thereby eliminating the interference of users' aimless clicks during fragmented time periods and obtaining a continuous behavioral time chain that reflects the evolution of users' true shopping intentions.

[0025] The climate association submodule, based on the time and product category content in the continuous behavior time chain, calls the temperature and humidity information under the corresponding geographic number, performs the corresponding link operation according to time, and clears the misaligned items in the operation category and climate to obtain the category climate connection pair; Based on the time series and product category information in the continuous behavior time chain, the system performs a deep search of the detailed environmental meteorological records under the corresponding geographic region number 360700, extracting serialized values ​​such as temperature (8.1 degrees Celsius) and humidity (64.5%) for each minute between the start and end points of the time chain. The system performs a precise chain alignment operation, using the product category field as the association primary key and performing equal-step join mapping with the climate sampling data at each second level. During execution, the system performs quality checks on each join pair. If a behavior record for a certain second corresponds to a missing meteorological value, or if the humidity sensor returns illegal negative data, it is determined to be an misaligned item and is immediately cleared. By matching and aligning the operation category field and the climate field one by one, the system eliminates all heterogeneous data gaps caused by sensor jitter or data transmission packet loss, ensuring an absolute one-to-one correspondence between the behavior flow and the climate flow on the time axis, and obtaining multi-dimensional structured category-climate join pairs composed of timestamps, product categories, temperature values, and humidity values.

[0026] The segment filtering submodule calls the time content and climate data combination in the category climate connection pair, judges whether there are missing items and incomplete connection content in the continuity, removes the missing parts from the operation trajectory, and only retains the temperature, humidity and behavior synchronization content to obtain the behavior climate link structure.

[0027] The system combines the timeline and climate data from the category-based climate link pair and performs strict data integrity checks to eliminate substandard behavioral segments. With an integrity ratio threshold of 1.0, the system detects segments containing complete user click records but with a 60-second data gap between 10:18 and 10:19 due to meteorological station maintenance. This segment is deemed incomplete. The system physically removes these segments with missing, discontinuous, or low signal-to-noise ratio data from the overall operation trajectory, excluding them from subsequent trend calculations. Only segments where temperature and humidity sampling points and behavioral touch points are perfectly synchronized within microsecond error ranges are retained, ensuring extremely high purity of the input data source, eliminating statistical bias caused by missing data, and obtaining a behavioral climate link structure with high technical confidence and analytical value.

[0028] Please see Figure 4 The consumer trend judgment module includes: The behavior direction extraction submodule obtains the continuous behavior path of each user in the behavior climate link structure of agricultural product category, extracts the change description between adjacent behaviors from the operation direction field, judges the difference of the operation direction of the same user in different time periods according to the time sequence, and obtains the behavior change direction group after filtering out content without direction change. The module acquires the continuous behavioral paths of each user in the behavioral climate link structure of agricultural product categories, transforming user interaction types into directional descriptions with vector meaning. The system defines browsing as baseline value 1, adding to cart as enhancement value 2, and placing an order as peak value 3. By extracting the state transition amounts between adjacent behaviors, it calculates the changing trends of user intentions. During execution, the system compares the user's operational directions at different time periods. If, in the first time period, the user only wanders between browsing pages with a continuous behavioral change of 0, it is judged as content with no directional change and filtered. If, in the second time period, the user quickly switches from browsing to placing an order with a change of positive 2, it is identified as a significant enhancement direction. The system performs difference judgments on the operational directions of the same user at different time periods in chronological order, retaining the change sequences with clear intention transitions and eliminating random walk data without purchase purpose, thus obtaining a group of behavioral change directions that accurately represent the dynamic evolution trajectory of user consumption desires. The climate trend comparison submodule extracts the temperature and humidity data for the corresponding time period based on the time period in the behavior change direction group, makes a directional judgment on the temperature and humidity between two time nodes, and then compares the directional judgment result with the behavior change direction to obtain a set of linked and consistent segments. Based on the time period corresponding to the behavioral change direction group, the system extracts the temperature change value of 2.5 degrees Celsius and the humidity change value of 3% for the corresponding time period and executes directional discrimination logic. The system calculates the difference between the climate values ​​at the start and end points. If the difference is positive, the temperature is determined to be on an upward trend; otherwise, it is on a downward trend. This climate change direction is compared with the user's behavioral intention change direction. A linkage consistency coefficient is introduced and compared with a preset threshold of 0.7. During the execution process, if it is detected that the user's purchase of warm agricultural products increases simultaneously with the temperature drop, and the two show a high degree of coordination in the logical direction, it is determined to be a linkage consistent segment. This step accurately identifies the real consumption behavior driven by climate fluctuations by eliminating noisy data that is inconsistent with the climate fluctuation direction (such as users browsing cold drinks when the temperature drops sharply), and obtains a set of linkage consistent segments that reflect the resonance pattern between climate and behavior.

[0029] The regional path filtering submodule calls the time and climate direction correspondence information in the set of linked and consistent segments, divides the geographical region into hot and humid regions and cold and dry regions, extracts the corresponding segment content according to the product category, removes data segments that are irrelevant to the region and category, and obtains the climate behavior linkage change segment.

[0030] The system retrieves time and climate direction information from the consistent data set and, based on long-term environmental benchmarks such as an average annual temperature of 18 degrees Celsius and an average annual relative humidity of 68% for the geographical location, divides the operational background into hot and humid or cold and dry regions to ensure that the analysis results conform to regional climate characteristics. Behavioral segments are extracted from the corresponding regions for the product category of Gannan navel oranges, and secondary noise reduction is performed. This process involves the system thoroughly removing data that does not conform to the logic of navel orange growth environment and segments from product categories unrelated to climate sensitivity (such as hardware tools). Through multiple spatial filters based on geographical characteristics, product climate attributes, and consistency, the system removes a large amount of irrelevant cross-regional and cross-category data interference, ensuring the purity of the analysis sample for specific regions and products, and obtaining climate behavior linkage change segments that can directly reveal the climate-driven consumption mechanism.

[0031] Please see Figure 5 The interest-based filtering module includes: The browsing behavior extraction submodule obtains the user ID and product category from the climate behavior linkage change segment information, extracts the user's browsing behavior content within the corresponding time range, retains the start and end time of the operation and the browsing category field, clears data that does not correspond to the time period, and obtains the browsing behavior group within the time period. The system retrieves user ID U2026011601 and product category Gannan navel oranges from the climate behavior linkage change segment information. Within the corresponding time window of 10:15 to 10:45, it deeply retrieves all browsing behavior content of this user. The system not only records the browsed category names, but also simultaneously extracts the number of clicks on the product attribute page, the page scroll depth, and the dwell time on specific climate keywords (such as "clearing heat" and "resolving phlegm"). During the execution process, the system implements strict time boundary verification, classifying all browsing records with start or end timestamps outside the segment boundary as background noise and removing them. Only browsing data with operation start and end times highly overlapping with the linkage change segment is retained, ensuring that the extracted behaviors are all generated under the inducement of specific climate changes. Through this step, the system isolates the user's normal browsing habits, accurately identifies specific interests stimulated by real-time climate fluctuations, and obtains browsing behavior groups within a time period with extremely rich data dimensions and close spatiotemporal correlation.

[0032] The text direction screening submodule extracts the climate type text content of the browsing area based on the product category field and operation time in the browsing behavior group within the time period. It then performs a direction comparison operation between the browsing category and the climate text to identify browsing items with different directional content and obtain the behavior climate direction difference fragments. Based on the product category field and corresponding operation time in the browsing behavior group, the system extracts the climate type text content of the browsing area, such as meteorological descriptions like "sudden drop in temperature" and "surge in humidity." The system performs a complex directional comparison operation, semantically comparing the attribute text of the browsing category with the real-time climate text to identify logical deviations behind the behavior. During execution, the system uses a preset word vector model to calculate the "coolness" or "warmth" features of the user's current browsing behavior and compares them with the actual trend of the ambient temperature. If the system detects that the ambient temperature is rising, but the user is frequently browsing agricultural products with "warm" or "hot" attributes, it determines that the directional content of the two is different and identifies it as a behavioral climate direction difference segment. This step can effectively capture consumption signals that violate conventional environmental common sense, such as the user's special physiological or psychological need to seek hot food in hot weather, thus obtaining highly valuable interest shift signals.

[0033] The trend offset classification submodule calls the product category field and user ID content in the behavioral climate direction difference segment, classifies similar browsing items according to user operation trends, assigns the direction change path to a unified category tag, organizes the classified content by user ID, and obtains a list of interest direction offset targets.

[0034] The system retrieves product category fields and user IDs from behavioral climate direction difference segments and clusters similar browsing items based on the user's operational trend evolution during the observation period. A baseline value of 0.5 for offset intensity is set. By calculating the cumulative offset weight of a user across multiple difference segments (e.g., reaching 0.75), the system determines that the user's consumption interests have undergone a significant atypical shift. The system then groups product items with similar directional change paths (e.g., shifting towards heat or high-heat directions) into a unified category label and organizes them structurally according to user IDs, removing temporary interference caused by accidental mis-prompting or advertising inducements. By deeply quantifying and categorizing the user's interest shifts under climate interference, the system effectively identifies the user's potential real needs under special weather conditions, thus obtaining a target list of interest direction shifts that can support personalized and accurate recommendations.

[0035] Please see Figure 6 The product sequence output module includes: The candidate product extraction submodule obtains the product category content in the target list of interest direction offset, extracts all product entries with corresponding climate tags, filters out items with missing climate tags and items that are inconsistent with the category, and retains the candidate set to obtain the product group corresponding to the climate category. The system retrieves product category information from the target list of user interests and performs high-precision climate tag matching in the global product database. It iterates through all product entries, extracting feature tags such as "Vitamin C supplement," "cold resistance aid," and "hydration," which align with the current climate context and the user's interest shift. During execution, the system implements strict quality screening: if a product's climate tag field is missing, or if the tag content completely contradicts the current low-temperature and low-humidity climate requirements, the product entry is deemed invalid and removed. The system retains only products with consistent categories and a climate regulation attribute score higher than 0.8, forming a core candidate set. Through this two-way screening mechanism, the system ensures that candidate product groups not only match the user's list in terms of category but also specifically alleviate or adapt to the current climate environment in terms of functional attributes, resulting in highly suitable climate-category-corresponding product groups.

[0036] The sequential partitioning submodule extracts the quantity text and transaction performance field from the warehousing field based on the content of each product in the product group corresponding to the climate category. After performing numerical conversion on the two fields, it performs superimposed sorting operation and partitions the products according to the merged result to obtain the product sorting order information. Based on the static and dynamic attributes of each product in the corresponding product group of the climate category, the inventory balance of 500 from the warehousing field and the average conversion rate of 0.15 from the transaction performance field are extracted. These two dimensions are then dimensionless, mapping the inventory quantity to a supply score of 0 to 1 and the conversion rate to a popularity score of 0 to 1. A warehousing weight of 0.4 and a sales weight of 0.6 are introduced, and an overlay sorting operation is performed. During this process, the system calculates the comprehensive ranking score for each product (e.g., a product with a score of 0.68) and sorts them globally in descending order based on the combined score, placing products with sufficient inventory and excellent market feedback at the forefront of the recommendation sequence. This step effectively balances the platform's inventory turnover pressure with the user's purchase conversion probability, avoiding the recommendation of out-of-stock or low-quality products to users, thus obtaining product sorting order information that can directly guide the front-end display logic.

[0037] The recommended content output submodule calls the product number and category content from the product sorting information, extracts the corresponding recommendation statement template content according to the sorting order, fills the recommendation statement into the corresponding product text output frame, and combines the recommendation content of each product to obtain the climate and consumption behavior linkage recommendation results.

[0038] The system retrieves product IDs and category information from the product sorting information and extracts corresponding recommendation templates in a predetermined order. Based on the current 8-degree Celsius low temperature and the characteristics of the Lesser Cold solar term, the system fills the product name and core climate benefits (such as "Winter Warmth" and "King of Vitamin C") into a preset text output framework, performing semantic combination and formatted layout. During execution, the system logically concatenates the recommendation statements for each product to form a persuasive and interconnected recommendation message. By transforming the sorting results into highly interactive graphic and textual output content, the system empowers the entire process from raw climate data collection to consumer behavior conversion. This ensures that the recommended content not only possesses product attributes but also demonstrates care for the current climate, resulting in climate- and consumer behavior-linked recommendation results that significantly enhance user engagement.

[0039] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An intelligent agricultural product recommendation system based on climate and consumer behavior, characterized in that, The system includes: The behavioral climate data acquisition module acquires user browsing, shopping, and ordering behaviors, extracts product category and time fields, associates and reads regional temperature, humidity and solar terms, retrieves agricultural product category and climate adaptation fields, filters the intersection of behavior and climate, and obtains the data content corresponding to climate behavior. The corresponding mapping construction module extracts continuous behavior time to form a timeline based on user ID, product category and operation time in the data content corresponding to the climate behavior, associates geographic ID, temperature and humidity, connects behavior and climate with category and time, removes discontinuous segments, and obtains behavior-climate link structure. The consumption trend judgment module extracts changes in behavioral direction based on user paths in the behavioral climate link structure, compares behavioral directions at different time periods, combines temperature and humidity changes, filters out behaviors with consistent directions, extracts agricultural product paths in hot and humid and cold and dry areas, and obtains information on climate behavior linkage segments. The interest direction filtering module relies on user and category data in the climate behavior linkage section information to extract browsing behavior within a time range, associate it with geographical and climate text, screen categories and climate text, extract content with inconsistent directions and classify it to obtain a list of interest direction offset targets.

2. The intelligent agricultural product recommendation system based on climate and consumer behavior according to claim 1, characterized in that: The climate behavior data includes a unique user identifier, product category, time of occurrence, geographical location, corresponding temperature value, relative humidity value, precipitation level, and solar term type. The behavior-climate link structure specifically includes continuous time nodes of the behavior, geographical location number, time-category mapping relationship, and behavior-climate cross-field set. The climate behavior linkage change segment information includes behavior direction change type, climate indicator change direction, path time period identifier, regional climate conditions, and agricultural product association type. The interest direction offset target list specifically includes user browsing preference categories, behavior trend characteristics, regional climate text tags, direction offset markers, and interest change categories.

3. The intelligent agricultural product recommendation system based on climate and consumer behavior according to claim 1, characterized in that, The behavioral climate acquisition module includes: The behavior extraction submodule acquires the user's operational behaviors during browsing, adding to cart, and placing orders. It collects product category and corresponding operation time information, calls device location content to locate the specific location, determines the region and date and time based on the location results and time content, and obtains spatiotemporal behavior location results. The regional climate association submodule collects the temperature, humidity, precipitation and solar term information of the corresponding region at the time point based on the region and time content in the spatiotemporal behavior positioning results, and sequentially associates each climate content with the product category content, extracts the corresponding items that are related to the operation time and climate information, and obtains the product climate association results. The climate suitability screening submodule calls each category and climate content involved in the commodity climate association results, compares the suitable temperature, humidity and precipitation ranges corresponding to each category in the agricultural product information, screens out the corresponding items that have intersection between the two types of data content, removes the remaining items, and obtains the data content corresponding to climate behavior.

4. The intelligent agricultural product recommendation system based on climate and consumer behavior according to claim 1, characterized in that, The corresponding mapping construction module includes: The timeline extraction submodule obtains the user ID, product category and operation time from the data content corresponding to the climate behavior. It categorizes the operation time sequence content of the same user by product category, filters out broken segments in the time field and retains continuous segments, arranges them by time to form an operation trajectory sequence, and obtains a continuous behavior time chain. The climate association submodule, based on the time and product category content in the continuous behavior time chain, calls the temperature and humidity information under the corresponding geographic number, performs the corresponding linking operation according to time, and clears the misaligned items in the operation category and climate to obtain category climate connection pairs. The segment filtering submodule calls the time content and climate data combination in the category climate connection pair, determines whether there are missing items or incomplete connection content in the continuity, removes the missing parts from the operation trajectory, and only retains the temperature, humidity and behavior synchronization content to obtain the behavior climate link structure.

5. The intelligent agricultural product recommendation system based on climate and consumer behavior according to claim 1, characterized in that, The consumption trend judgment module includes: The behavior direction extraction submodule obtains the continuous behavior path of each user in the behavior climate link structure of the agricultural product category, extracts the change description between adjacent behaviors from the operation direction field, judges the difference of the operation direction of the same user in different time periods according to the time sequence, and obtains the behavior change direction group after filtering out content with no direction change. The climate trend comparison submodule extracts the temperature and humidity data for the corresponding time period based on the time period in the behavior change direction group, makes a directional judgment on the temperature and humidity between two time nodes, and then compares the directional judgment result with the behavior change direction to obtain a set of linked and consistent segments. The regional path filtering submodule calls the time and climate direction correspondence information in the linked consistent segment set, divides the geographical region into hot and humid regions and cold and dry regions, extracts the corresponding segment content according to the product category, removes data segments that are irrelevant to the region and category, and obtains the climate behavior linkage change segment.

6. The intelligent agricultural product recommendation system based on climate and consumer behavior according to claim 1, characterized in that, The interest direction filtering module includes: The browsing behavior extraction submodule obtains the user ID and product category from the climate behavior linkage change segment information, extracts the user's browsing behavior content within the corresponding time range, retains the operation start and end time and browsing category fields, clears data that does not correspond to the time period, and obtains the browsing behavior group within the time period. The text direction screening submodule extracts the climate type text content of the browsing area based on the product category field and operation time in the browsing behavior group within the time period, performs a direction comparison operation between the browsing category and the climate text, identifies browsing items with different directional content, and obtains the behavior climate direction difference segment. The trend offset classification submodule calls the product category field and user ID content in the behavior climate direction difference segment, classifies similar browsing items according to user operation trends, assigns the direction change path to a unified category label, organizes the classification content by user ID, and obtains the interest direction offset target list.

7. The intelligent agricultural product recommendation system based on climate and consumer behavior according to claim 1, characterized in that, The system also includes: The product order output module, based on the user's interest direction offset target clearing category, obtains products with climate tags, extracts the storage quantity text and transaction performance field, calls the content of the two fields to divide the candidate product order, adds recommendation descriptions to them and compiles them into the output template, and obtains the climate and consumption behavior linkage recommendation results. The climate and consumer behavior linked recommendation results include matching product categories, recommendation priority sequence, inventory quantity information, sales performance indicators, and recommendation prompt text.

8. The intelligent agricultural product recommendation system based on climate and consumer behavior according to claim 7, characterized in that, The product sequence output module includes: The candidate product extraction submodule obtains the product category content in the target list of interest direction offset, extracts all product entries with corresponding climate tags, filters out items with missing climate tags and items that are inconsistent with the category, and retains the candidate set to obtain the product group corresponding to the climate category. The sequential division processing submodule extracts the quantity text and transaction performance field content from the warehousing field according to the content of each product in the product group corresponding to the climate category. After performing numerical conversion processing on the two fields respectively, it performs superimposed sorting operation, divides the products into order according to the merged result, and obtains the product sorting order information. The recommended content output submodule calls the product number and category content in the product sorting order information, extracts the corresponding recommendation statement template content according to the sorting order, fills the recommendation statement into the corresponding product text output frame, and obtains the climate and consumption behavior linkage recommendation result by combining the recommendation content of each product.

Images