Data resource management system and method based on artificial intelligence

Abstract

The invention discloses a data resource management system and method based on artificial intelligence, and belongs to the technical field of artificial intelligence. According to the method, multi-source data is collected, and data cleaning and preprocessing are carried out; selecting meteorological indexes and soil indexes, and calculating an influence factor of each index in the meteorological indexes and the soil indexes; determining the weight of each influence factor, and calculating to obtain an agricultural product external environment influence factor; historical data are collected, meteorological data and soil data are matched with agricultural product planting areas and planting periods, and a multi-dimensional data set is constructed; performing feature engineering on the multi-dimensional data set to obtain a feature set; training a long short-term memory network model, and performing model evaluation and optimization; after training is completed, meteorological data and soil data are input into the model, and agricultural product yield prediction data are output; and performing data alignment, comparing the predicted agricultural product yield value with the average value of the market demand, and analyzing a confidence interval to obtain a decision suggestion.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based data resource management system and method. Background Art

[0002] In today's complex and highly data-driven agricultural product supply chain, accurate agricultural product output forecasting and reasonable market supply and demand matching have always been core challenges. With the rapid development of science and technology, technologies such as big data and artificial intelligence have gradually penetrated into agricultural production and circulation.

[0003] In the past, the impact of meteorological and soil conditions on the growth of agricultural products was mostly empirical and qualitative, lacking quantitative analysis methods, making it difficult to accurately measure the impact of each factor and unable to provide accurate numerical basis for planting adjustments. Existing technologies often view individual meteorological or soil factors in isolation, ignoring their interactions and their dynamic impact on different growth stages of agricultural products. Existing technologies often use simple linear regression or empirical formulas to estimate agricultural product yields, which cannot effectively capture the nonlinear relationship between complex factors such as meteorology and soil and yields, resulting in poor prediction accuracy, especially in years with extreme climate and soil anomalies. Summary of the invention

[0004] The purpose of the present invention is to provide a data resource management system and method based on artificial intelligence to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions: A data resource management method based on artificial intelligence, the method comprising the following steps: Collect multi-source data, including meteorological data, soil data, market demand data and agricultural product data, and perform data cleaning and preprocessing; Select meteorological indicators and soil indicators, calculate the influencing factors of each of the meteorological indicators and soil indicators; determine the weight of each influencing factor, and calculate the influencing factors of the external environment of agricultural products; Collect historical data, match meteorological data and soil data with agricultural product planting areas and planting cycles according to time and space dimensions, and construct a multidimensional data set; perform feature engineering on the multidimensional data set to obtain a feature set, and construct interactive features based on the external environmental influencing factors of the agricultural products and incorporate them into the feature set; Train the long short-term memory network model, and conduct model evaluation and optimization; after training, input meteorological data and soil data into the model, and output agricultural product yield forecast data, including the predicted agricultural product yield value and confidence interval; The agricultural product output forecast data and the market demand data are aligned in time and space dimensions; the average value of the predicted agricultural product output values ​​is compared with the average value of the predicted market demand, the confidence interval is analyzed, and decision-making recommendations are obtained.

[0006] Access the meteorological agency network platform through API to obtain meteorological data; use the soil sensor network to monitor soil data; Through web crawler technology, we collect sales data related to agricultural products on various e-commerce platforms every day and perform structured processing; collect transaction data from wholesale agricultural product markets, including transaction volume, transaction price, and circulation of agricultural products from different origins; obtain predicted market demand and consumer trend analysis based on agricultural product industry reports and information released by market research institutions; and integrate to form market demand data; Obtain corresponding agricultural product data during the planting stage, growth process and after harvesting of agricultural products.

[0007] Furthermore, in the meteorological data, temperature, humidity, rainfall, sunshine duration and wind speed are selected as meteorological indicators, and in the soil data, soil moisture, soil fertility, soil pH and soil texture are selected as soil indicators; According to the types of different agricultural products, the meteorological index ranges and soil index ranges suitable for the growth of the agricultural products are collected, it is determined at which stage of the meteorological index range and the soil index range the current meteorological index and the soil index are, and the impact of meteorological index and soil index on the growth of agricultural products is qualitatively graded.

[0008] Furthermore, obtain the meteorological data and soil data related to the current planting areas and varieties of agricultural products in the past few years, as well as the corresponding agricultural product yield data and quality data; conduct preliminary statistical analysis on the collected historical data, and calculate the correlation coefficients between various meteorological indicators and soil indicators and agricultural product yield and quality; Furthermore, the impact factor is calculated for each indicator based on the quantitative classification results of the impact of meteorological indicators and soil indicators and the statistical analysis of historical data; the weight of each impact factor is determined based on the judgment of the degree of each indicator based on the strength of the correlation between each indicator and the output and quality of agricultural products in the statistical analysis of historical data, combined with the current growth stage of agricultural products; the calculated impact factor of each meteorological indicator is multiplied by its corresponding weight, and the impact factor of each soil indicator is multiplied by its corresponding weight, and all the products are added together to obtain the external environment impact factor of agricultural products.

[0009] Collect historical meteorological data, soil data and corresponding agricultural product yield data, and match meteorological data and soil data with planting periods in chronological order based on the planting cycle of agricultural products; for annual crops, correspond daily meteorological data and soil data to corresponding growth stages starting from the sowing date, including sowing period, seedling period, growth period, flowering period, fruiting period and harvest period; for perennial crops, match meteorological data and soil data for many years according to their growth rhythm; Furthermore, based on the geographic coordinate information of the agricultural product planting plots, the meteorological data and soil data are located to the corresponding plots; using the geographic information system, the location of each plot is marked on the map to obtain the geographic location data, and the corresponding meteorological data and soil data are associated with the geographic location data; when there is a meteorological station covering several plots, the inverse distance weighted method is used to extrapolate the meteorological data to each plot.

[0010] Furthermore, the matched meteorological data, soil data and agricultural product yield data are stored in the form of database tables to construct a multi-dimensional data set; Performing feature engineering on the multidimensional data set, including extracting numerical features, categorical features, time series features and correlation coefficients between the features; calculating the correlation coefficients between the external environmental influencing factors of agricultural products and the extracted different features, constructing interactive features, and incorporating the interactive features into the feature set.

[0011] The number of features in the feature set is used as the number of nodes in the input layer of the long short-term memory network, the number of hidden layers and the number of nodes are set, the output layer is configured, the Identity function is selected as the activation function, the Adam optimizer is selected as the optimizer, the feature set is divided into a training set, a validation set and a test set according to the proportion, and the long short-term memory network model is trained; the model is evaluated using the test set to optimize the hyperparameters; After the training is completed, the meteorological data and soil data collected in real time are subjected to feature engineering and input into the trained long short-term memory network model; the model outputs the predicted agricultural product yield value; for the calculation of the confidence interval, the Bootstrap method is used to sample multiple times with replacement from the training set to generate multiple pseudo training sets with the same size as the original training set, and the pseudo training sets are used to retrain multiple long short-term memory network models to obtain a series of yield prediction values, and the confidence interval is calculated by statistically analyzing the distribution of the yield prediction values.

[0012] Furthermore, for agricultural product output forecast data, the starting time is extracted and end time ; Collect sales data from e-commerce platforms on a daily basis and record the earliest collection time and the latest collection time ; For the transaction data of the agricultural product wholesale market, determine the earliest transaction record time and the latest trading time ; Obtain forecasts for a period of time in the future from industry reports and market research information, and extract the starting forecast time involved and end prediction time ; Determine the common start time and common end time ; The time resolution of e-commerce platform data is daily. When the yield forecast data is given by week, month or planting season, time aggregation is required. For wholesale market transaction data, and , intercept the data of the corresponding time period, and when its time scale is inconsistent with the output forecast, perform aggregation or split processing; when the industry report data gives an annual forecast, the annual forecast is split into quarterly forecasts and monthly forecasts through linear interpolation method; for the confidence interval of the output forecast, when time aggregation is performed, the upper and lower limits of the confidence interval are also aggregated accordingly; Furthermore, the yield forecast data is coded by administrative division to obtain the coding set ,in, For the 1st to The coding of production forecast data; mapping the purchase area data of the e-commerce platform to the standard geographical area, using the address resolution function of the geographic information system, converting the fuzzy purchase address into the standard administrative area code to form a regional set ,in, For the 1st to The standard administrative region code for purchasing regional data on e-commerce platforms; the different origins in the agricultural product wholesale market transaction data are mapped to standard regional codes, recorded as ,in, For the 1st to Standard administrative region codes for agricultural product wholesale market transaction data; when and , aggregate the sales data of the e-commerce platform and wholesale market corresponding to the same area, and match the production forecast area; and , based on the planting area and historical production proportion, the e-commerce data is divided into sub-areas for production forecast.

[0013] Calculate the average of the predicted agricultural production values and the average of the predicted market demand ,Compare and ,like , the probability of oversupply increases by A%; if , the probability of supply exceeding demand increases by A%; obtain the confidence interval, if the lower limit of the confidence interval of agricultural product output is higher than the upper limit of market demand, the probability of oversupply increases by B%; if the upper limit of the confidence interval of agricultural product output is lower than the lower limit of market demand, the probability of supply exceeding demand increases by B%; where A and B are positive integers set by the user; draw the change curve of agricultural product output forecast and market demand in time series, calculate the supply and demand ratio of agricultural product output and market demand; input the supply and demand ratio, the probability of oversupply and the probability of oversupply into the decision suggestion database for search to obtain decision suggestions.

[0014] A data resource management system based on artificial intelligence, comprising: Data collection and preprocessing module: including: a data collection unit and a data preprocessing unit; wherein the data collection unit collects multi-source data, the multi-source data includes meteorological data, soil data, market demand data and agricultural product data, and the data preprocessing unit performs data cleaning and preprocessing; Impact factor calculation module: including: indicator impact factor calculation unit and external environment impact factor comprehensive calculation unit; wherein, the indicator impact factor calculation unit selects meteorological indicators and soil indicators, and calculates the impact factor of each indicator in the meteorological indicators and soil indicators; the external environment impact factor comprehensive calculation unit determines the weight of each impact factor, and calculates the external environment impact factor of agricultural products; Dataset construction and feature engineering module: including: a multi-dimensional dataset construction unit and a feature engineering unit; wherein the multi-dimensional dataset construction unit collects historical data, matches meteorological data and soil data with agricultural product planting areas and planting cycles according to time and space dimensions, and constructs a multi-dimensional dataset; the feature engineering unit performs feature engineering on the multi-dimensional dataset to obtain a feature set, constructs interactive features based on the external environmental influencing factors of the agricultural products, and incorporates them into the feature set; Model training and prediction module: including: model construction and training unit, model evaluation and optimization unit and yield prediction execution unit; among them, the model construction and training unit trains the long short-term memory network model, and the model evaluation and optimization unit performs model evaluation and optimization; after the training is completed, the yield prediction execution unit inputs meteorological data and soil data into the model, and outputs agricultural product yield prediction data, including the predicted agricultural product yield value and confidence interval; Data alignment and decision-making module: including: a data alignment unit and a decision suggestion generation unit; wherein the data alignment unit aligns the agricultural product output forecast data with the market demand data in time and space dimensions; the decision suggestion generation unit compares the average value of the predicted agricultural product output value with the average value of the predicted market demand, analyzes the confidence interval, and obtains decision suggestions.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention selects key meteorological and soil indicators, combines historical data, agricultural knowledge and real-time monitoring, and quantifies the impact of each indicator on the growth of agricultural products. For example, it accurately classifies the temperature impact according to the suitable growth range of different crops, and provides a precise numerical reference for planting adjustments.

[0016] 2. The present invention calculates the influencing factors of various indicators and reasonably allocates weights to obtain the external environmental influencing factors of agricultural products, comprehensively considers the synergistic and dynamic effects of meteorological and soil factors, and accurately reflects the advantages and disadvantages of the environment. Growers can predict the yield trend in advance and optimize agricultural operations in a targeted manner.

[0017] 3. The present invention adopts a long short-term memory network deep learning model. With its powerful learning and prediction capabilities for time series data, it deeply mines the complex nonlinear relationship between meteorological and soil historical data and agricultural product yields, and outputs yield prediction values ​​and confidence intervals. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the steps of a data resource management method based on artificial intelligence of the present invention; Figure 2 It is a system structure diagram of a data resource management system based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution. According to one embodiment of the present invention, Figure 1 A step diagram of a data resource management method based on artificial intelligence is shown in the figure, a data resource management method based on artificial intelligence, the method comprises the following steps: Collect multi-source data, including meteorological data, soil data, market demand data and agricultural product data, and perform data cleaning and preprocessing; Select meteorological indicators and soil indicators, calculate the influencing factors of each of the meteorological indicators and soil indicators; determine the weight of each influencing factor, and calculate the influencing factors of the external environment of agricultural products; Collect historical data, match meteorological data and soil data with agricultural product planting areas and planting cycles according to time and space dimensions, and construct a multidimensional data set; perform feature engineering on the multidimensional data set to obtain a feature set, and construct interactive features based on the external environmental influencing factors of the agricultural products and incorporate them into the feature set; Train the long short-term memory network model, and conduct model evaluation and optimization; after training, input meteorological data and soil data into the model, and output agricultural product yield forecast data, including the predicted agricultural product yield value and confidence interval; The agricultural product output forecast data and the market demand data are aligned in time and space dimensions; the average value of the predicted agricultural product output values ​​is compared with the average value of the predicted market demand, the confidence interval is analyzed, and decision-making recommendations are obtained.

[0021] Access the meteorological agency network platform through API to obtain meteorological data; use the soil sensor network to monitor soil data; Through web crawler technology, we collect sales data related to agricultural products on various e-commerce platforms every day and perform structured processing; collect transaction data from wholesale agricultural product markets, including transaction volume, transaction price, and circulation of agricultural products from different origins; obtain predicted market demand and consumer trend analysis based on agricultural product industry reports and information released by market research institutions; and integrate to form market demand data; Obtain corresponding agricultural product data during the planting stage, growth process and after harvesting of agricultural products.

[0022] In this embodiment, through cooperation with the local meteorological department and the five micro-meteorological stations set up in the tomato planting area, the daily temperature, humidity, rainfall, sunshine duration and wind speed data of the past five years (2019-2023) were collected. The weather station data is updated every 30 minutes to ensure that subtle meteorological changes can be captured. In the plantation area, 50 soil sensors are deployed at a density of one sampling point per 10 mu to monitor soil moisture, soil fertility (nitrogen, phosphorus, potassium content), soil acidity (pH value) and soil texture in real time (the proportion of sand, loam and clay is determined by particle analysis sensors). At the same time, professional soil sampling is carried out once a quarter and sent to the laboratory for testing of trace element content for calibration of sensor data. The planting area, variety, planting time, expected yield, actual yield and quality classification of tomatoes in the park in these five years are recorded (based on fruit size, color, sweetness, hardness and other indicators, divided into four grades: excellent, good, medium and poor). Capture daily tomato sales data from mainstream e-commerce platforms, including sales volume, sales price, and purchase region; establish data connection with local agricultural product wholesale markets to obtain daily transaction volume, transaction price, and circulation of tomatoes from different origins; collect annual market demand forecasts and consumer trend analysis for tomatoes from agricultural product industry reports.

[0023] Based on agricultural research literature and the park's many years of planting experience, the range of meteorological indicators suitable for tomato growth is determined as follows: Temperature: 15℃-30℃ is the suitable range. Below 10℃ means severe low temperature inhibition, 10℃-15℃ means slight low temperature inhibition, above 30℃ means slight high temperature impact, and above 35℃ means severe high temperature impact.

[0024] Humidity: 60%-80% is the suitable range, below 40% is severely insufficient humidity, 40%-60% is insufficient humidity, above 80% is high humidity, and above 90% is severely high humidity.

[0025] Rainfall: During the seedling stage (1-20 days after sowing), 20-30 mm per week is appropriate. Less than 10 mm indicates severe drought stress, 10-20 mm indicates mild drought stress, more than 30 mm indicates mild flood risk, and more than 40 mm indicates severe flood risk; during the flowering period (21-40 days after sowing)... and so on, different growth stages have different suitable rainfall ranges and classifications.

[0026] Light exposure time: 8-12 hours a day is the suitable range. Less than 6 hours is severely insufficient light, 6-8 hours is insufficient light, more than 12 hours is slightly excessive light, and more than 14 hours is severely excessive light.

[0027] Wind speed: During the fruiting period (41-60 days after sowing), a wind speed less than 5 m / s means no impact, 5-8 m / s means low risk, 8-10 m / s means medium risk, and greater than 10 m / s means high risk.

[0028] Determine the soil index range suitable for tomato growth: Soil moisture: 40%-60% is the suitable range, below 30% the soil moisture is severely insufficient, 30%-40% the soil moisture is insufficient, above 60% the soil moisture is high, and above 70% the soil moisture is severely high.

[0029] Soil fertility: Nitrogen content is in the suitable range of 0.1%-0.2%, phosphorus content is in the suitable range of 0.05%-0.1%, and potassium content is in the suitable range of 0.15%-0.25%. 50% below the lower limit of the suitable range is severely insufficient, 20% below the lower limit is insufficient, and 20% above the upper limit is excessive... Phosphorus, potassium and other trace elements are graded similarly.

[0030] Soil pH: The suitable range is 6-7. Less than 5 is too acidic, 5-6 is suitable acidity, 7-8 is suitable alkalinity, and greater than 8 is too alkaline.

[0031] Soil texture: Loam is the most suitable, sandy soil has a suitability of 0.8, and clay has a suitability of 0.7. The more it deviates from the characteristics of loam, the lower the suitability.

[0032] Using statistical software, we conducted a correlation analysis on the meteorological data, soil data, and tomato yield and quality data over the past five years. The results showed that in terms of yield, the Pearson correlation coefficient between temperature and yield was 0.6 during the flowering period (positive correlation, indicating that suitable temperature promotes flowering and fruiting, thereby increasing yield), and the correlation coefficient between nitrogen content in soil fertility and yield during the fruit expansion period was 0.7 (strong positive correlation, nitrogen has a significant effect on fruit growth); in terms of quality, the correlation coefficient between light duration and fruit sweetness was 0.5 (positive correlation, sufficient light is conducive to sugar accumulation), and the correlation coefficient between soil pH and fruit hardness was -0.4 (negative correlation, excessive acidity or alkalinity affects the formation of fruit cell walls and reduces hardness).

[0033] Taking the real-time data of a certain day as an example, the temperature on that day is 25℃, which is in the suitable range, and its temperature impact factor is 1; the humidity is 70%, which is in the suitable range, and the humidity impact factor is 1; the cumulative rainfall this week is 25 mm, which is in the suitable range during the seedling stage, and the rainfall impact factor is 1... and so on, determine the impact factor of the day for each meteorological and soil indicator.

[0034] Based on the correlation analysis of historical data and the current growth stage of tomatoes (in the flowering stage), the weights of each indicator are determined. The temperature weight is set to 0.25, the humidity weight is 0.15, the rainfall weight is 0.1 (the rainfall weight is higher in the seedling stage and relatively lower in the flowering stage), the light duration weight is 0.2, and the wind speed weight is 0.05; the soil moisture weight is 0.1, the soil fertility weight is 0.1 (focusing on the role of nitrogen, phosphorus, and potassium in promoting flowering), the soil pH weight is 0.03, and the soil texture weight is 0.02. The total weight is 1.

[0035] Multiply each meteorological index influence factor by its weight, multiply each soil index influence factor by its weight, and then add them together. The temperature influence factor of the day is 1×0.25+humidity influence factor 1×0.15+…+soil texture influence factor (0.8×0.02)=0.78, and the external environment influence factor of agricultural products on that day is obtained.

[0036] Collect historical meteorological data, soil data and corresponding agricultural product yield data, and match meteorological data and soil data with planting periods in chronological order based on the planting cycle of agricultural products; for annual crops, correspond daily meteorological data and soil data to corresponding growth stages starting from the sowing date, including sowing period, seedling period, growth period, flowering period, fruiting period and harvest period; for perennial crops, match meteorological data and soil data for many years according to their growth rhythm; Furthermore, based on the geographic coordinate information of the agricultural product planting plots, the meteorological data and soil data are located to the corresponding plots; using the geographic information system, the location of each plot is marked on the map to obtain the geographic location data, and the corresponding meteorological data and soil data are associated with the geographic location data; when there is a meteorological station covering several plots, the inverse distance weighted method is used to extrapolate the meteorological data to each plot.

[0037] Furthermore, the matched meteorological data, soil data and agricultural product yield data are stored in the form of database tables to construct a multi-dimensional data set; Performing feature engineering on the multidimensional data set, including extracting numerical features, categorical features, time series features and correlation coefficients between the features; calculating the correlation coefficients between the external environmental influencing factors of agricultural products and the extracted different features, constructing interactive features, and incorporating the interactive features into the feature set.

[0038] The number of features in the feature set is used as the number of nodes in the input layer of the long short-term memory network, the number of hidden layers and the number of nodes are set, the output layer is configured, the Identity function is selected as the activation function, the Adam optimizer is selected as the optimizer, the feature set is divided into a training set, a validation set and a test set according to the proportion, and the long short-term memory network model is trained; the model is evaluated using the test set to optimize the hyperparameters; After the training is completed, the meteorological data and soil data collected in real time are subjected to feature engineering and input into the trained long short-term memory network model; the model outputs the predicted agricultural product yield value; for the calculation of the confidence interval, the Bootstrap method is used to sample multiple times with replacement from the training set to generate multiple pseudo training sets with the same size as the original training set, and the pseudo training sets are used to retrain multiple long short-term memory network models to obtain a series of yield prediction values, and the confidence interval is calculated by statistically analyzing the distribution of the yield prediction values. In this embodiment, the average yield of tomatoes harvested this week is predicted to be 10 tons, and the 95% confidence interval is [8,12] tons.

[0039] Furthermore, for agricultural product output forecast data, the starting time is extracted and end time ; Collect sales data from e-commerce platforms on a daily basis and record the earliest collection time and the latest collection time ; For the transaction data of the agricultural product wholesale market, determine the earliest transaction record time and the latest trading time ; Obtain forecasts for a period of time in the future from industry reports and market research information, and extract the starting forecast time involved and end prediction time ; Determine the common start time and common end time ; The time resolution of e-commerce platform data is daily. When the yield forecast data is given by week, month or planting season, time aggregation is required. For wholesale market transaction data, and , intercept the data of the corresponding time period, and when its time scale is inconsistent with the output forecast, perform aggregation or split processing; when the industry report data gives an annual forecast, the annual forecast is split into quarterly forecasts and monthly forecasts through linear interpolation method; for the confidence interval of the output forecast, when time aggregation is performed, the upper and lower limits of the confidence interval are also aggregated accordingly; Furthermore, the yield forecast data is coded by administrative division to obtain the coding set ,in, For the 1st to The coding of production forecast data; mapping the purchase area data of the e-commerce platform to the standard geographical area, using the address resolution function of the geographic information system, converting the fuzzy purchase address into the standard administrative area code to form a regional set ,in, For the 1st to The standard administrative region code for purchasing regional data on e-commerce platforms; the different origins in the agricultural product wholesale market transaction data are mapped to standard regional codes, recorded as ,in, For the 1st to Standard administrative region codes for agricultural product wholesale market transaction data; when and , aggregate the sales data of the e-commerce platform and wholesale market corresponding to the same area, and match the production forecast area; and , based on the planting area and historical production proportion, the e-commerce data is divided into sub-areas for production forecast.

[0040] Calculate the average of the predicted agricultural production values and the average of the predicted market demand ,Compare and ,like , the probability of oversupply increases by A%; if , the probability of supply exceeding demand increases by A%; obtain the confidence interval, if the lower limit of the confidence interval of agricultural product output is higher than the upper limit of market demand, the probability of oversupply increases by B%; if the upper limit of the confidence interval of agricultural product output is lower than the lower limit of market demand, the probability of supply exceeding demand increases by B%; where A and B are positive integers set by the user; draw the change curve of agricultural product output forecast and market demand in time series, calculate the supply and demand ratio of agricultural product output and market demand; input the supply and demand ratio, the probability of oversupply and the probability of oversupply into the decision suggestion database for search to obtain decision suggestions.

[0041] According to another embodiment of the present invention, Figure 2 As shown in the system structure diagram of a data resource management system based on artificial intelligence, a data resource management system based on artificial intelligence includes: Data collection and preprocessing module: including: a data collection unit and a data preprocessing unit; wherein the data collection unit collects multi-source data, the multi-source data includes meteorological data, soil data, market demand data and agricultural product data, and the data preprocessing unit performs data cleaning and preprocessing; Impact factor calculation module: including: indicator impact factor calculation unit and external environment impact factor comprehensive calculation unit; wherein, the indicator impact factor calculation unit selects meteorological indicators and soil indicators, and calculates the impact factor of each indicator in the meteorological indicators and soil indicators; the external environment impact factor comprehensive calculation unit determines the weight of each impact factor, and calculates the external environment impact factor of agricultural products; Dataset construction and feature engineering module: including: a multi-dimensional dataset construction unit and a feature engineering unit; wherein the multi-dimensional dataset construction unit collects historical data, matches meteorological data and soil data with agricultural product planting areas and planting cycles according to time and space dimensions, and constructs a multi-dimensional dataset; the feature engineering unit performs feature engineering on the multi-dimensional dataset to obtain a feature set, constructs interactive features based on the external environmental influencing factors of the agricultural products, and incorporates them into the feature set; Model training and prediction module: including: model construction and training unit, model evaluation and optimization unit and yield prediction execution unit; among them, the model construction and training unit trains the long short-term memory network model, and the model evaluation and optimization unit performs model evaluation and optimization; after the training is completed, the yield prediction execution unit inputs meteorological data and soil data into the model, and outputs agricultural product yield prediction data, including the predicted agricultural product yield value and confidence interval; Data alignment and decision-making module: including: a data alignment unit and a decision suggestion generation unit; wherein the data alignment unit aligns the agricultural product output forecast data with the market demand data in time and space dimensions; the decision suggestion generation unit compares the average value of the predicted agricultural product output value with the average value of the predicted market demand, analyzes the confidence interval, and obtains decision suggestions.

[0042] In this embodiment, the apple yield forecast data is based on the forecast of different growth stages of the current planting season (March-October 2024), with the record start time being March 1, 2024, the end time being October 31, 2024, and the forecast time interval being once a week.

[0043] Consumer purchase data captured by e-commerce platforms will be collected starting from January 1, 2024, and updated daily, with the latest date being October 31, 2024; the earliest recorded transaction data in the wholesale market is February 1, 2024, and the latest is October 31, 2024; the market demand forecast given in the industry report starts on March 1, 2024 and ends on February 28, 2025, with a forecast interval of once a month.

[0044] The calculated common start time is March 1, 2024, and the common end time is October 31, 2024.

[0045] Perform spatial dimension alignment, mean value comparison, confidence interval analysis, and dynamic change considerations.

[0046] Based on a comprehensive judgment based on the above results, it is determined that there is an oversupply of apples.

[0047] For growers, it is recommended to plan the planting area for the next season in advance and reduce the planting scale by 10%-15%; optimize the planting structure and appropriately increase the planting proportion of early-maturing varieties with strong market demand; strengthen orchard management, control the amount of fertilizer applied, reasonably arrange irrigation, and reduce production costs.

[0048] In the warehousing stage, warehouse managers are reminded to prepare to expand storage capacity and allocate storage resources in advance to deal with possible apple backlogs; at the same time, based on the characteristics of apples, storage conditions are optimized, such as adjusting temperature and humidity, and using preservation technology to slow down the deterioration of apples.

[0049] On the sales side, collaborate with e-commerce platforms and wholesalers to plan joint promotional activities, use big data for precise marketing, and push preferential information to target consumers; expand emerging sales channels, such as community group buying and live streaming; guide apple processing companies to increase purchasing efforts and develop diversified processed products, such as apple juice, dried apples, apple cider vinegar, etc., to extend the industrial chain and digest excess production.

[0050] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A data resource management method based on artificial intelligence, characterized in that: The method comprises the following steps: Collect multi-source data, including meteorological data, soil data, market demand data and agricultural product data, and perform data cleaning and preprocessing; Select meteorological indicators and soil indicators, calculate the influencing factors of each of the meteorological indicators and soil indicators; determine the weight of each influencing factor, and calculate the influencing factors of the external environment of agricultural products; Collect historical data, match meteorological data and soil data with agricultural product planting areas and planting cycles according to time and space dimensions, and construct a multidimensional data set; perform feature engineering on the multidimensional data set to obtain a feature set, and construct interactive features based on the external environmental influencing factors of the agricultural products and incorporate them into the feature set; Train the long short-term memory network model, and conduct model evaluation and optimization; after training, input meteorological data and soil data into the model, and output agricultural product yield forecast data, including the predicted agricultural product yield value and confidence interval; The agricultural product output forecast data and the market demand data are aligned in time and space dimensions; the average value of the predicted agricultural product output values ​​is compared with the average value of the predicted market demand, the confidence interval is analyzed, and decision-making recommendations are obtained.

2. The data resource management method based on artificial intelligence according to claim 1, characterized in that: Access the meteorological agency network platform through API to obtain meteorological data; use the soil sensor network to monitor soil data; Through web crawler technology, we collect sales data related to agricultural products on various e-commerce platforms every day and perform structured processing; Collect transaction data from the wholesale agricultural product market, including transaction volume, transaction price, and circulation of agricultural products from different origins; obtain forecasted market demand and consumer trend analysis based on agricultural product industry reports and information released by market research institutions; integrate and form market demand data; Obtain corresponding agricultural product data during the planting stage, growth process and after harvesting of agricultural products.

3. The data resource management method based on artificial intelligence according to claim 2, characterized in that: Among the meteorological data, temperature, humidity, rainfall, sunshine duration and wind speed are selected as meteorological indicators; among the soil data, soil moisture, soil fertility, soil pH and soil texture are selected as soil indicators; According to the types of different agricultural products, the meteorological index ranges and soil index ranges suitable for the growth of the agricultural products are collected, it is determined at which stage of the meteorological index range and the soil index range the current meteorological index and the soil index are, and the impact of meteorological index and soil index on the growth of agricultural products is qualitatively graded.

4. The data resource management method based on artificial intelligence according to claim 3 is characterized in that: Obtain meteorological data and soil data related to the current agricultural product planting areas and varieties in the past few years, as well as the corresponding agricultural product yield data and quality data; conduct preliminary statistical analysis on the collected historical data, and calculate the correlation coefficients between various meteorological indicators and soil indicators and agricultural product yield and quality; The impact factor is calculated for each indicator based on the quantitative classification results of the impact of meteorological indicators and soil indicators and the statistical analysis of historical data; the weight of each influencing factor is determined based on the judgment of the degree of each indicator based on the strength of the correlation between each indicator and the output and quality of agricultural products in the statistical analysis of historical data, combined with the current growth stage of agricultural products; the calculated impact factor of each meteorological indicator is multiplied by its corresponding weight, and the impact factor of each soil indicator is multiplied by its corresponding weight, and all the products are added together to obtain the external environment impact factor of agricultural products.

5. The data resource management method based on artificial intelligence according to claim 1, characterized in that: Collect historical meteorological data, soil data and corresponding agricultural product yield data, and match meteorological data and soil data with planting periods in chronological order based on the planting cycle of agricultural products; for annual crops, correspond daily meteorological data and soil data to corresponding growth stages starting from the sowing date, including sowing period, seedling period, growth period, flowering period, fruiting period and harvest period; for perennial crops, match meteorological data and soil data for many years according to their growth rhythm; According to the geographic coordinate information of the agricultural product planting plots, the meteorological data and soil data are located to the corresponding plots; using the geographic information system, the location of each plot is marked on the map to obtain the geographic location data, and the corresponding meteorological data and soil data are associated with the geographic location data; when there is a meteorological station covering several plots, the inverse distance weighted method is used to extrapolate the meteorological data to each plot.

6. The data resource management method based on artificial intelligence according to claim 5, characterized in that: The matched meteorological data, soil data and agricultural product yield data are stored in the form of database tables to construct a multi-dimensional data set; Performing feature engineering on the multidimensional data set, including extracting numerical features, categorical features, time series features and correlation coefficients between the features; calculating the correlation coefficients between the external environmental influencing factors of agricultural products and the extracted different features, constructing interactive features, and incorporating the interactive features into the feature set.

7. The data resource management method based on artificial intelligence according to claim 1, characterized in that: The number of features in the feature set is used as the number of nodes in the input layer of the long short-term memory network, the number of hidden layers and the number of nodes are set, the output layer is configured, the Identity function is selected as the activation function, the Adam optimizer is selected as the optimizer, the feature set is divided into a training set, a validation set and a test set according to the proportion, and the long short-term memory network model is trained; the model is evaluated using the test set to optimize the hyperparameters; After the training is completed, the meteorological data and soil data collected in real time are subjected to feature engineering and input into the trained long short-term memory network model; the model outputs the predicted agricultural product yield value; for the calculation of the confidence interval, the Bootstrap method is used to sample multiple times with replacement from the training set to generate multiple pseudo training sets with the same size as the original training set, and the pseudo training sets are used to retrain multiple long short-term memory network models to obtain a series of yield prediction values, and the confidence interval is calculated by statistically analyzing the distribution of the yield prediction values.

8. The data resource management method based on artificial intelligence according to claim 1, characterized in that: For agricultural product output forecast data, extract the start time and end time ; Collect sales data from e-commerce platforms on a daily basis and record the earliest collection time and the latest collection time ; For the transaction data of the agricultural product wholesale market, determine the earliest transaction record time and the latest trading time ; Obtain forecasts for a period of time in the future from industry reports and market research information, and extract the starting forecast time involved and end prediction time ; Determine the common start time and common end time ; The time resolution of e-commerce platform data is daily. When the yield forecast data is given by week, month or planting season, time aggregation is required. For wholesale market transaction data, and , intercept the data of the corresponding time period, and when its time scale is inconsistent with the output forecast, perform aggregation or split processing; when the industry report data gives an annual forecast, the annual forecast is split into quarterly forecasts and monthly forecasts through linear interpolation method; for the confidence interval of the output forecast, when time aggregation is performed, the upper and lower limits of the confidence interval are also aggregated accordingly; For the yield forecast data, the code set is obtained by coding the administrative divisions ,in, For the 1st to The coding of production forecast data; mapping the purchase area data of the e-commerce platform to the standard geographical area, using the address resolution function of the geographic information system, converting the fuzzy purchase address into the standard administrative area code to form a regional set ,in, For the 1st to The standard administrative region code for purchasing regional data on e-commerce platforms; the different origins in the agricultural product wholesale market transaction data are mapped to standard regional codes, recorded as ,in, For the 1st to Standard administrative region codes for agricultural product wholesale market transaction data; when and , aggregate the sales data of the e-commerce platform and wholesale market corresponding to the same area, and match the production forecast area; and , based on the planting area and historical production proportion, the e-commerce data is divided into sub-areas for production forecast.

9. The data resource management method based on artificial intelligence according to claim 8, characterized in that: Calculate the average of the predicted agricultural production values and the average of the predicted market demand ,Compare and ,like , the probability of oversupply increases by A%; if , the probability of supply exceeding demand increases by A%; obtain the confidence interval, if the lower limit of the confidence interval of agricultural product output is higher than the upper limit of market demand, the probability of oversupply increases by B%; if the upper limit of the confidence interval of agricultural product output is lower than the lower limit of market demand, the probability of supply exceeding demand increases by B%; where A and B are positive integers set by the user; draw the change curve of agricultural product output forecast and market demand in time series, calculate the supply and demand ratio of agricultural product output and market demand; input the supply and demand ratio, the probability of oversupply and the probability of oversupply into the decision suggestion database for search to obtain decision suggestions.

10. An artificial intelligence-based data resource management system, using an artificial intelligence-based data resource management method according to any one of claims 1 to 9, characterized in that: include: Data collection and preprocessing module: including: a data collection unit and a data preprocessing unit; wherein the data collection unit collects multi-source data, the multi-source data includes meteorological data, soil data, market demand data and agricultural product data, and the data preprocessing unit performs data cleaning and preprocessing; Impact factor calculation module: including: indicator impact factor calculation unit and external environment impact factor comprehensive calculation unit; wherein, the indicator impact factor calculation unit selects meteorological indicators and soil indicators, and calculates the impact factor of each indicator in the meteorological indicators and soil indicators; the external environment impact factor comprehensive calculation unit determines the weight of each impact factor, and calculates the external environment impact factor of agricultural products; Dataset construction and feature engineering module: including: a multi-dimensional dataset construction unit and a feature engineering unit; wherein the multi-dimensional dataset construction unit collects historical data, matches meteorological data and soil data with agricultural product planting areas and planting cycles according to time and space dimensions, and constructs a multi-dimensional dataset; the feature engineering unit performs feature engineering on the multi-dimensional dataset to obtain a feature set, constructs interactive features based on the external environmental influencing factors of the agricultural products, and incorporates them into the feature set; Model training and prediction module: including: model construction and training unit, model evaluation and optimization unit and yield prediction execution unit; among them, the model construction and training unit trains the long short-term memory network model, and the model evaluation and optimization unit performs model evaluation and optimization; after the training is completed, the yield prediction execution unit inputs meteorological data and soil data into the model, and outputs agricultural product yield prediction data, including the predicted agricultural product yield value and confidence interval; Data alignment and decision-making module: including: a data alignment unit and a decision suggestion generation unit; wherein the data alignment unit aligns the agricultural product output forecast data with the market demand data in time and space dimensions; the decision suggestion generation unit compares the average value of the predicted agricultural product output value with the average value of the predicted market demand, analyzes the confidence interval, and obtains decision suggestions.

Images