Soil fertility prediction method based on data mining

Abstract

The invention relates to the technical field of data mining, and provides a soil fertility prediction method based on data mining. The method comprises the following steps: acquiring a target soil sample set, and extracting a first soil sample; reading a predetermined fertility characteristic index, and performing fertility detection on the first soil sample to obtain first detection data; calling a porosity plan to carry out porosity detection analysis on the first soil sample to obtain second detection data; inputting the first detection data and the second detection data as an integrated fusion fertility prediction model to obtain output information; and taking a first predicted fertility index mean value in the output information to obtain a target fertility index. According to the method, the technical problems that the fertility prediction precision is relatively low and the method cannot adapt to the change of different soil characteristics due to the fact that multi-source data cannot be fully utilized for comprehensive analysis in a traditional soil fertility prediction method are solved, and comprehensive analysis on various fertility characteristics through an integrated fusion model is realized; and the accuracy and applicability of soil fertility prediction are improved.

Description

Technical Field

[0001] The present application relates to the technical field of data mining, and in particular to a soil fertility prediction method based on data mining. Background Art

[0002] With the continuous development of agricultural production, the monitoring and prediction of soil fertility has become the key to improving soil utilization efficiency and crop yields. Traditional soil fertility assessment methods usually rely on manual sampling and laboratory analysis, which is not only time-consuming but also has certain limitations, and it is difficult to fully and accurately reflect the real-time changes in soil fertility. At the same time, with the increasing complexity of factors such as soil type, environmental factors and fertilization management, traditional methods are difficult to cope with the comprehensive analysis and accurate prediction of multidimensional data, resulting in inaccurate and delayed soil fertility assessment, and unable to provide timely and effective decision-making basis for agricultural production. In addition, many current soil fertility prediction methods lack comprehensive consideration of key characteristics such as soil spatial distribution and porosity, and ignore the interaction of various soil components and their impact on fertility. This makes the fertility prediction results unable to accurately reflect the actual fertility status of the soil, affecting the scientific nature of fertilization strategies, soil improvement measures and crop growth regulation in agricultural production. Summary of the invention

[0003] This application provides a soil fertility prediction method based on data mining, aiming to solve the technical problem that traditional soil fertility prediction methods cannot fully utilize multi-source data for comprehensive analysis, resulting in low fertility prediction accuracy and inability to adapt to changes in different soil characteristics.

[0004] In view of the above problems, the present application provides a soil fertility prediction method based on data mining, which includes: introducing a partitioned and stratified sampling mechanism to obtain a soil sample set of the target soil, and extracting a first soil sample from the soil sample set; reading a predetermined fertility characteristic index, and performing a fertility test on the first soil sample based on the predetermined fertility characteristic index to obtain first test data; retrieving a porosity plan to perform a porosity test and analysis on the first soil sample to obtain second test data; using the first test data and the second test data as input information of an integrated fusion fertility prediction model, and obtaining output information through the integrated fusion fertility prediction model; taking the average of the first predicted fertility index in the output information to obtain a target fertility index, wherein the target fertility index is used to characterize the fertility of the target soil.

[0005] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0006] The above-mentioned soil fertility prediction method based on data mining first obtains a soil sample set of the target soil and extracts a first soil sample from it. This sampling mechanism ensures the representativeness and diversity of the soil samples, which helps to improve the comprehensiveness of the data and the accuracy of the prediction. Subsequently, according to the predetermined fertility characteristic index, the first soil sample is tested for fertility to obtain the first test data. At the same time, the porosity plan is retrieved, and the porosity test analysis is performed on the first soil sample to obtain the second test data. After that, the two sets of data are combined and passed into the integrated fusion fertility prediction model as input information. The model can comprehensively analyze these multidimensional data, thereby improving the accuracy and reliability of soil fertility prediction. Finally, the mean of the first predicted fertility index is extracted through the information output by the model to obtain the target fertility index, which is used to characterize the fertility of the target soil. Through this method, the soil fertility level can be evaluated in real time and accurately, avoiding the errors caused by a single data source or incomplete analysis in traditional methods. At the same time, this integrated fusion prediction model can flexibly adapt to the characteristics of different types of soils, improve the adaptability and practicality of soil fertility prediction, and thus provide a scientific basis for precise fertilization and soil improvement in agricultural production.

[0007] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 The figure is a flow chart of a soil fertility prediction method based on data mining in one embodiment.

[0010] Figure 2 The present invention is a schematic diagram of a process of obtaining second detection data in a soil fertility prediction method based on data mining in an embodiment. DETAILED DESCRIPTION

[0011] The embodiments of the present application provide a soil fertility prediction method based on data mining to solve the technical problem that traditional soil fertility prediction methods cannot fully utilize multi-source data for comprehensive analysis, resulting in low fertility prediction accuracy and inability to adapt to changes in different soil characteristics.

[0012] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0013] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0014] Examples, such as Figure 1 As shown, the present application provides a soil fertility prediction method based on data mining, the method comprising:

[0015] A partitioned and stratified sampling mechanism is introduced to obtain a soil sample set of the target soil, and a first soil sample in the soil sample set is extracted.

[0016] In the embodiment of the present application, a zoning and stratified sampling mechanism is first used to obtain a soil sample set of the target soil. The core of this zoning and stratified sampling mechanism is to divide the target area into several small blocks or layers according to different soil characteristics (such as soil type, terrain, climate, etc.). Each block or layer represents a specific type or condition of soil (such as soil with different soil pH, soil organic matter content or soil with different terrain elevations). This method can ensure that different characteristic areas of the soil are covered during sample collection, thereby obtaining more comprehensive and representative soil samples; after obtaining the soil sample set of the target soil, a soil sample is randomly extracted from the soil sample set of the target soil as the first soil sample. This sample will serve as the basis for preliminary analysis and provide key data for subsequent soil fertility testing; through this zoning and stratified sampling, the deviation that may be caused by a single sampling can be effectively reduced, ensuring that the sample is more representative, and thus providing a more reliable basis for accurately predicting soil fertility.

[0017] A predetermined fertility characteristic index is read, and a fertility test is performed on the first soil sample based on the predetermined fertility characteristic index to obtain first test data.

[0018] In one embodiment, a set of predetermined fertility characteristic indicators are first read, and these indicators include parameters of the chemical composition, physical characteristics, etc. of the soil. For example, the fertility characteristic indicators may include the organic matter content in the soil, the concentration of nitrogen, phosphorus, potassium, soil pH, water content, etc., and these indicators are important bases for judging the soil fertility status; then, based on these predetermined fertility characteristic indicators, a fertility test is performed on the first soil sample. The fertility test process involves a series of experimental methods, such as chemical analysis, spectral analysis or other modern detection technologies. By measuring and analyzing various fertility characteristics in the soil sample, a set of specific test data, namely the first test data, is obtained. This first test data contains specific parameter values ​​corresponding to the fertility characteristic indicators. These parameter values ​​reflect the fertility level of the soil and provide quantitative basic data for subsequent analysis and prediction.

[0019] Furthermore, the present application provides that the predetermined fertility characteristic indicators include fertility characteristic indicators of soil composition dimension and soil characteristic dimension, wherein the soil composition dimension includes various organic matter content indicators and various mineral content indicators, and the soil characteristic dimension includes at least soil pH, soil moisture and soil temperature.

[0020] Preferably, the predetermined fertility characteristic index includes two main dimensions, namely, soil composition dimension and soil characteristic dimension. The soil composition dimension focuses on the chemical composition of the soil, mainly including various organic matter content indicators and various mineral content indicators in the soil. Various organic matter content indicators reflect the richness of organic matter (such as plant residues, humus, etc.) in the soil, and various mineral content indicators include the concentration of important elements (such as nitrogen, phosphorus, potassium, etc.) in the soil. These components directly affect the fertility and nutrient supply capacity of the soil. The soil characteristic dimension focuses on the physical properties of the soil, mainly including Including soil acidity (pH value), soil moisture and soil temperature. Soil acidity affects the plant's ability to absorb nutrients. Too acidic or too alkaline soil may lead to restricted plant growth. Soil moisture is an important factor in soil fertility. The moisture content in the soil affects the growth environment of the plant. Too much or too little moisture will affect the fertility performance. Soil temperature affects the activity of soil microorganisms and the growth of plant roots, thereby indirectly affecting soil fertility. Through these two dimensions of fertility characteristic indicators, the soil fertility status can be comprehensively evaluated, providing a more accurate basis for soil fertility prediction and agricultural management.

[0021] The porosity plan is retrieved to perform porosity detection and analysis on the first soil sample to obtain second detection data.

[0022] In one embodiment, the porosity plan is first retrieved, and a porosity detection and analysis is performed on the first soil sample. Porosity is an indicator of the proportion of voids (such as pores and water pores) in the soil, which reflects the soil's air permeability, water permeability, and water retention capacity. By analyzing the porosity of the soil sample, the soil's performance in terms of water and air flow can be understood, which is crucial for determining soil fertility. When performing porosity detection and analysis, a direct volume detection is performed through the volume detection plan in the porosity plan, and then a liquid penetration detection is performed through the liquid penetration detection plan in the porosity plan. Based on the results of these two tests, important indicators such as the air permeability coefficient and water permeability coefficient of the soil can be calculated. These indicators will be used as the second test data for subsequent soil fertility assessments. Porosity detection provides more detailed physical property data for soil fertility predictions, helping to comprehensively assess the state of the soil.

[0023] Further, if Figure 2 As shown, the present application provides a method for calling a porosity plan to perform porosity detection and analysis on the first soil sample to obtain second detection data, including:

[0024] Extract the volume detection plan in the porosity plan; perform direct volume detection on the first soil sample according to the volume detection plan to obtain a first detection result; extract the liquid penetration detection plan in the porosity plan; perform liquid penetration detection on the first soil sample according to the liquid penetration detection plan to obtain a second detection result; and construct the second detection data based on the first air permeability coefficient obtained by analyzing the first detection result and the first water permeability coefficient and the first water retention coefficient obtained by analyzing the second detection result.

[0025] Preferably, before conducting the porosity test, a volume test plan related to soil volume test is first extracted from the porosity plan. This volume test plan usually includes technical details such as how to measure the volume of the soil and how to calculate the ratio of the pore volume to the solid volume. This plan provides operational guidance for the subsequent direct volume test. Subsequently, direct volume test is performed according to the extracted volume test plan. In this step, the total volume of the soil is obtained by measuring the mass and volume of the first soil sample. Usually, the first soil sample is placed in a container of known volume, its mass is measured, and then the solid volume of the soil is measured by drying. By summarizing these data, the first test result can be obtained; in addition, the plan content related to liquid penetration testing is extracted from the porosity plan. The purpose of liquid penetration testing is to evaluate the water permeability of soil. This process usually involves the penetration rate of liquid (such as water or a specific solution) through the first soil sample. The plan will include the type of liquid, the time period of the penetration test, the penetration path, etc. According to the liquid penetration test plan, the dry first soil sample is tested for permeability. This process usually involves injecting a certain amount of liquid on the soil surface and observing the penetration process of the liquid in the soil. During the test, it is necessary to record the mass of the dry soil sample, the permeability The volume of water that permeates the soil sample, the cross-sectional area of ​​the soil sample, the penetration time, the mass of the soil sample after penetration, etc., are used to obtain the penetration capacity data of the liquid through the soil, thereby obtaining the second test result; after obtaining the results of the volume test and the liquid penetration test, the total volume of the soil in the first test result and the solid volume of the dry soil are calculated by difference to obtain the pore volume, and then the pore volume is calculated by ratio to the total volume of the soil to obtain the first aeration coefficient of the soil, that is, the proportion of space in the soil pores that can be used for gas circulation, and the volume of water that permeates the soil sample in the second test result is calculated by the cross-sectional area and the penetration coefficient of the soil sample. The first water permeability coefficient, that is, the water flow capacity of the soil, is obtained by multiplying the product of the time and calculating the ratio. At the same time, the first water retention coefficient of the soil sample, that is, the ability of the soil to retain water, is obtained by calculating the difference between the mass of the infiltrated soil sample and the mass of the dry soil sample, and then calculating the ratio of the calculated result to the mass of the dry soil sample. Finally, the first aeration coefficient, the first water permeability coefficient and the first water retention coefficient obtained through the analysis are summarized to form the second test data. These data reflect the performance of the soil in air and water flow, are important physical characteristics for evaluating soil fertility, and provide key inputs for subsequent fertility prediction models.

[0026] The first detection data and the second detection data are used as input information of an integrated fusion fertility prediction model, and output information is obtained through the integrated fusion fertility prediction model.

[0027] In one embodiment, the first detection data and the second detection data are firstly input as input information into a pre-built integrated fusion fertility prediction model. Specifically, the first detection data includes soil fertility characteristics (such as organic matter content, mineral content, pH, etc.), while the second detection data reflects the physical properties of the soil (such as porosity, air permeability, water permeability, etc.). These two types of data respectively analyze the fertility status of the soil in detail from different angles, so they need to be integrated. After the two types of data are input into the integrated fusion fertility prediction model, the model will conduct in-depth analysis and integration of various data according to the internal primary prediction layer and secondary prediction layer to generate more accurate fertility prediction results. Through the integrated fusion model, multiple fertility characteristics and soil physical properties can be comprehensively considered to obtain more comprehensive and accurate soil fertility prediction results. Finally, the model outputs the soil fertility prediction results as output information, which is usually expressed as a predicted fertility index, reflecting the fertility level of the target soil. This output information can provide a scientific basis for agricultural decision-making and help select appropriate fertilization strategies, soil improvement methods, etc.

[0028] Further, the present application provides using the first detection data and the second detection data as input information of an integrated fusion fertility prediction model, and obtaining output information through the integrated fusion fertility prediction model, and also includes:

[0029] Obtain a soil fertility detection log, and extract a first fertility detection record in the soil fertility detection log; establish a first training data group based on the first fertility characteristic information and the first fertility index in the first fertility detection record, and obtain a first prediction model by supervised learning on the first training data group; establish a second training data group based on the first porosity characteristic information and the first fertility index in the first fertility detection record, and obtain a second prediction model by supervised learning on the second training data group; the first prediction model and the second prediction model constitute a primary prediction layer; obtain a secondary prediction layer, the secondary prediction layer is used to perform normalized weighted output on the prediction results of the first prediction model and the second prediction model; the primary prediction layer and the secondary prediction layer are constructed as the integrated fusion fertility prediction model.

[0030] Preferably, all test records are first obtained from the stored soil fertility test log, which records the detailed data and results of each soil fertility test; then, a first fertility test record is extracted from the soil fertility test log, which contains fertility characteristic information of a specific soil sample (such as soil organic matter content, mineral composition, pH, etc.) and a corresponding fertility index. According to the extracted first fertility test record, the first fertility characteristic information is combined with the first fertility index group to construct a first training data group. This training data group will be used to train a supervised learning model to predict the soil fertility level, specifically For example, taking a fully connected neural network as an example, other machine learning models can also be used to construct an initial first prediction model through a fully connected neural network, including an input layer, a hidden layer, an output layer, etc., and then use a random initialization method (such as Xavier initialization or He initialization) to assign an initial value to the weight of each neuron to ensure that the initial first prediction model does not produce too large or too small gradients when training begins; then, the first fertility feature information in the first training data group is input layer by layer as an input feature into the initial first prediction model, and after layer-by-layer calculations of the input layer and the hidden layer, the predicted fertility index is output, and the calculation of each layer includes linear weighting of the input and the weight. , add bias and perform nonlinear transformation through activation function, then use mean square error (MSE) as loss function, calculate the error between the predicted fertility index and the first fertility index, and transfer the error from the output layer to the input layer through back propagation algorithm, calculate gradient layer by layer, and determine the contribution of each weight and bias to the loss, and use Adam optimizer (or other optimizers such as SGD) to update the weights in the initial first prediction model. Adam optimizer combines momentum and adaptive learning rate, which can effectively accelerate convergence and avoid overfitting. Perform multiple iterations through the above steps until the loss function of the first training data group converges or reaches the set maximum iteration times, generate the first prediction model. In each round of training, the loss function is minimized by adjusting the weights and biases to improve the accuracy of the prediction. In the first fertility detection record, in addition to the fertility characteristic information, it also contains the first porosity characteristic information (such as physical properties of soil aeration and water permeability). These characteristics also have an important impact on soil fertility. Therefore, based on the porosity characteristic information and the fertility index, a second training data set is constructed. Similarly, the second training data set is also trained by supervised learning. The input feature is the first porosity characteristic information, and the label is the first fertility index. The second prediction model is obtained by training through the same steps as above.Then, the first prediction model (based on fertility characteristics) and the second prediction model (based on porosity characteristics) obtained through training are combined into a first-level prediction layer. The function of this first-level prediction layer is to predict soil fertility according to two different input characteristics respectively, and obtain two independent prediction results. In the first-level prediction layer, each model processes the input data separately to generate prediction results, but does not directly output the final results. Instead, it is handed over to the second-level prediction layer for further processing. In the second-level prediction layer, the two groups of prediction results of the first-level prediction layer are normalized and weighted. Normalized weighting means that the prediction results of the first prediction model and the second prediction model are standardized (such as the maximum and minimum method) and adjusted to the same scale or range for comparison. Different weights are assigned to each model according to the accuracy of each model in the data not used for training, and the determined weights are built into the secondary prediction layer. The secondary prediction layer weights the prediction results of the two models through these two weights to obtain a fused prediction result, namely the first prediction fertility index. This first prediction fertility index is more comprehensive and accurate, and can better reflect the true fertility status of the soil; finally, the primary prediction layer and the secondary prediction layer are integrated to form an integrated fusion fertility prediction model. This integrated model combines the prediction results of fertility characteristics and porosity characteristics, and obtains a more accurate and stable soil fertility prediction through weighted fusion, providing a scientific basis for soil management and fertilizer application in agricultural production. ;

[0031] The average of the first predicted fertility index in the output information is taken to obtain a target fertility index, wherein the target fertility index is used to characterize the fertility of the target soil.

[0032] In one embodiment, after obtaining the output information of each soil sample in the soil sample set, the first predicted fertility index in these output information is averaged to obtain the target fertility index of the target soil. This target fertility index is used to describe the overall fertility status of the target soil, so that the user can understand the overall fertility status of the soil and thus formulate reasonable fertilization plans and improvement measures.

[0033] Further, the present application provides using the first detection data and the second detection data as input information of an integrated fusion fertility prediction model, and obtaining output information through the integrated fusion fertility prediction model, and then further includes:

[0034] Extract the second fertility detection record in the soil fertility detection log; generate a soil fertility time series based on the first fertility index and the second fertility index in the second fertility detection record; activate the prediction support rate function to analyze the soil fertility time series to obtain a first prediction support rate of the first prediction fertility index; when the first prediction support rate does not reach a predetermined support rate threshold, retest and analyze the first prediction fertility index.

[0035] Optionally, first, obtain a second soil fertility test record from the stored soil fertility test log. The second soil fertility test record is data after the first fertility test record, and contains soil fertility data at another moment, including a second fertility index and related fertility characteristics and other data. These data provide a soil fertility reference at another point in time for subsequent analysis; then, generate a soil fertility time series according to the first fertility index and the second fertility index extracted previously in accordance with the chronological relationship. The fertility time series reflects the changing trend of soil fertility at different time points; after obtaining the soil fertility time series, use a prediction support rate function to analyze the time series data. The function of the prediction support rate function is to measure the reliability and accuracy of the model prediction results. Usually, the trend of the time series data is used to calculate the first prediction support rate of the first predicted fertility index. If the first prediction support rate does not reach the predetermined support rate threshold, it means that the model prediction may have a large error and the reliability of the prediction result is low. At this time, in order to improve the accuracy of the prediction, it is necessary to analyze the first prediction support rate of the first predicted fertility index. A predicted fertility index is retested and analyzed. During the retest and analysis process, the input first test data and the second test data will be judged first to identify whether there are erroneous data in the first test data and the second test data, such as missing values ​​and outliers. If there are errors, the fertility test and porosity test will be re-performed, the first test data and the second test data will be updated, and the prediction process will be re-executed. If there are no errors, it means that the hyperparameter settings of the model itself do not conform to the current scenario. At this time, the current first test data and the second test data will be used as retrieval targets, and records with deviations from the first test data and the second test data within the tolerance range (determined according to business needs and expert decisions) will be obtained from the soil fertility test log, and these records will be used to perform secondary training on the current integrated fusion fertility prediction model. The training process is consistent with the above. Through further data collection and model optimization, it is ensured that the predicted fertility index is more accurate and reliable, thereby providing more scientific data support for precision fertilization and soil management in agricultural production.

[0036] Further, the present application provides activating a prediction support rate function to analyze the soil fertility time series to obtain a first prediction support rate of the first prediction fertility index, including:

[0037] Performing trend analysis on the soil fertility time series according to the prediction support rate function to obtain a soil fertility trend line; obtaining a first distance from the first prediction fertility index to the soil fertility trend line; and using the normalized first distance as the first prediction support rate.

[0038] Optionally, when using the prediction support rate function to calculate the support rate of the first predicted fertility index, it is first necessary to perform a trend analysis on the soil fertility time series. Specifically, from the existing soil fertility time series, extract the paired data of time and fertility index to construct a time series data set. These data will serve as the basis for regression analysis to reveal the changing trend of soil fertility; then, select a suitable regression model to fit these data. In the soil fertility time series data, the fertility changes over time usually show certain nonlinear characteristics, which means that polynomial regression can better capture the changing pattern of the fertility index. In order to avoid overfitting and maintain the accuracy of the model, you can start with a quadratic regression model. Quadratic regression can effectively fit fertility data with a single change curve. For example, fertility rises or falls in some periods and remains stable or fluctuates slightly in other periods. Using the quadratic regression model, it can be expressed as y=β0+β1t+β2t 2, where β0 is the intercept, β1 and β2 are regression coefficients, representing the linear part and quadratic part of the fertility change over time, t is the time variable, and y is the soil fertility index; then, the data is fitted by the least squares method to solve the best regression coefficient, so that the fitted trend line can reduce the error between the predicted value and the actual data as much as possible, thereby obtaining a trend line describing the change in soil fertility. After completing the trend line fitting, it is necessary to evaluate the fitting effect, and use the determination coefficient to evaluate the goodness of fit of the model. The closer the determination coefficient is to 1, the better the model fits the data, and the more accurately it can reflect the long-term change trend of fertility. If the determination coefficient value is low, it means that quadratic regression may not be sufficient to fit the data, and you may need to consider increasing the order of the model (such as cubic regression) or trying other nonlinear regression methods. At the same time, you also need to evaluate the residual The residuals are analyzed to check the error distribution of the fitted data. If the residuals show a random distribution, it means that the model fits well. On the contrary, if the residuals show systematic deviations, the model should be adjusted or other methods should be selected. Once it is confirmed that the trend line fits well, the soil fertility trend line can be drawn based on the regression coefficient obtained by fitting. This trend line can not only accurately reflect the change pattern of soil fertility over time, but also provide a reliable basis for subsequent fertility prediction. After obtaining the soil fertility trend line, the absolute difference between the first predicted fertility index and the soil fertility index at that moment in the soil fertility trend line is calculated to obtain the first distance from the first predicted fertility index to the soil fertility trend line. The first distance is then ratio-calculated with the historical maximum distance to complete the normalization of the first distance, thereby obtaining the first prediction support rate, which provides a basis for subsequent retesting or optimization.

[0039] The above process of calculating the first predicted support rate is the specific calculation process of the predicted support rate function. Therefore, the specific expression of the predicted support rate function is: Among them, S is the normalized first prediction support rate, is the first predicted fertility index predicted by the model, y trend is the value of the soil fertility trend line at the corresponding time point, D max is the maximum distance between the forecast value and the trend line in the historical data.

[0040] Further, the present application provides taking the mean of the first predicted fertility index in the output information to obtain a target fertility index, including:

[0041] Reading a predetermined human impact index; analyzing human activity records based on the predetermined human impact index to obtain human impact index parameters; performing normalized weighted analysis on the human impact index parameters to obtain a human impact coefficient; and calibrating and adjusting the first predicted fertility index based on the human impact coefficient.

[0042] Preferably, in soil fertility prediction, it is crucial to consider the influence of human factors, because soil fertility is not only affected by natural factors, but also significantly affected by human activities (such as tillage, fertilization, irrigation, etc.). Therefore, it is necessary to calibrate the first predicted fertility index according to the human impact index. Specifically, first obtain predetermined human impact indicators, which usually include various human activity data related to soil fertility changes, such as crop rotation, fertilization records, tillage, pollution, etc.; then, obtain human activity records through survey records, agricultural management systems or other data sources, and quantify the influence of these human impact indicators in combination with the knowledge of field experts and existing agricultural research literature to obtain human impact indicator parameters; then, since the impact intensity of different human activities on soil fertility varies greatly, it is necessary to normalize the extracted human impact indicator parameters (such as the maximum value). Minimum method), the purpose of normalization is to convert data of different units and dimensions into the same scale (usually between 0 and 1) for easy comparison and weighting; after normalization, each indicator needs to be weighted. Different human impact indicators may have different impact intensities on soil fertility, and these differences need to be reflected through weighting. For example, the impact of fertilization on fertility may be stronger than that of tillage, so the weight of fertilization can be set larger. The weight of each human impact indicator is also determined based on business needs and expert decisions. By weighting and summing the obtained human impact indicator parameters through these weights, an artificial impact coefficient can be obtained. By adding this artificial impact coefficient to the first predicted fertility index, the calibration adjustment of the first predicted fertility index is completed. The adjusted first predicted fertility index combines the influence of natural factors and human factors, and provides a more accurate and comprehensive fertility prediction.

[0043] Furthermore, the present application provides that the predetermined human influence indicators include influence indicators of crop rotation dimension, fertilization management dimension and social activity dimension, wherein the crop rotation dimension refers to a rotation record with a crop type identifier, the fertilization management dimension refers to a fertilization record with a fertilization type and usage identifier, and the social activity dimension includes at least tillage, desertification and pollution.

[0044] Optionally, the predetermined human impact indicators involve multiple aspects, mainly including dimensions such as crop rotation, fertilizer management and social activities, which directly or indirectly affect the soil fertility level. The crop rotation dimension refers to records related to crop rotation. Crop rotation refers to planting different crops in a specific order on the same land. Crop rotation can effectively improve the fertility and health of the soil and avoid excessive consumption of soil nutrients. The crop rotation dimension usually includes the identification of crop types and the planting order of crops in each cycle, which helps analyze the long-term impact of planting different crops on soil fertility. For example, some crops (such as beans) can fix nitrogen and help increase the nitrogen content in the soil, while some crops Crops (such as grains) may consume more nitrogen. Therefore, crop rotation records are crucial to understanding changes in soil fertility. The fertilization management dimension covers detailed records of fertilization, including information such as the type, amount, and method of fertilization. Fertilization management is an important factor affecting soil fertility. Reasonable fertilization can supplement the nutrients lacking in the soil and enhance soil fertility. The fertilization management dimension records the type of fertilizer (such as organic fertilizer, chemical fertilizer, compound fertilizer, etc.), the amount of fertilizer (usually measured in kilograms / hectare or tons / hectare) and the method of fertilization (such as broadcasting, strip application, drip irrigation, etc.) used each time fertilization is applied. Different fertilization methods and amounts have different effects on soil fertility. Reasonable fertilization Fertilizer can maintain or improve soil fertility, while excessive fertilization may lead to nutrient imbalance or soil pollution. The social activity dimension includes the direct or indirect impact of human activities on soil fertility, especially activities that may cause soil degradation or pollution during agricultural production, mainly including tillage, desertification, and pollution. Tillage refers to the tillage methods adopted in the process of cultivating land, such as deep tillage, shallow tillage, and rotational tillage. Different tillage methods will affect the structure, air permeability, and water retention capacity of the soil. For example, excessive tillage may lead to soil compaction, affect the water permeability of the soil, and reduce the accumulation of organic matter, thereby reducing soil fertility. Desertification refers to the deterioration of soil due to factors such as overgrazing, over-tillage, or vegetation destruction. The process in which the soil surface loses its soil-fixing ability, leading to wind erosion or water erosion, and forming sandy land. Desertification will aggravate the decline of soil fertility, especially in arid areas, affecting the cultivability of land and agricultural productivity. Pollution is usually caused by chemical substances (such as pesticides, heavy metals, excessive fertilizers, etc.), which will reduce the ecological function of the soil, destroy the soil structure, affect the activity of microorganisms, and thus affect soil fertility. Industrial emissions in social activities, improper garbage disposal, agricultural pollution, etc. may all lead to soil pollution. These human-influenced indicators conduct a comprehensive assessment of soil fertility from different angles, help to more comprehensively understand the changing laws of soil fertility, and provide a basis for precise soil management and fertilization decisions.

[0045] In summary, the embodiments of the present application have at least the following technical effects:

[0046] The embodiment of the present application aims to accurately predict soil fertility by integrating the fertility characteristics, porosity characteristics and human activity influences of soil samples. The present application includes introducing a partitioned and stratified sampling mechanism, obtaining a sample set of the target soil and conducting fertility detection, conducting porosity detection and analysis on the soil samples in combination with the porosity plan, and obtaining prediction results by integrating the fertility prediction model and taking the fertility detection data and the porosity data as inputs; further analyzing the soil fertility time series data, activating the prediction support rate function, and conducting trend analysis to obtain the prediction support rate of soil fertility, and optimizing the model based on this; to ensure the accuracy of the prediction results, human factors such as crop rotation, fertilization management and social activities are also considered, and the human influence coefficient is obtained by normalizing and weighting the parameters of these human factors, and the predicted fertility index is calibrated and adjusted, thereby providing a more accurate soil fertility prediction. These technical effects jointly solve the technical problem that traditional soil fertility prediction methods cannot fully utilize multi-source data for comprehensive analysis, resulting in low fertility prediction accuracy and inability to adapt to changes in different soil characteristics. They achieve the technical effect of comprehensively analyzing multiple fertility characteristics through integrated fusion models, thereby improving the accuracy and applicability of soil fertility prediction.

[0047] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0048] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0049] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A soil fertility prediction method based on data mining, characterized in that: include: Introducing a partitioned and stratified sampling mechanism to obtain a soil sample set of the target soil, and extracting a first soil sample from the soil sample set; Reading a predetermined fertility characteristic index, and performing a fertility test on the first soil sample based on the predetermined fertility characteristic index to obtain first test data; Retrieving the porosity plan to perform porosity detection and analysis on the first soil sample to obtain second detection data; Using the first detection data and the second detection data as input information of an integrated fusion fertility prediction model, and obtaining output information through the integrated fusion fertility prediction model; The average of the first predicted fertility index in the output information is taken to obtain a target fertility index, wherein the target fertility index is used to characterize the fertility of the target soil.

2. The soil fertility prediction method based on data mining according to claim 1 is characterized in that: The predetermined fertility characteristic index includes fertility characteristic indexes of soil composition dimension and soil characteristic dimension, wherein the soil composition dimension includes various organic matter content indexes and various mineral content indexes, and the soil characteristic dimension includes at least soil pH, soil moisture and soil temperature.

3. The soil fertility prediction method based on data mining according to claim 1 is characterized in that: Retrieving the porosity plan to perform porosity detection and analysis on the first soil sample to obtain second detection data, including: Extracting a volume detection plan from the porosity plan; Performing a direct volume test on the first soil sample according to the volume test plan to obtain a first test result; Extracting a liquid penetration testing plan from the porosity plan; Performing a liquid penetration test on the first soil sample according to the liquid penetration test plan to obtain a second test result; The second detection data is constructed based on the first ventilation coefficient obtained by analyzing the first detection result and the first water permeability coefficient and the first water retention coefficient obtained by analyzing the second detection result.

4. The soil fertility prediction method based on data mining according to claim 1 is characterized in that: The first detection data and the second detection data are used as input information of an integrated fusion fertility prediction model, and output information is obtained through the integrated fusion fertility prediction model, and the above also includes: Obtaining a soil fertility detection log, and extracting a first fertility detection record in the soil fertility detection log; A first prediction model is obtained by forming a first training data group based on the first fertility feature information and the first fertility index in the first fertility detection record and performing supervised learning on the first training data group; A second prediction model is obtained by forming a second training data set based on the first porosity characteristic information in the first fertility detection record and the first fertility index, and performing supervised learning on the second training data set; The first prediction model and the second prediction model form a primary prediction layer; Acquire a secondary prediction layer, where the secondary prediction layer is used to perform normalized weighted output on the prediction results of the first prediction model and the second prediction model; The primary prediction layer and the secondary prediction layer are constructed as the integrated fusion fertility prediction model.

5. The soil fertility prediction method based on data mining according to claim 4 is characterized in that: The first detection data and the second detection data are used as input information of an integrated fusion fertility prediction model, and output information is obtained through the integrated fusion fertility prediction model, and then the following further includes: Extracting a second fertility test record from the soil fertility test log; generating a soil fertility time series based on the first fertility index and the second fertility index in the second fertility detection record; activating a prediction support rate function to analyze the soil fertility time series to obtain a first prediction support rate of the first prediction fertility index; When the first prediction support rate does not reach a predetermined support rate threshold, the first predicted fertility index is retested and analyzed.

6. The soil fertility prediction method based on data mining according to claim 5 is characterized in that: Activating the prediction support rate function to analyze the soil fertility time series to obtain a first prediction support rate of the first prediction fertility index includes: Performing trend analysis on the soil fertility time series according to the prediction support rate function to obtain a soil fertility trend line; Obtaining a first distance from the first predicted fertility index to the soil fertility trend line; The normalized first distance is used as the first prediction support rate.

7. The soil fertility prediction method based on data mining according to claim 1 is characterized in that: Taking the mean of the first predicted fertility index in the output information to obtain the target fertility index includes: Reading predetermined human impact indicators; Analyze the human activity records based on the predetermined human impact index to obtain human impact index parameters; Performing normalized weighted analysis on the human impact index parameters to obtain a human impact coefficient; The first predicted fertility index is calibrated and adjusted based on the human influence coefficient.

8. The soil fertility prediction method based on data mining according to claim 7 is characterized in that: The predetermined human impact indicators include impact indicators in the crop rotation dimension, the fertilization management dimension and the social activity dimension, wherein the crop rotation dimension refers to the rotation record with the crop type identification, the fertilization management dimension refers to the fertilization record with the fertilization type and usage identification, and the social activity dimension at least includes tillage, desertification and pollution.

Images