Cultivated land quality evaluation method and system based on random forest

Abstract

The invention discloses a farmland quality evaluation method and system based on a random forest, and the method comprises the steps: S1, obtaining the basic condition of an existing farmland, and obtaining the farmland information; s2, based on the cultivated land information, classifying the cultivated land information by using an ensemble learning classification algorithm to obtain a classification result; s3, constructing a cultivated land quality evaluation model based on a random forest algorithm, selecting a training set based on a classification result, and inputting the training set into the cultivated land quality evaluation model for training to obtain a trained cultivated land quality evaluation model; and S4, collecting the basic condition of the cultivated land to be measured, obtaining the information of the cultivated land to be measured, inputting the information of the cultivated land to be measured into the cultivated land quality evaluation model for evaluation, and obtaining a quality evaluation result. The cultivated land information is preprocessed and classified through the integrated learning classification algorithm, different features and conditions of cultivated land are distinguished more finely, more accurate input is provided for follow-up construction of a cultivated land quality evaluation model, and therefore the accuracy and pertinence of overall evaluation are improved.

Description

[0001] Technology Neighborhood

[0002] The present invention relates to the field of cultivated land investigation and evaluation technology, and in particular to a cultivated land quality evaluation method and system based on random forest. Background Art

[0003] As a key national resource, cultivated land is directly related to the security of food supply, the prosperity of rural economy and the balance of ecological environment. Its quality assessment is a yardstick to measure the potential and current status of agricultural development in a region. This work is extremely critical. It not only helps the government and agricultural departments to clearly understand the actual situation of cultivated land, but also guides how to improve the quality of cultivated land, optimize the agricultural production structure, ensure food self-sufficiency, and maintain food quality and ecological security. It is the cornerstone of the sustainable development strategy. Traditionally, cultivated land quality assessment focuses on the physical and chemical properties of the soil itself, such as fertility level, but this single perspective ignores factors such as transportation convenience and reasonable layout of cultivated land that also have a significant impact on agricultural production efficiency. Therefore, the assessment system is not comprehensive enough and it is difficult to accurately reflect the overall picture of cultivated land quality. Given the multidimensional characteristics of cultivated land quality, the current challenge is the lack of a universally recognized standard to integrate various indicators and make fair comparisons across regions. This means that even if we recognize that the evaluation should cover a wider range of dimensions, it is still difficult to achieve direct quality comparisons between different regions in actual operations, which affects the practicality and guiding value of the assessment results. Therefore, it is imperative to improve the arable land quality evaluation system and explore a comprehensive evaluation model that can comprehensively consider multiple factors such as soil characteristics, terrain conditions, water resources, transportation accessibility, land use pattern, etc., and establish a standardized, quantifiable and comparable evaluation framework. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method for evaluating farmland quality based on random forests, the method comprising:

[0005] Step S1, obtaining basic information of existing cultivated land and obtaining cultivated land information;

[0006] Step S2: Based on the cultivated land information, the cultivated land information is classified using an integrated learning classification algorithm to obtain a classification result;

[0007] Step S3, constructing a cultivated land quality evaluation model based on a random forest algorithm, selecting a training set based on the classification result and inputting it into the cultivated land quality evaluation model for training, and obtaining a trained cultivated land quality evaluation model;

[0008] Step S4: Collect basic information of the cultivated land to be tested to obtain information of the cultivated land to be tested, input the information of the cultivated land to be tested into the cultivated land quality evaluation model for evaluation, and obtain a quality evaluation result.

[0009] Optionally, in step S1, the cultivated land information mainly includes cultivated land soil color, cultivated land soil layer thickness, cultivated land soil texture, cultivated land soil organic matter content, cultivated land inorganic pollution, cultivated land altitude, cultivated land irrigation guarantee rate and cultivated land environment conditions.

[0010] Optionally, in S2, based on the cultivated land information, the cultivated land information is classified using an integrated learning classification algorithm, and the process of obtaining the classification result specifically includes:

[0011] Taking the cultivated land information as an original data set, granulating the original data set based on null value information of the original data set to obtain a missing attribute set;

[0012] Based on the missing attribute set, a maximized information granule is obtained, and neighborhood tolerance classes are divided, and tolerance information entropy is calculated based on the neighborhood tolerance classes;

[0013] Calculate neighborhood tolerance conditional entropy based on the missing attribute set and the tolerance information entropy;

[0014] Based on the maximized information granules, a non-missing attribute set is used as input, and the BP algorithm is used as a base classifier for ensemble learning to obtain several classifiers;

[0015] Calculate the weight of the sub-classifier using the neighborhood tolerance condition entropy, the size of the information granule, and the sub-classifier accuracy;

[0016] Based on the sub-classifiers and the weights of the sub-classifiers, classifier integration is performed to obtain a classification result of cultivated land information.

[0017] Optionally, the weight calculation method of the sub-classifier includes:

[0018] ω i =(Acc i |Gra i |NTE i ) / (∑Acc i |Gra i |NTE i )

[0019] Among them, ω i The weight assigned to the prediction of the i-th base classifier, Acc i is the accuracy of the i-th base classifier, |Gra i | is the size of the i-th information particle, NTE i is the neighborhood tolerance conditional entropy of the category attribute corresponding to the missing attribute set of the i-th information particle.

[0020] Optionally, in step S3, the process of constructing a farmland quality evaluation model based on the random forest algorithm specifically includes:

[0021] The Gini coefficient is used to characterize the data characteristics after classification, and the binary cabinet partitioning technology is used to generate a simple binary tree as the influencing factor to calculate the Gini coefficient value based on the influencing factor;

[0022] Factor screening is performed based on Pearson correlation analysis and the Gini coefficient value to generate a random forest;

[0023] Genetic algorithm is used to adaptively optimize the random forest parameters to obtain a constructed arable land quality evaluation model.

[0024] The present invention also discloses a farmland quality evaluation system based on random forests, the system comprising: a historical information acquisition module, an information classification module, an evaluation model construction module and a farmland evaluation result module;

[0025] The historical information acquisition module is used to obtain the basic situation of existing cultivated land and obtain cultivated land information;

[0026] The information classification module is used to classify the cultivated land information based on the cultivated land information using an integrated learning classification algorithm to obtain a classification result;

[0027] The evaluation model construction module is used to construct a cultivated land quality evaluation model based on a random forest algorithm, select a training set based on the classification result and input it into the cultivated land quality evaluation model for training, so as to obtain a trained cultivated land quality evaluation model;

[0028] The farmland evaluation result module is used to collect basic information of the farmland to be tested, obtain information of the farmland to be tested, input the information of the farmland to be tested into the farmland quality evaluation model for evaluation, and obtain a quality evaluation result.

[0029] Optionally, the cultivated land information mainly includes cultivated land soil color, cultivated land soil layer thickness, cultivated land soil texture, cultivated land soil organic matter content, cultivated land inorganic pollution, cultivated land altitude, cultivated land irrigation guarantee rate and cultivated land environment conditions.

[0030] Optionally, the workflow of the information classification module includes:

[0031] Taking the cultivated land information as an original data set, granulating the original data set based on null value information of the original data set to obtain a missing attribute set;

[0032] Based on the missing attribute set, a maximized information granule is obtained, and neighborhood tolerance classes are divided, and tolerance information entropy is calculated based on the neighborhood tolerance classes;

[0033] Calculate neighborhood tolerance conditional entropy based on the missing attribute set and the tolerance information entropy;

[0034] Based on the maximized information granules, a non-missing attribute set is used as input, and the BP algorithm is used as a base classifier for ensemble learning to obtain several classifiers;

[0035] Calculate the weight of the sub-classifier using the neighborhood tolerance condition entropy, the size of the information granule, and the sub-classifier accuracy;

[0036] Based on the sub-classifiers and the weights of the sub-classifiers, classifier integration is performed to obtain a classification result of cultivated land information.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The present invention pre-processes and classifies cultivated land information through an integrated learning classification algorithm, which can more finely distinguish the different characteristics and conditions of cultivated land, and provide more accurate input for the subsequent construction of a cultivated land quality evaluation model, thereby improving the accuracy and pertinence of the overall evaluation. This method can help agricultural decision makers identify high-quality and low-quality cultivated land areas by quickly and accurately evaluating the quality of the cultivated land to be tested, and optimize resource allocation accordingly, such as strengthening protection measures in high-yield fields and implementing improvement plans in low-yield fields, thereby improving agricultural production efficiency and land use efficiency. Accurate evaluation of cultivated land quality helps to formulate scientific and reasonable farming systems and protection measures, reduce over-exploitation and unreasonable use, protect cultivated land resources, support the sustainable development of agriculture, and ensure long-term food security and ecological and environmental balance. High-quality cultivated land evaluation results can provide a scientific basis for the government to formulate agricultural policies, subsidy issuance, land transfer policies, etc., help precise policy implementation, and promote the process of agricultural modernization. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in 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 paying creative labor.

[0040] Figure 1 This is a method step diagram of a method for evaluating cultivated land quality according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0042] Embodiment 1

[0043] Random forest-based methods for evaluating farmland quality, such as Figure 1 As shown, the method includes:

[0044] Step S1, obtain the basic situation of existing cultivated land and obtain cultivated land information. In this embodiment, the cultivated land information mainly includes cultivated land soil color, cultivated land soil thickness, cultivated land soil texture, cultivated land soil organic matter content, cultivated land inorganic pollution, cultivated land altitude, cultivated land irrigation guarantee rate and cultivated land environment. In the present invention, historical cultivated land data is selected, but because the cultivated land data collection period is long, some data are small in number or even absent during the collection process, so it is necessary to process the data reasonably to avoid overfitting in the future.

[0045] Step S2, based on the cultivated land information, using an integrated learning classification algorithm to classify the cultivated land information to obtain a classification result. Specifically, the method includes: taking the cultivated land information as the original data set, granulating the original data set based on the null value information of the original data set to obtain a missing attribute set; obtaining a maximized information granule based on the missing attribute set, dividing the neighborhood tolerance class, and calculating the tolerance information entropy based on the neighborhood tolerance class; calculating the neighborhood tolerance conditional entropy based on the missing attribute set and the tolerance information entropy; based on the maximized information granule, taking the non-missing attribute set as input, using the BP algorithm as the base classifier for integrated learning to obtain several classifiers; using the neighborhood tolerance conditional entropy, the size of the information granule and the sub-classifier accuracy to calculate the weight of the sub-classifier; integrating the classifiers based on the sub-classifiers and the weights of the sub-classifiers to obtain the cultivated land information classification result.

[0046] Let x 1 ,x 2 ,...,x n Cultivated land information 4 , c 1 ,c 2 ,c 3 ,c 4 is the conditional attribute, c 5 To make a decision attribute, the samples are granulated according to the missing attributes in the incomplete mixed data set. Several information granules are obtained. In order to improve the accuracy of the prediction and effectively use the data with missing values, we adopt a strategy to expand the coverage of information granules. This process starts with re-examining the original data set and distinguishing two types of information granules: one is a data set without missing attributes, and the other is a set with missing attributes. Then, if it is found that a certain information granule with missing attributes has a set of attributes that can be completely covered by the attribute set of another information granule, we will do the following: set the missing attribute value of the sample in this information granule with missing attributes according to the missing attribute set of the larger information granule that contains it. In this way, we not only retain the original missing information, but also make reasonable inferences about the missing data based on a broader context, forming a "maximized information granule". There are four cases for maximizing information granules:

[0047] The first type: does not contain missing attributes: X 1 ={x 1 ,x 4 ,x 5};

[0048] The second type: missing attribute c 4 , remove attribute c from samples that do not contain missing attributes 4 , then we get X 2 ={x 1 ,x 2 ,x 3 ,x 5};

[0049] The third type: missing attribute c 2 、c 3 , remove attribute c from samples that do not contain missing attributes 2 、c 3 , then we get X 3 ={x 1 ,x 3 ,x 4 ,x 5};

[0050] The fourth type: missing attribute c 3 、c 4 , remove attribute c from samples without missing attributes 3 、c 4 , then we get X 4 ={x 1 ,x 2 ,x 4 ,x 5 ,x 6}.

[0051] First, the neighborhood tolerance classes are divided according to the definition, and the neighborhood tolerance information entropy is calculated according to the following formula:

[0052]

[0053] Calculate the neighborhood tolerance joint entropy:

[0054]

[0055] Finally, the missing attributes, including continuous attributes and discrete attributes, are taken as known conditions to calculate the neighborhood tolerance conditional entropy based on category attributes:

[0056]

[0057] On each maximized information granule, non-missing attributes are used as input, and ensemble learning is performed using an integrated classification algorithm with the BP algorithm as the base classifier to obtain several classification prediction models. The weights of each sub-classifier are calculated using the neighborhood tolerance conditional entropy corresponding to the missing attributes of each information granule, the size of the information granule, and the accuracy of the sub-classifier. The missing attribute set of the prediction data set is a subset of the missing attribute set of some information granules. Then, the sample of the missing attribute set and the attribute set corresponding to these information granules are used as the input of the corresponding sub-classifier. After training, the predicted category of the sample on these sub-classifiers is obtained, and then according to the analysis results of these base classifiers, the weight calculation method of the sub-classifier includes:

[0058] ω i =(Acc i |Gra i |NTE i ) / (∑Acc i |Gra i |NTE i )

[0059] Among them, ω i The weight assigned to the prediction of the i-th base classifier, Acc i is the accuracy of the i-th base classifier, |Gra i | is the size of the i-th information particle, NTE i is the neighborhood tolerance conditional entropy of the category attribute corresponding to the missing attribute set of the i-th information particle.

[0060] Step S3, constructing a cultivated land quality evaluation model based on a random forest algorithm, selecting a training set based on the classification result and inputting it into the cultivated land quality evaluation model for training, and obtaining a trained cultivated land quality evaluation model. Specifically, it includes: using the Gini coefficient to characterize the data characteristics after classification, using the binary cabinet segmentation technology to generate a simple binary tree as an influencing factor to calculate the Gini coefficient value based on the influencing factor; screening factors based on Pearson correlation analysis and the Gini coefficient value to generate a random forest; using a genetic algorithm to adaptively optimize the random forest parameters to obtain a constructed cultivated land quality evaluation model.

[0061] Step S4: Collect basic information of the cultivated land to be tested to obtain information of the cultivated land to be tested, input the information of the cultivated land to be tested into the cultivated land quality evaluation model for evaluation, and obtain a quality evaluation result.

[0062] Embodiment 2

[0063] The farmland quality evaluation system based on random forest includes: historical information acquisition module, information classification module, evaluation model construction module and farmland evaluation result module;

[0064] The historical information acquisition module is used to obtain the basic situation of existing cultivated land and obtain cultivated land information. The cultivated land information mainly includes cultivated land soil color, cultivated land soil layer thickness, cultivated land soil texture, cultivated land soil organic matter content, cultivated land inorganic pollution, cultivated land altitude, cultivated land irrigation guarantee rate and cultivated land environment.

[0065] The information classification module is used to classify the cultivated land information based on the cultivated land information using an integrated learning classification algorithm to obtain a classification result.

[0066] The workflow of the information classification module includes: taking the cultivated land information as the original data set, granulating the original data set based on the null value information of the original data set to obtain a missing attribute set; obtaining a maximized information granule based on the missing attribute set, dividing the neighborhood tolerance class, and calculating the tolerance information entropy based on the neighborhood tolerance class; calculating the neighborhood tolerance conditional entropy based on the missing attribute set and the tolerance information entropy; based on the maximized information granule, taking the non-missing attribute set as input, using the BP algorithm as the base classifier for ensemble learning to obtain several classifiers; using the neighborhood tolerance conditional entropy, the size of the information granule and the sub-classifier accuracy to calculate the weight of the sub-classifier; integrating the classifiers based on the sub-classifiers and the weights of the sub-classifiers to obtain the cultivated land information classification results.

[0067] The evaluation model construction module is used to construct a cultivated land quality evaluation model based on a random forest algorithm, select a training set based on the classification result and input it into the cultivated land quality evaluation model for training, so as to obtain a trained cultivated land quality evaluation model.

[0068] Specifically, it includes: using the Gini coefficient to characterize the characteristics of classified data, using binary floor cabinet segmentation technology to generate a simple binary tree as an influencing factor to calculate the Gini coefficient value based on the influencing factor; screening factors based on Pearson correlation analysis and the Gini coefficient value to generate a random forest; using a genetic algorithm to adaptively optimize the random forest parameters to obtain a constructed arable land quality evaluation model.

[0069] The farmland evaluation result module is used to collect basic information of the farmland to be tested, obtain information of the farmland to be tested, input the information of the farmland to be tested into the farmland quality evaluation model for evaluation, and obtain a quality evaluation result.

[0070] The embodiments described above are only descriptions of the preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the protection scope of the claims of the present invention.

Claims

1. The method for evaluating cultivated land quality based on random forest is characterized by: The method comprises: Step S1, obtaining basic information of existing cultivated land and obtaining cultivated land information; Step S2: Based on the cultivated land information, the cultivated land information is classified using an integrated learning classification algorithm to obtain a classification result; Step S3, constructing a cultivated land quality evaluation model based on a random forest algorithm, selecting a training set based on the classification result and inputting it into the cultivated land quality evaluation model for training, and obtaining a trained cultivated land quality evaluation model; Step S4: Collect basic information of the cultivated land to be tested to obtain information of the cultivated land to be tested, input the information of the cultivated land to be tested into the cultivated land quality evaluation model for evaluation, and obtain a quality evaluation result.

2. The method for evaluating cultivated land quality based on random forest according to claim 1, characterized in that: In step S1, the cultivated land information mainly includes cultivated land soil color, cultivated land soil layer thickness, cultivated land soil texture, cultivated land soil organic matter content, cultivated land inorganic pollution, cultivated land altitude, cultivated land irrigation guarantee rate and cultivated land environment.

3. The method for evaluating cultivated land quality based on random forest according to claim 1, characterized in that: In S2, based on the cultivated land information, the cultivated land information is classified using an integrated learning classification algorithm, and the process of obtaining the classification result specifically includes: Taking the cultivated land information as an original data set, granulating the original data set based on null value information of the original data set to obtain a missing attribute set; Based on the missing attribute set, a maximized information granule is obtained, and neighborhood tolerance classes are divided, and tolerance information entropy is calculated based on the neighborhood tolerance classes; Calculate neighborhood tolerance conditional entropy based on the missing attribute set and the tolerance information entropy; Based on the maximized information granules, a non-missing attribute set is used as input, and the BP algorithm is used as a base classifier for ensemble learning to obtain several classifiers; Calculate the weight of the sub-classifier using the neighborhood tolerance condition entropy, the size of the information granule, and the sub-classifier accuracy; Based on the sub-classifiers and the weights of the sub-classifiers, classifier integration is performed to obtain a classification result of cultivated land information.

4. The method for evaluating cultivated land quality based on random forest according to claim 3, characterized in that: The weight calculation method of the sub-classifier includes: oh i =(Acc i |Gra i |NTE i ) / (∑Acc i |Gra i |NTE i ) Among them, ω i The weight assigned to the prediction of the i-th base classifier, Acc i is the accuracy of the i-th base classifier, |Gra i | is the size of the i-th information particle, NTE i is the neighborhood tolerance conditional entropy of the category attribute corresponding to the missing attribute set of the i-th information particle.

5. The method for evaluating cultivated land quality based on random forest according to claim 1, characterized in that: In step S3, the process of constructing a farmland quality evaluation model based on the random forest algorithm specifically includes: The Gini coefficient is used to characterize the data characteristics after classification, and the binary cabinet partitioning technology is used to generate a simple binary tree as the influencing factor to calculate the Gini coefficient value based on the influencing factor; Factor screening is performed based on Pearson correlation analysis and the Gini coefficient value to generate a random forest; Genetic algorithm is used to adaptively optimize the random forest parameters to obtain a constructed arable land quality evaluation model.

6. A farmland quality evaluation system based on random forest, wherein the farmland quality evaluation system is used to implement the farmland quality evaluation method according to any one of claims 1 to 5, characterized in that: The system includes: historical information acquisition module, information classification module, evaluation model construction module and cultivated land evaluation result module; The historical information acquisition module is used to obtain the basic situation of existing cultivated land and obtain cultivated land information; The information classification module is used to classify the cultivated land information based on the cultivated land information using an integrated learning classification algorithm to obtain a classification result; The evaluation model construction module is used to construct a cultivated land quality evaluation model based on a random forest algorithm, select a training set based on the classification result and input it into the cultivated land quality evaluation model for training, so as to obtain a trained cultivated land quality evaluation model; The farmland evaluation result module is used to collect basic information of the farmland to be tested, obtain information of the farmland to be tested, input the information of the farmland to be tested into the farmland quality evaluation model for evaluation, and obtain a quality evaluation result.

7. The farmland quality evaluation system based on random forest according to claim 6 is characterized in that: The cultivated land information mainly includes cultivated land soil color, cultivated land soil layer thickness, cultivated land soil texture, cultivated land soil organic matter content, cultivated land inorganic pollution, cultivated land altitude, cultivated land irrigation guarantee rate and cultivated land environment conditions.

8. The farmland quality evaluation system based on random forest according to claim 6 is characterized in that: The workflow of the information classification module includes: Taking the cultivated land information as an original data set, granulating the original data set based on null value information of the original data set to obtain a missing attribute set; Based on the missing attribute set, a maximized information granule is obtained, and neighborhood tolerance classes are divided, and tolerance information entropy is calculated based on the neighborhood tolerance classes; Calculate neighborhood tolerance conditional entropy based on the missing attribute set and the tolerance information entropy; Based on the maximized information granules, a non-missing attribute set is used as input, and the BP algorithm is used as a base classifier for ensemble learning to obtain several classifiers; Calculate the weight of the sub-classifier using the neighborhood tolerance condition entropy, the size of the information granule, and the sub-classifier accuracy; Based on the sub-classifiers and the weights of the sub-classifiers, classifier integration is performed to obtain a classification result of cultivated land information.

Images