Seed production field prediction processing method, device, storage medium and electronic device

Abstract

The embodiments of the present application provide a seed field prediction processing method, device, storage medium and electronic device, the method comprising: determining the feature weight of the signaling position data of the field to be identified; obtaining the target feature variable related to the seed field identification according to the feature weight of the signaling position data to be identified; inputting the target feature variable into a pre-trained target recognition model to perform seed field prediction and obtain the prediction result of the field to be identified. This can solve the problem in the related art that seed fields are identified through remote sensing images, that manual identification is required based on the stripe image presented by the seed corn field which is different from the field corn, and that a large number of manual efforts are required for image recognition, and that non-striped sowing cannot be identified. This method can be applied to large-scale and accurate seed field identification, and the identification of seed fields can be completed without manual participation, thereby achieving the effect of reducing labor costs.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of communications, and more specifically, to a seed field prediction processing method, device, storage medium and electronic device. Background Art

[0002] Seed management is the top priority of the Central Document No. 1 in 2021, and is also one of the important tasks of national ministries and local towns. At present, in the large-scale corn planting process, it is often divided into field corn and seed corn. Field corn will be circulated to the market for sale after maturity, while seed corn will be used as seed, which is one of the foundations of agricultural activities in the next year.

[0003] At present, the industry generally identifies illegal seed production farms by excluding legal seed production farms from all seed production farms, and the remaining farms are illegal seed production farms. The data of legal seed production farms can be obtained accurately from seed companies, while the full amount of seed production farms are currently often identified by on-site inspections and remote sensing image recognition.

[0004] On-site inspections and surveys rely solely on labor costs and experience for identification, and there are the following problems: there is no obvious difference between corn seed fields and field corn for a long time, the transportation conditions of agricultural fields are poor, and the geographical location is remote. Therefore, a large amount of manpower is often required for field inspections, which cannot meet the needs of large-scale seed field identification.

[0005] Related technologies have proposed remote sensing image identification. Remote sensing technology can be used to photograph large areas of corn fields, but it requires manual identification based on the stripe images of seed corn fields that are different from those of field corn, and a large number of people are required to perform image recognition. In addition, some illegal seed corn fields will avoid stripe sowing when sowing in order to avoid monitoring, resulting in inability to identify. Summary of the invention

[0006] The embodiments of the present application provide a seed field prediction processing method, device, storage medium and electronic device to solve the problem in the related art that seed fields are identified through remote sensing images, which requires manual identification based on the stripe images presented by the seed corn fields that are different from those of field corn, and also requires a large number of manual efforts for image recognition, and non-striped sowing cannot be identified.

[0007] According to one embodiment of the present application, a seed production field prediction processing method is provided, comprising:

[0008] Determining feature weights of signaling location data of a field to be identified;

[0009] Acquire target characteristic variables related to seed field identification according to the characteristic weights of the signaling position data to be identified;

[0010] The target characteristic variables are input into a pre-trained target recognition model to perform seed field prediction and obtain prediction results of the fields to be identified.

[0011] Optionally, the method further comprises:

[0012] Determining feature weights of signaling location training data;

[0013] Acquire training feature variables related to seed production field identification according to feature weights of the signaling position training data;

[0014] The seed field prediction model is trained according to the training characteristic variables of the signaling position training data to obtain the trained target recognition model.

[0015] Optionally, the seed field prediction model is trained according to the training feature variables of the signaling position training data to obtain the trained target recognition model, which includes:

[0016] Determine feature thresholds;

[0017] Binarizing the training feature variable according to the feature threshold to obtain a binary feature of the training feature variable;

[0018] The seed field prediction model is trained according to the binary features and corresponding labels to obtain the trained target recognition model.

[0019] Optionally, determining the feature threshold comprises:

[0020] Perform binning processing on the training feature variables, and determine the WOE value and IV value of the training feature variables;

[0021] The feature threshold is determined according to the WOE value, IV value and sample proportion distribution.

[0022] Optionally, binning the training feature variables and determining the WOE value and IV value of the training feature variables includes:

[0023] Performing binning processing on the training feature variables to obtain multiple grouping samples;

[0024] The WOE value corresponding to each grouped sample i is determined in the following way:

[0025]

[0026] The IV value corresponding to each grouped sample i is determined in the following way:

[0027]

[0028] Determine the sum of the IV values ​​of the multiple grouped samples as the IV value of the target feature variable;

[0029] py i pn represents the proportion of seed-producing farms in grouped sample i to the total number of seed-producing farms in the sample. i represents the proportion of non-seed-producing farms in group i to the total number of non-seed-producing farms in the sample, y i ,n i Respectively represent the number of seed-producing fields and non-seed-producing fields in group i, y T ,n T They represent the number of seed-producing fields and non-seed-producing fields in the total sample respectively, and m represents the number of bin groups.

[0030] Optionally, determining the feature threshold according to the WOE value, the IV value, and the sample proportion distribution includes:

[0031] According to the WOE value distribution, IV value, and positive / negative sample ratio of the multiple grouped samples, detecting whether a certain continuous interval and the positive / negative sample ratio show a monotonic distribution;

[0032] If so, the upper limit of the continuous interval is determined as the feature threshold.

[0033] Optionally, determining the feature weight of the signaling location data of the field to be identified includes:

[0034] The decision-making objectives, decision-making criteria and decision-making objects are divided into the highest level, the middle level and the lowest level according to their mutual relationships, and a hierarchical model is established;

[0035] Comparing the factors in each level of the hierarchical structure model in pairs, and constructing a discriminant matrix according to the comparison results;

[0036] Verifying the consistency of the discriminant matrix;

[0037] Determine the feature weights of all factors in each layer from the highest level to the lowest level in order of relative importance to the highest level.

[0038] Optionally, checking the consistency of the discriminant matrix includes:

[0039] Determine the consistency index CI value of the discriminant matrix;

[0040] Compare the consistency index CI and the random consistency index RI to determine the test coefficient CR;

[0041] If the check coefficient is less than a preset value, it is determined that the discriminant matrix passes the consistency check;

[0042] If the check coefficient is greater than or equal to the preset value, it is determined that the discriminant matrix fails the consistency check.

[0043] Optionally, obtaining target characteristic variables related to seed production field identification according to characteristic weights of the signaling position data to be identified includes:

[0044] The feature variables whose feature weights in the discriminant matrix are greater than a preset threshold are selected as the target feature variables.

[0045] According to another embodiment of the present application, a seed production field prediction processing device is provided, comprising:

[0046] A first determination module is used to determine the feature weight of the signaling position data of the field to be identified;

[0047] A first acquisition module, used for acquiring target characteristic variables related to seed production field identification according to the characteristic weights of the signaling position data to be identified;

[0048] The prediction module is used to input the target feature variables into a pre-trained target recognition model to perform seed field prediction and obtain the prediction results of the field to be identified.

[0049] Optionally, the device further comprises:

[0050] A second determination module is used to determine the feature weights of the signaling location training data;

[0051] A second acquisition module is used to acquire training feature variables related to seed production field identification according to the feature weights of the signaling position training data;

[0052] The training module is used to train the seed field prediction model according to the training characteristic variables of the signaling position training data to obtain the trained target recognition model.

[0053] Optionally, the training module includes:

[0054] A first determination submodule is used to determine a feature threshold;

[0055] A binarization processing submodule, used for performing binarization processing on the training feature variable according to the feature threshold value to obtain the binary feature of the training feature variable;

[0056] The training submodule is used to train the seed field prediction model according to the binary features and corresponding labels to obtain the trained target recognition model.

[0057] Optionally, the first determining submodule includes:

[0058] A binning processing unit, used for binning the training feature variables and determining the WOE value and IV value of the training feature variables;

[0059] The determination unit is used to determine the feature threshold according to the WOE value, the IV value and the sample proportion distribution.

[0060] Optionally, the sharing processing unit is further used for:

[0061] Performing binning processing on the training feature variables to obtain multiple grouping samples;

[0062] The WOE value corresponding to each grouped sample i is determined in the following way:

[0063]

[0064] The IV value corresponding to each grouped sample i is determined in the following way:

[0065]

[0066] Determine the sum of the IV values ​​of the multiple grouped samples as the IV value of the target feature variable;

[0067] py i pn represents the proportion of seed-producing farms in grouped sample i to the total number of seed-producing farms in the sample. i represents the proportion of non-seed-producing farms in group i to the total number of non-seed-producing farms in the sample, y i ,n i Respectively represent the number of seed-producing fields and non-seed-producing fields in group i, y T ,n T They represent the number of seed-producing fields and non-seed-producing fields in the total sample respectively, and m represents the number of bin groups.

[0068] Optionally, the determining unit is further configured to:

[0069] According to the WOE value distribution, IV value, and positive / negative sample ratio of the multiple grouped samples, detecting whether a certain continuous interval and the positive / negative sample ratio show a monotonic distribution;

[0070] If so, the upper limit of the continuous interval is determined as the feature threshold.

[0071] Optionally, the first determining module includes:

[0072] Establish submodules to divide decision goals, decision criteria and decision objects into the highest level, the middle level and the lowest level according to their mutual relationships, and establish a hierarchical model;

[0073] A comparison submodule, used for comparing the factors in each level of the hierarchical model in pairs, and constructing a discriminant matrix according to the comparison results;

[0074] A check submodule, used to check the consistency of the discriminant matrix;

[0075] The second determination submodule is used to determine the feature weights of the relative importance of all factors in each layer to the highest layer in order from the highest layer to the lowest layer.

[0076] Optionally, the syndrome module is further used to:

[0077] Determine the consistency index CI value of the discriminant matrix;

[0078] Compare the consistency index CI and the random consistency index RI to determine the test coefficient CR;

[0079] If the check coefficient is less than a preset value, it is determined that the discriminant matrix passes the consistency check;

[0080] If the check coefficient is greater than or equal to the preset value, it is determined that the discriminant matrix fails the consistency check.

[0081] Optionally, the first acquisition module is further used to:

[0082] The feature variables whose feature weights in the discriminant matrix are greater than a preset threshold are selected as the target feature variables.

[0083] According to another embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0084] According to another embodiment of the present application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0085] In an embodiment of the present application, feature weights of signaling position data of a field to be identified are determined; target feature variables related to seed field identification are obtained according to the feature weights of the signaling position data to be identified; the target feature variables are input into a pre-trained target recognition model to perform seed field prediction, and a prediction result of the field to be identified is obtained. This can solve the problem in the related art of identifying seed fields through remote sensing images, which requires manual identification based on stripe images of seed corn fields that are different from those of field corn, and also requires a large number of manual efforts for image recognition, and the problem that non-striped sowing cannot be identified. This can be applied to large-scale, accurate seed field identification, and seed field identification can be completed without manual participation, thereby reducing labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 It is a hardware structure block diagram of a mobile terminal of the seed production field prediction processing method according to an embodiment of the present application;

[0087] Figure 2 is a flow chart of a seed production field prediction processing method according to an embodiment of the present application;

[0088] Figure 3 is a flow chart of a seed production field prediction processing method according to an optional embodiment of the present application;

[0089] Figure 4 is a flow chart of corn seed field identification based on telecommunication operator mobile phone signaling according to an embodiment of the present application;

[0090] Figure 5 It is a schematic diagram of the ratio of the average number of people per day in the ear removal and emasculation period to the sowing period and the proportion of marginal seed production fields according to an embodiment of the present application;

[0091] Figure 6 It is a schematic diagram of the ratio of the average number of people per day during the ear removal and emasculation period to the average number of people per day during the sowing period and the proportion of marginal seed production fields according to an embodiment of the present application;

[0092] Figure 7 It is a schematic diagram of the ratio of the average number of people per day in the ear removal and emasculation period to the sowing period and the proportion of marginal seed production fields according to an embodiment of the present application;

[0093] Figure 8 It is a schematic diagram of four Woe values ​​and the proportion of marginal seed production fields according to the ratio of the average daily residence time during the de-tasseling period to the growth period according to the embodiment of the present application;

[0094] Fig. 9 is a schematic diagram of a ROC curve according to an embodiment of the present application;

[0095] Fig.10 is a block diagram of a seed production field prediction processing device according to an embodiment of the present application;

[0096] Fig.11 It is a block diagram of a seed field prediction processing device according to an optional embodiment of the present application. DETAILED DESCRIPTION

[0097] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0098] It should be noted that the terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0099] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 is a hardware structure block diagram of a mobile terminal of the seed field prediction processing method of an embodiment of the present application, such as Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0100] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the seed farm prediction processing method in the embodiment of the present application. The processor 102 executes various functional applications and service chain address pool slice processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the mobile terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0101] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0102] In this embodiment, a seed field prediction processing method running on the above mobile terminal or network architecture is provided. Figure 2 is a flowchart of a seed field prediction processing method according to an embodiment of the present application, such as Figure 2 As shown, the process includes the following steps:

[0103] Step S202, determining the feature weight of the signaling location data of the field to be identified;

[0104] In an embodiment of the present application, the above-mentioned step S202 can specifically determine the feature weight in the following manner: divide the decision objectives, decision criteria and decision objects into the highest layer, the middle layer and the lowest layer according to their mutual relationships, and establish a hierarchical model; compare the factors in each layer in the hierarchical model in pairs, and construct a discriminant matrix based on the comparison results; verify the consistency of the discriminant matrix, specifically, determine the consistency index CI value of the discriminant matrix, compare the consistency index CI with the random consistency index RI to determine the test coefficient CR; if the test coefficient is less than a preset value, determine that the discriminant matrix passes the consistency check, if the test coefficient is greater than or equal to the preset value, determine that the discriminant matrix fails the consistency check; finally, determine the feature weights of the relative importance of all factors in each layer to the highest layer from the highest layer to the lowest layer.

[0105] Step S204, obtaining target characteristic variables related to seed production field identification according to the characteristic weights of the signaling position data to be identified;

[0106] Correspondingly, in the above step S204, the feature variables whose feature weights in the above discriminant matrix are greater than a preset threshold are selected as target feature variables.

[0107] Step S206, inputting the target feature variable into a pre-trained target recognition model to perform seed field prediction and obtain the prediction result of the field to be identified.

[0108] Through the above steps S202 to S206, the problem that seed fields are identified through remote sensing images in related technologies needs to be manually identified based on the stripe images of seed corn fields that are different from those of field corn, and a large number of people need to be invested in image recognition, and non-striped sowing cannot be identified can be solved. This can be applied to large-scale and accurate seed field identification, and seed field identification can be completed without human participation, thereby reducing labor costs.

[0109] The embodiment of the present invention is based on real-time and full telecommunication signaling data, and can make judgments on suspected corn seed-producing fields according to the identification business model in real time, thereby effectively providing accurate digital management tools. While ensuring explainability, it provides more efficient and accurate management capabilities for illegal seed-producing fields, and liberates a large amount of manpower.

[0110] In the related art, seed production fields can only be identified during the heading period from July to August, when the sowing season has been missed, and legal crops cannot be replanted after being discovered. The embodiment of the present invention distinguishes based on cultivation differences. In practice, it is found that in addition to the heading period (July-August), there are also differences in cultivation characteristics between field corn and seed corn during the sowing period (April-May). Further exploration can be carried out on the sowing period to advance the time window for discovering illegal seed production fields and maximize the losses of the country and farmers.

[0111] The embodiment of the present application is based on the identification of corn seed fields by telecommunications mobile phone signaling, and uses the AHP hierarchical analysis method to screen features in combination with actual business conditions to obtain feature indicators that are more relevant to seed field identification. The information contained in a single category is calculated by the WOE and IV values ​​of the binary classification scenario to obtain the threshold of the feature indicator, and then the indicator is binarized. Finally, the data is divided into a training set and a validation set according to a certain ratio, a supervised learning model is constructed, and it is applied to corn seed field identification.

[0112] The present application also provides a training method for the target recognition model. After training the model, it is possible to predict whether the field belongs to a seed production field. Figure 3 is a flowchart of a seed field prediction processing method according to an optional embodiment of the present application, such as Figure 3 As shown, including:

[0113] Step S302, determining feature weights of signaling location training data;

[0114] Step S304, obtaining training feature variables related to seed production field identification according to the feature weights of the signaling position training data;

[0115] Step S306, training the seed field prediction model according to the training feature variables of the signaling position training data to obtain the trained target recognition model.

[0116] The target recognition model obtained by training through the above steps S302 to S306 is more stable, and the prediction result is more accurate when it is used to identify whether a field is a seed production field.

[0117] In the above step S306, training can be performed in the following manner:

[0118] S3061, determining a feature threshold;

[0119] S3062, binarizing the training feature variable according to the feature threshold to obtain a binary feature of the training feature variable;

[0120] S3063: Train the seed field prediction model according to the binary features and corresponding labels to obtain the trained target recognition model.

[0121] In the embodiment of the present application, S3061 may include: binning the training feature variables, and determining the WOE value and IV value of the training feature variables. Further, binning the training feature variables to obtain multiple grouping samples, and determining the WOE value corresponding to each grouping sample i in the following manner:

[0122]

[0123] The IV value corresponding to each grouped sample i is determined in the following way:

[0124]

[0125] The sum of the IV values ​​of the multiple grouped samples is determined as the IV value of the target feature variable, where py i pn represents the proportion of seed-producing farms in grouped sample i to the total number of seed-producing farms in the sample. i represents the proportion of non-seed-producing farms in group i to the total number of non-seed-producing farms in the sample, y i ,n i Respectively represent the number of seed-producing fields and non-seed-producing fields in group i, y T ,n T Respectively represent the number of seed-producing fields and non-seed-producing fields in the total sample, and m represents the number of bin groups. Then, the characteristic threshold is determined according to the WOE value, IV value and sample proportion distribution. Specifically, according to the WOE value distribution, IV value and positive / negative sample proportion of the multiple grouped samples, it is detected whether a certain continuous interval and the positive / negative sample proportion show a monotonic distribution; if so, the upper limit of the continuous interval is determined as the characteristic threshold.

[0126] The seed production field identification method based on the signaling data of telecom operators has solved the problem of short time window in the traditional seed production field identification process. In the traditional identification method, the identification time is concentrated in the emasculation stage, but this method can advance the identification time to the sowing period, helping relevant departments to judge suspected seed production fields and illegal seed production fields as soon as possible. Figure 4 is a flow chart of corn seed field identification based on telecommunication operator mobile phone signaling according to an embodiment of the present application, such as Figure 4 As shown, including:

[0127] S401, using AHP hierarchical analysis method to calculate feature weights and obtain feature indicators that are strongly related to seed field identification

[0128] The AHP method decomposes the problem into different components according to the nature of the problem and the overall goal to be achieved, and clusters and combines the factors at different levels according to the interrelated influences and affiliations between the factors to form a multi-level analysis structure model, so that the problem is ultimately reduced to the determination of the relative importance of the lowest level (schemes and measures for decision-making, etc.) relative to the highest level (overall goal) or the ranking of the relative order of superiority and inferiority. The complete AHP method usually includes four steps:

[0129] Establish a hierarchical model, divide the decision-making goals, factors to be considered (decision-making criteria) and decision-making objects into the highest level, middle level and lowest level according to their mutual relationships, and draw a hierarchical diagram. The highest level refers to the purpose of decision-making and the problem to be solved. The lowest level refers to the alternative plans when making decisions. The middle level refers to the factors to be considered and the criteria for decision-making. For the two adjacent levels, the higher level is called the goal level and the lower level is called the factor level.

[0130] When constructing a discriminant matrix, if the weights between factors at each level are determined, it is often not easy to be accepted by others if the result is only qualitative. Therefore, Saaty et al. proposed the consistent matrix method, that is, not comparing all factors together, but comparing them two by two. At this time, relative scales are used to minimize the difficulty of comparing factors of different natures, so as to improve accuracy. For example, for a certain criterion, the schemes under it are compared two by two, and the levels are evaluated according to their importance. ij Table 1 lists the nine importance levels and their values ​​given by Saaty. The matrix formed by the pairwise comparison results is called the discriminant matrix. The discriminant matrix has the properties shown in Table 1.

[0131]

[0132] Table 1

[0133] Factor i is better than factor j Quantization value Equally important 1 Slightly important 3 Strong and important 5 Strongly important 7 Extremely important 9 The middle value of two adjacent judgments 2，4，6，8

[0134] Hierarchical single sorting and its consistency test, corresponding to the eigenvector of the maximum eigenvalue of the discriminant matrix, is normalized (so that the sum of the elements in the vector is equal to 1) and recorded as W. The elements of W are the ranking weights of the relative importance of the factors at the same level to the factors at the previous level. This process is called hierarchical single sorting. Whether the hierarchical single sorting can be confirmed requires a consistency test. The so-called consistency test refers to the allowable range of inconsistency determined for A. Among them, the only non-zero eigenvalue of the n-order consistent matrix is ​​n; the maximum eigenvalue of the n-order positive reciprocal matrix A, A is a consistent matrix if and only if λ=n.

[0135] Since λ continuously depends on a ij , the more λ is larger than n, the more serious the inconsistency of A. The consistency index is calculated by CI. The smaller the CI, the greater the consistency. The eigenvector corresponding to the maximum eigenvalue is used as the weight vector of the influence of the compared factor on a certain factor in the upper layer. The greater the inconsistency, the greater the judgment error caused. Therefore, the value of λ-n can be used to measure the inconsistency of A. The consistency index is defined as:

[0136]

[0137] CI = 0, there is complete consistency; CI is close to 0, there is satisfactory consistency; the larger the CI, the more serious the inconsistency. In order to measure the size of CI, the random consistency index RI is introduced:

[0138]

[0139] Among them, the random consistency index RI is related to the order of the discriminant matrix. Generally speaking, the larger the matrix order, the greater the possibility of random deviation from consistency. The corresponding relationship is shown in Table 2.

[0140] Table 2

[0141]

[0142] Considering that the deviation from consistency may be caused by random reasons, when testing whether the discriminant matrix has satisfactory consistency, it is also necessary to compare CI and random consistency index RI to obtain the test coefficient CR, the formula is as follows:

[0143]

[0144] Generally, if CR < 0.1, the discriminant matrix is ​​considered to pass the consistency test, otherwise it does not have satisfactory consistency.

[0145] The total ranking of levels and its consistency test, which calculates the relative importance of all factors at a certain level to the highest level (total goal), is called the total ranking of levels. This process is carried out from the highest level to the lowest level.

[0146] In this embodiment, according to the relevant feedback from the Ministry of Agriculture and on-site surveys and communications with seed companies and local farmers, it is learned that the biggest difference between seed corn and field corn is that seed corn needs to arrange labor to remove ears and tassels, while field corn does not need it. Therefore, the difference in population thermal data during the de-tasseling and tasseling period is quite obvious. The characteristic variables shown in Table 3 are preliminarily obtained through analysis of population thermal data in different farming periods and differences in farming modes between seed fields and non-seed fields.

[0147] Table 3

[0148]

[0149] The original feature data of signaling location data and label data during the sowing period (April 10 to May 10), the growing period (June 1 to June 30), and the ear removal and emasculation period (July 1 to July 31) were filtered by day.

[0150] The original features include information related to human activities (the ratio of the average daily number of people during the de-tasseling period to the sowing period, the ratio of the average daily residence time during the de-tasseling period to the sowing period, the ratio of the average daily number of people during the de-tasseling period to the growing period, and the ratio of the average daily residence time during the de-tasseling period to the growing period), plot identification (plot area), and attributes of users who stop / reside in the plot (average age of users, average network access time of users, top 1 of users' Internet preferences, and daily average of the proportion of users suspected of working). Feature factors are set and sorted;

[0151] The discriminant matrix is ​​constructed according to the satty score, as shown in Table 4.

[0152] Table 4

[0153]

[0154]

[0155] After consistency test and hierarchical sorting, the consistency test passed and the following weights can be used for analytic hierarchy process calculation. The maximum eigenvalue is 9.4717, the CI value is 0.0590, and the CR value is 0.0404;

[0156] The judgment results are shown in Table 5.

[0157] Table 5

[0158]

[0159] According to the consistency test results, characteristic indicators that are strongly correlated with seed production field identification can be obtained, including the ratio of the average daily number of people during the ear removal and tassel removal period to the sowing period, the ratio of the average daily residence time during the ear removal and tassel removal period to the sowing period, the ratio of the average daily number of people during the ear removal and tassel removal period to the growing period, and the ratio of the average daily residence time during the ear removal and tassel removal period to the growing period.

[0160] S402, using the WOE and IV values ​​of the binary classification scenario to calculate the feature to determine the amount of information contained in a single category.

[0161] Considering that the target variable in the business scenario is a binary variable, namely seed-producing fields and non-seed-producing fields, the WOE method is used to bin the four human activity characteristics of the fields, and the IV value is used to evaluate the predictive ability of the characteristics. The statistical formulas of WOE and IV values ​​are as follows:

[0162] For the WOE value corresponding to group i, the calculation formula is as follows:

[0163]

[0164] Similarly, for the IV value corresponding to group i, the calculation formula is as follows:

[0165]

[0166] The characteristic IV value is the sum of the IV values ​​of each group:

[0167]

[0168] Among them, py i It indicates the proportion of “seed-producing farms” in group i to the total sample of seed-producing farms;

[0169] pn i It indicates the proportion of “non-seed-producing fields” in group i to the total number of non-seed-producing fields in the sample;

[0170] y i and n i Respectively represent the number of “seed-producing fields” and “non-seed-producing fields” in group i;

[0171] y T and n T Respectively represent the number of “seed-producing fields” and “non-seed-producing fields” in the total sample;

[0172] m represents the number of bin groups.

[0173] S403, obtaining a feature threshold according to the WOE and the sample proportion distribution.

[0174] According to the WOE distribution of the binning interval and the proportion of positive / negative samples, check whether a certain continuous interval and the proportion of positive / negative samples show a monotonic distribution. If so, it indicates that the continuous interval has a clear division ability for the target variable, and the upper limit of the continuous interval is used as the threshold for feature preprocessing.

[0175] Based on the results of the field survey, the time range of the local corn planting period, corn growing period and ear and tassel removal period were determined, and it was clarified that there were obvious characteristics of increased working time and more workers during the corn planting period and ear and tassel removal period.

[0176] The WOE value and IV value were calculated by binning each feature of the variables "ratio of the average daily number of people during the de-tasseling period to the sowing period", "ratio of the average daily residence time during the de-tasseling period to the sowing period", "ratio of the average daily number of people during the de-tasseling period to the growing period", and "ratio of the average daily residence time during the de-tasseling period to the growing period".

[0177] When the characteristic variables "the ratio of the average number of people per day during the de-tasseling period to the sowing period" and "the ratio of the average daily residence time during the de-tasseling period to the sowing period" were greater than -10%, the response ratios for whether it was a corn seed field were significantly different;

[0178] When the characteristic variables "ratio of average daily number of people during de-tasseling and growing period" and "ratio of average daily stay time during de-tasseling and growing period" were greater than 0%, the response proportions for whether it was a corn seed field were significantly different.

[0179] According to woe, the relationship between woe and the proportion of marginal seed production fields corresponding to each characteristic was obtained. When the ratio of the average daily number of people in the ear removal and tassel removal period to the sowing period, and the ratio of the average daily residence time in the ear removal and tassel removal period to the sowing period were greater than -10%, the response proportion of whether it was a seed production field was significantly different (there was only one minimum woe value interval, and it remained monotonic overall); when the ratio of the average daily number of people in the ear removal and tassel removal period to the growth period, and the ratio of the average daily residence time in the ear removal and tassel removal period to the growth period were greater than 0%, the response proportion of whether it was a corn seed production field was significantly different (there was more than one minimum woe value interval, and the overall performance was a fluctuating distribution, and the interval [0, 1] showed an overall downward trend), the woe value of the average daily number of people in the ear removal and tassel removal period to the sowing period and the proportion of marginal seed production fields were Figure 5 shown.

[0180] The ratio of the average number of people per day during the de-tasseling period to the average number of people per day during the growing period and the proportion of marginal seed production fields, such as Figure 6 shown.

[0181] The woe value of the average daily number of people during the detasseling period and the sowing period and the proportion of marginal seed production fields, such as Figure 7 shown.

[0182] The ratio of the average daily residence time during the detasseling period to the growth period, the four Woe values ​​and the proportion of marginal seed production fields Figure 8 shown.

[0183] S404, performing binarization processing on the feature variables using the feature threshold.

[0184] After obtaining the feature threshold, feature transformation is immediately carried out to achieve variable binarization processing.

[0185] According to the characteristic threshold, the ratio of the average number of people per day between the de-tasseling period and the sowing period, and the ratio of the average daily residence time between the de-tasseling period and the sowing period were marked as 1 when they were greater than -10%, otherwise they were marked as 0; the ratio of the average number of people per day between the de-tasseling period and the growing period, and the ratio of the average daily residence time between the de-tasseling period and the growing period were marked as 1 when they were greater than 0%, otherwise they were marked as 0. The results of the binarization processing are shown in Table 6.

[0186] Table 6

[0187]

[0188] S405, using GBDT to predict cultivated land.

[0189] GBDT (Gradient Boosting Decision Tree) is an iterative decision tree algorithm that consists of multiple decision trees (CART regression trees) (each decision tree is iterated serially), and the conclusions of all trees are accumulated to get the final answer.

[0190] GBDT is different from traditional Boost ing in that each calculation is to reduce the residual of the previous time, and in order to eliminate the residual, the model is built in the gradient direction where the residual decreases. Therefore, in GBDT, each new decision tree model is built to make the residual of the previous model descend in the gradient, which is very different from the traditional Boost ing that focuses on the weighting of correct and incorrect samples. In the GradientBoost ing algorithm, the key is to use the value of the negative gradient direction of the loss function in the current model as an approximate value of the residual, and then fit a CART regression tree.

[0191] The model retention validation was carried out on the training samples (the validation set accounted for 40%), and the results are shown in Table 7.

[0192] Table 7

[0193] type Precision Recall F1 Non-seed farm 0.50 0.29 0.36 Seed Farm 0.89 0.95 0.92

[0194] The AUC value is 0.81, and the ROC curve is as follows Fig. 9 shown.

[0195] Based on the verification result data, the results of the real seed production field are identified and the positive sample recognition rate is calculated.

[0196] Verification results: Taking the verification data of Ganzhou District as an example, 114 real and scattered seed-producing fields were recorded in batches, 98 of which were determined by the model to be suspected seed-producing fields, and the positive sample recognition rate was 85.96%, which is similar to the internal verification of the model.

[0197] The embodiment of the present application utilizes Python language, machine learning algorithm and Spark distributed computing framework to perform distributed encapsulation of relevant technical details, and performs containerized management to ensure the isolation of user data and resources; the seed field evaluation page displayed on the front end is designed to distinguish the field identification in a front-end and back-end linkage manner. Based on the feature screening method of AHP analytic hierarchy process, characteristic indicators that are more relevant to seed field identification are obtained, and the information contained in a single category is calculated by the WOE and IV value of the binary classification scenario to obtain the threshold of the characteristic indicator, thereby realizing the binary processing of the indicator. On the premise that the above processing is completed, the data is divided into a training set and a validation set according to a certain ratio, a supervised learning model is constructed, and it is applied to the identification of corn seed fields in large quantities of fields.

[0198] According to another embodiment of the present application, a seed production field prediction processing device is provided. Fig.10 is a block diagram of a seed field prediction processing device according to an embodiment of the present application, such as Fig.10 As shown, including:

[0199] A first determination module 102 is used to determine the feature weight of the signaling position data of the field to be identified;

[0200] A first acquisition module 104, configured to acquire target characteristic variables related to seed production field identification according to characteristic weights of the signaling position data to be identified;

[0201] The prediction module 106 is used to input the target feature variables into a pre-trained target recognition model to perform seed field prediction and obtain the prediction results of the field to be identified.

[0202] Fig.11 is a block diagram of a seed field prediction processing device according to an optional embodiment of the present application, such as Fig.11 As shown, the device also includes:

[0203] A second determination module 112, configured to determine feature weights of signaling location training data;

[0204] A second acquisition module 114, configured to acquire training feature variables related to seed production field identification according to feature weights of the signaling position training data;

[0205] The training module 116 is used to train the seed field prediction model according to the training feature variables of the signaling position training data to obtain the trained target recognition model.

[0206] Optionally, the training module 116 includes:

[0207] A first determination submodule, used to determine a feature threshold;

[0208] A binarization processing submodule, used for performing binarization processing on the training feature variable according to the feature threshold value to obtain the binary feature of the training feature variable;

[0209] The training submodule is used to train the seed field prediction model according to the binary features and corresponding labels to obtain the trained target recognition model.

[0210] Optionally, the first determining submodule includes:

[0211] A binning processing unit, used for binning the training feature variables and determining the WOE value and IV value of the training feature variables;

[0212] The determination unit is used to determine the feature threshold according to the WOE value, the IV value and the sample proportion distribution.

[0213] Optionally, the sharing processing unit is further used for:

[0214] Performing binning processing on the training feature variables to obtain multiple grouping samples;

[0215] The WOE value corresponding to each grouped sample i is determined in the following way:

[0216]

[0217] The IV value corresponding to each grouped sample i is determined in the following way:

[0218]

[0219] Determine the sum of the IV values ​​of the multiple grouped samples as the IV value of the target feature variable;

[0220] py i pn represents the proportion of seed-producing farms in grouped sample i to the total number of seed-producing farms in the sample. i represents the proportion of non-seed-producing farms in group i to the total number of non-seed-producing farms in the sample, y i ,n i Respectively represent the number of seed-producing fields and non-seed-producing fields in group i, y T ,n TThey represent the number of seed-producing fields and non-seed-producing fields in the total sample respectively, and m represents the number of bin groups.

[0221] Optionally, the determining unit is further configured to:

[0222] According to the WOE value distribution, IV value, and positive / negative sample ratio of the multiple grouped samples, detecting whether a certain continuous interval and the positive / negative sample ratio show a monotonic distribution;

[0223] If so, the upper limit of the continuous interval is determined as the feature threshold.

[0224] Optionally, the first determining module 102 includes:

[0225] Establish submodules to divide decision goals, decision criteria and decision objects into the highest level, the middle level and the lowest level according to their mutual relationships, and establish a hierarchical model;

[0226] A comparison submodule, used for comparing the factors in each level of the hierarchical model in pairs, and constructing a discriminant matrix according to the comparison results;

[0227] A check submodule, used to check the consistency of the discriminant matrix;

[0228] The second determination submodule is used to determine the feature weights of the relative importance of all factors in each layer to the highest layer in order from the highest layer to the lowest layer.

[0229] Optionally, the syndrome module is further used to:

[0230] Determine the consistency index CI value of the discriminant matrix;

[0231] Compare the consistency index CI and the random consistency index RI to determine the test coefficient CR;

[0232] If the check coefficient is less than a preset value, it is determined that the discriminant matrix passes the consistency check;

[0233] If the check coefficient is greater than or equal to the preset value, it is determined that the discriminant matrix fails the consistency check.

[0234] Optionally, the first acquisition module 104 is further configured to:

[0235] The feature variables whose feature weights in the discriminant matrix are greater than a preset threshold are selected as the target feature variables.

[0236] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute at least the following steps in any one of the above method embodiments when running:

[0237] S1, determining the feature weight of the signaling location data of the field to be identified;

[0238] S2, obtaining target characteristic variables related to seed production field identification according to the characteristic weights of the signaling position data to be identified;

[0239] S3, inputting the target feature variable into a pre-trained target recognition model to perform seed field prediction and obtain the prediction result of the field to be identified.

[0240] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0241] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to perform at least the following steps in any one of the above method embodiments:

[0242] S1, determining the feature weight of the signaling location data of the field to be identified;

[0243] S2, obtaining target characteristic variables related to seed production field identification according to the characteristic weights of the signaling position data to be identified;

[0244] S3, inputting the target feature variable into a pre-trained target recognition model to perform seed field prediction and obtain the prediction result of the field to be identified.

[0245] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0246] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0247] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0248] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A seed production field prediction processing method, characterized in that: include: Determining feature weights of signaling location data of a field to be identified; Acquire target characteristic variables related to seed production field identification according to the characteristic weights of the signaling position data to be identified; Inputting the target characteristic variable into a pre-trained target recognition model to perform seed production field prediction and obtain the prediction result of the field to be identified; The method further comprises: Determining feature weights of signaling location training data; Acquire training feature variables related to seed field identification according to feature weights of the signaling position training data; The seed field prediction model is trained according to the training characteristic variables of the signaling position training data to obtain the trained target recognition model, including: Determining the feature threshold includes: binning the training feature variables and determining the WOE value and IV value of the training feature variables; determining the feature threshold according to the WOE value, IV value and sample proportion distribution; Binarizing the training feature variable according to the feature threshold to obtain a binary feature of the training feature variable; The seed field prediction model is trained according to the binary features and corresponding labels to obtain the trained target recognition model; The step of binning the training feature variables and determining the WOE value and IV value of the training feature variables includes: binning the training feature variables to obtain a plurality of grouping samples; and determining the WOE value corresponding to each grouping sample i in the following manner: The IV value corresponding to each grouped sample i is determined in the following way: Determine the sum of the IV values ​​of the multiple grouped samples as the IV value of the target feature variable; py i pn represents the proportion of seed-producing farms in grouped sample i to the total number of seed-producing farms in the sample. i represents the proportion of non-seed-producing farms in group i to the total number of non-seed-producing farms in the sample, y i ,n i Respectively represent the number of seed-producing fields and non-seed-producing fields in group i, y T ,n T They represent the number of seed-producing fields and non-seed-producing fields in the total sample, respectively, and m represents the number of bin groups; Among them, determining the feature threshold according to the WOE value, IV value and sample proportion distribution includes: detecting whether a certain continuous interval and the positive / negative sample proportion show a monotonic distribution according to the WOE value distribution, IV value, and positive / negative sample proportion of the multiple grouped samples; if so, determining the upper limit of the continuous interval as the feature threshold.

2. The method according to claim 1, characterized in that The feature weights for determining the signaling location data of the field to be identified include: The decision-making objectives, decision-making criteria and decision-making objects are divided into the highest level, the middle level and the lowest level according to their mutual relationships, and a hierarchical model is established; Comparing the factors in each level of the hierarchical model in pairs, and constructing a discriminant matrix according to the comparison results; Verifying the consistency of the discriminant matrix; Determine the feature weights of all factors in each layer from the highest level to the lowest level in order of relative importance to the highest level.

3. The method according to claim 2, characterized in that Checking the consistency of the discriminant matrix includes: Determine the consistency index CI value of the discriminant matrix; Compare the consistency index CI and the random consistency index RI to determine the test coefficient CR; If the check coefficient is less than a preset value, it is determined that the discriminant matrix passes the consistency check; If the check coefficient is greater than or equal to the preset value, it is determined that the discriminant matrix fails the consistency check.

4. The method according to claim 3, characterized in that Acquiring target feature variables related to seed field identification according to the feature weights of the signaling position data to be identified includes: The feature variables whose feature weights in the discriminant matrix are greater than a preset threshold are selected as the target feature variables.

5. A seed field prediction processing device, characterized in that: include: A determination module, used to determine the feature weight of the signaling position data of the field to be identified; An acquisition module, used for acquiring target characteristic variables related to seed production field identification according to the characteristic weights of the signaling position data to be identified; A prediction module is used to input the target feature variable into a pre-trained target recognition model to perform seed production field prediction and obtain the prediction result of the field to be identified; The device also includes: A second determination module is used to determine the feature weights of the signaling location training data; A second acquisition module is used to acquire training feature variables related to seed production field identification according to the feature weights of the signaling position training data; A training module, used for training the seed field prediction model according to the training characteristic variables of the signaling position training data to obtain the trained target recognition model; Wherein, the training module includes: A first determination submodule is used to determine a feature threshold; A binarization processing submodule, used for performing binarization processing on the training feature variable according to the feature threshold value to obtain the binary feature of the training feature variable; A training submodule, used for training the seed field prediction model according to the binary features and corresponding labels to obtain the trained target recognition model; Wherein, the first determining submodule includes: A binning processing unit, used for binning the training feature variables and determining the WOE value and IV value of the training feature variables; A determination unit, used to determine the feature threshold according to the WOE value, the IV value and the sample proportion distribution; The binning processing unit is further used to: bin the training feature variables to obtain a plurality of grouping samples; and determine the WOE value corresponding to each grouping sample i by the following method: The IV value corresponding to each grouped sample i is determined in the following way: Determine the sum of the IV values ​​of the multiple grouped samples as the IV value of the target feature variable; py i pn represents the proportion of seed-producing farms in grouped sample i to the total number of seed-producing farms in the sample. i represents the proportion of non-seed-producing farms in group i to the total number of non-seed-producing farms in the sample, y i ,n i Respectively represent the number of seed-producing fields and non-seed-producing fields in group i, y T ,n T They represent the number of seed-producing fields and non-seed-producing fields in the total sample, respectively, and m represents the number of bin groups; Among them, the determination unit is also used to: detect whether a certain continuous interval and the positive / negative sample ratio show a monotonic distribution according to the WOE value distribution, IV value, and positive / negative sample ratio of the multiple grouped samples; if so, determine the upper limit of the continuous interval as the feature threshold.

6. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 4 when executed.

7. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 4.

Images