A jujube fruit planting method and system based on multi-source data analysis

Abstract

The present disclosure relates to a jujube planting method and system based on multi-source data analysis, comprising: analyzing a multi-source planting sample data set by a jujube planting optimization model of the last round of jujube planting analysis to generate planting analysis supervision data, determining a target jujube planting optimization model from the corresponding jujube planting optimization model of each round according to the planting analysis supervision data; determining a target planting data set from the corresponding multi-source planting data set of each round according to the planting analysis supervision data; loading the planting data in the multi-source planting data set corresponding to each round of planting into the jujube planting optimization model generated by the last round of jujube planting analysis iteration to generate a global planting supervision data set; comparing the global planting supervision data set and the multi-source planting data set corresponding to the target jujube to generate an expanded planting supervision data set; inputting the expanded planting supervision data set and the target planting data set into the target jujube planting optimization model to obtain a planting analysis result.

Description

Technical Field

[0001] This disclosure relates to the field of data analysis technology, and in particular to a method and system for jujube cultivation based on multi-source data analysis. Background Technology

[0002] In jujube cultivation, traditional methods often rely on farmers' experience and intuition, lacking scientific data support and precise decision-making basis. This leads to low planting efficiency, resource waste, and unstable yield and quality. To address these issues, intelligent planting management systems based on multi-source data analysis and machine learning technologies are gradually emerging. These systems optimize planting decisions and improve yield and quality by collecting and analyzing various data sources during the planting process, such as soil data, climate data, and pest and disease data.

[0003] However, existing intelligent planting management systems still have some problems. First, they often focus only on the analysis of planting data from a single round, ignoring the correlations and differences between multiple planting rounds. Second, when determining planting optimization models, these systems typically only consider data from the current round, failing to fully utilize the valuable information in historical data. This results in limited accuracy and applicability of the recommendations. Summary of the Invention

[0004] To overcome the problems existing in related technologies, this disclosure provides a jujube planting method and system based on multi-source data analysis. By comprehensively utilizing multi-source datasets from multiple planting cycles, a jujube planting model is constructed and optimized to generate more accurate and applicable planting recommendations.

[0005] According to a first aspect of the present disclosure, a method for jujube cultivation based on multi-source data analysis is provided, the method comprising: Obtain the multi-source planting dataset for each round of planting under the multiple planting conditions corresponding to the jujube fruit to be planted, and use it as the multi-source planting sample dataset called in this iteration of jujube planting data analysis; The jujube planting optimization model generated by the last round of jujube planting analysis iteration is used to analyze the multi-source planting sample dataset of each round, generate planting analysis supervision data, and determine the target jujube planting optimization model from the jujube planting optimization models generated by the corresponding jujube planting analysis iteration in each round based on the planting analysis supervision data. Based on the planting analysis and supervision data, the target planting dataset used for configuring the target jujube planting optimization model analysis iteration is determined from the multi-source planting sample dataset corresponding to each round; The planting data in the multi-source planting sample dataset corresponding to each round of planting is filtered to generate a global planting supervision dataset. The global planting supervision dataset is compared with the multi-source planting dataset corresponding to the target jujube of the same variety as the jujube to be planted, and an extended planting supervision dataset is generated. The target jujube is a jujube with planting conditions similar to those of the jujube to be planted. The extended planting supervision dataset and the target planting dataset are input into the target jujube planting optimization model to obtain the planting analysis results output by the target jujube planting optimization model. The jujube planting optimization model generated by this round of jujube planting analysis iteration is also obtained. The planting analysis results include the fertilization time nodes, irrigation time nodes, and pesticide spraying time nodes of this round of planting, as well as the amount of fertilizer applied, the amount of water applied, and the amount of pesticide sprayed for each application. This enables the jujube planting to be completed by applying fertilizer according to the fertilization time nodes and the amount of fertilizer applied, irrigating according to the irrigation time nodes and the amount of water applied, and spraying according to the pesticide spraying time nodes and the amount of pesticide sprayed.

[0006] In one possible implementation, determining the target jujube planting optimization model from the jujube planting optimization models generated by the corresponding iterative jujube planting analysis in each round based on the planting analysis supervision data includes: The jujube planting optimization models generated by the iterative analysis of jujube planting in each round are connected in time sequence to obtain the model sequence; Pre-set the sample data weights for each data point in the planting analysis and monitoring data; Based on the pre-set sample data weights, the planting analysis supervision data, and the planting labeling results for the planting analysis supervision data, weak learning training is performed in the order of each jujube planting optimization model in the model sequence to obtain the corresponding predicted planting results, weak learning error rate, and updated sample data weights. The jujube planting optimization model with the smallest weak learning error rate is determined as the target jujube planting optimization model.

[0007] In one possible implementation, the step of performing weak learning training according to the pre-set sample data weights, the planting analysis supervision data, and the planting labeling results for the planting analysis supervision data, in the order of each jujube planting optimization model in the model sequence, to obtain the corresponding predicted planting results, weak learning error rate, and updated sample data weights, includes: The planting analysis and supervision data are input into the first jujube planting optimization model in the model sequence, so that the first jujube planting optimization model is trained by weak learning according to the pre-set sample data weights to obtain the corresponding predicted planting results. The predicted planting results are compared with the planting labeling results to obtain the weak learning error rate corresponding to the first jujube planting optimization model; The pre-set sample data weights are updated based on the weak learning error rate corresponding to the first jujube planting optimization model, so as to obtain the updated sample data weights corresponding to the next jujube planting optimization model connected to the first jujube planting optimization model in the model sequence. Repeat the following steps: input the planting analysis supervision data, the planting labeling results, and the updated sample data weights corresponding to the previous jujube planting optimization model into the next jujube planting optimization model in sequence for weak learning training, to obtain the corresponding predicted planting results, weak learning error rate, and updated sample data weights, until all jujube planting optimization models have completed weak learning training.

[0008] In one possible implementation, the step of filtering the planting data in the multi-source planting sample dataset corresponding to each round of planting to generate a global planting supervision dataset includes: Expert knowledge labeling is performed on the planting data in the multi-source planting sample dataset corresponding to each round of planting to generate expert labels for the planting data in the multi-source planting sample dataset corresponding to each round of planting. Based on the expert labels, determine the conflict and consistency among the planting data in the multi-source planting sample dataset corresponding to each round of planting; If the conflict between any two planting data sets meets the preset conflict conditions, one of the two planting data sets is deleted based on the consistency between the two planting data sets and each of the other planting data sets. A global planting supervision dataset is generated based on all the planting data that has not been deleted.

[0009] In one possible implementation, the step of matching the global planting supervision dataset with the multi-source planting dataset corresponding to the target jujube of the same variety to be planted, to generate an extended planting supervision dataset, includes: The expert labels of each planting data in the global planting supervision dataset are matched with the expert labels of each planting data in the multi-source planting dataset corresponding to the target jujube of the same variety to be planted. The planting data in the multi-source planting dataset corresponding to the target jujube that do not achieve a match are used as the extended planting supervision dataset.

[0010] In one possible implementation, the jujube planting optimization model generated through the last round of jujube planting analysis iterations analyzes the multi-source planting sample dataset of each round to generate planting analysis supervision data, including: Set the initial sample data weights for each round of the multi-source planting sample dataset; Based on the initial sample data weights, the planting data in each round of the multi-source planting sample dataset is input into the jujube planting optimization model generated by the last round of jujube planting analysis iteration, and the first predicted planting analysis result of the jujube planting optimization model generated by the last round of jujube planting analysis iteration predicts the multi-source planting dataset corresponding to each round of planting. The first predicted planting analysis result is compared with the data in the actual multi-source planting dataset during the last round of planting to generate the first error rate for each round of planting; Based on the first error rate, update the initial sample data weights of the multi-source planting sample dataset for each round; Repeat the following steps: Based on the updated sample data weights, input the planting data in the multi-source planting sample dataset of each round into the jujube planting optimization model generated by the last round of jujube planting analysis iteration, and obtain the new predicted planting analysis results of the jujube planting optimization model generated by the last round of jujube planting analysis iteration for predicting the multi-source planting dataset corresponding to each round of planting. The new predictive planting analysis results are compared with the data in the actual multi-source planting dataset during the last round of planting to generate a new error rate for each round of planting. Based on the new error rate, update the sample data weights of the multi-source planting sample dataset in each round until the preset number of repetitions is reached; Data from the multi-source planting sample dataset whose average change in sample data weights does not exceed a preset change rate threshold are used as the planting analysis supervision data.

[0011] In one possible implementation, updating the initial sample data weights of the multi-source planting sample dataset for each round based on the first error rate includes: Based on the first error rate, the initial sample data weights of the multi-source planting sample dataset in each round are updated using a strategy gradient method that maximizes expected returns. The step of updating the sample data weights of the multi-source planting sample dataset in each round according to the new error rate includes: Calculate the weighted error rate of the new error rate, and compare the magnitude of the weighted error rates after two consecutive updates of the sample data weights; Based on the relationship between the weighted error rates after two consecutive updates of the sample data weights, adjust the gradient direction and gradient amount in the gradient method for maximizing expected return; Based on the adjusted gradient change direction and gradient change amount in the strategy gradient method for maximizing expected return, the sample data weights of the multi-source planting sample dataset in each round are updated according to the sample data weights of the last round.

[0012] In one possible implementation, the jujube planting optimization model generated through the last round of jujube planting analysis iterations analyzes the multi-source planting dataset for each round to generate planting analysis supervision data, including: The data from the multi-source planting dataset corresponding to each round of planting is loaded as input data into the jujube planting optimization model generated by the last round of jujube planting analysis iteration, and the second prediction planting analysis result is obtained by the jujube planting optimization model generated by the last round of jujube planting analysis iteration predicting the multi-source planting dataset corresponding to each round of planting. The second predicted planting analysis result corresponding to each round of planting is compared with the data in the actual multi-source planting dataset during that round of planting to generate the second error rate corresponding to each round of planting. Planting analysis and supervision data are generated by removing the multi-source planting datasets of the rounds to be removed from the multi-source planting datasets whose second error rate exceeds the preset error rate threshold.

[0013] In one possible implementation, the multi-source planting dataset includes, but is not limited to, data in the following dimensions: soil data, climate data, pest and disease data, fertilization data, irrigation data, pesticide application data, and jujube fruit growth data.

[0014] According to a second aspect of the present disclosure, a jujube fruit planting system based on multi-source data analysis is provided, comprising: Processor and computer-readable storage medium; The computer-readable storage medium stores machine-executable instructions, which, when executed by the processor, implement the jujube fruit planting method based on multi-source data analysis as described in any one of the first aspects.

[0015] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects: By acquiring multi-source planting datasets corresponding to multiple planting cycles of the jujube trees to be planted, this dataset serves as the multi-source sample set used in this iteration of jujube planting data analysis. This effectively utilizes information from historical planting data, improving the accuracy of model construction and optimization. The jujube planting optimization model generated in the final iteration of jujube planting analysis is used to analyze the multi-source planting datasets from each cycle, generating supervised planting analysis data. This not only considers the characteristics of the current cycle's data but also improves the model's generalization ability and accuracy by comparing the differences and correlations between data from multiple cycles.

[0016] Furthermore, based on the supervised planting analysis data, a target jujube planting optimization model was determined from the jujube planting optimization models generated in each round of corresponding jujube planting analysis iterations. This ensured the optimal match between the selected model and the current planting conditions, thereby improving the accuracy and applicability of the planting recommendations. The target planting dataset used for configuring the target jujube planting optimization model analysis iterations was determined from the multi-source planting datasets corresponding to each round based on the supervised planting analysis data. This refined the dataset selection process, enabling the model to be trained and optimized on more representative data.

[0017] Furthermore, by loading the planting data from the multi-source planting datasets corresponding to each round of planting into the jujube planting optimization model generated by the last round of jujube planting analysis iteration, a global planting supervision dataset is generated. This dataset is then compared with the multi-source planting datasets corresponding to the target jujube variety to be planted to generate an extended planting supervision dataset, thereby enhancing the model's generalization ability and adaptability to similar planting conditions.

[0018] Finally, by inputting the extended planting supervision dataset and the target planting dataset into the target jujube planting optimization model, the planting analysis results were obtained. These results include key information such as the timing of fertilization and irrigation in this planting cycle, as well as the amount of fertilizer applied and the amount of water used for each irrigation. This information provides scientific and precise planting recommendations, which help improve the yield and quality of jujubes while reducing the risk of resource waste and environmental pollution.

[0019] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0020] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0021] Figure 1 This is a flowchart illustrating a jujube fruit planting method and system based on multi-source data analysis, according to an exemplary embodiment.

[0022] Figure 2 This is an implementation illustrated according to an exemplary embodiment. Figure 1 The flowchart for step S12.

[0023] Figure 3 This is an implementation illustrated according to an exemplary embodiment. Figure 2 The flowchart for step S123. Detailed Implementation

[0024] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0025] Figure 1 This is a flowchart illustrating a jujube cultivation method based on multi-source data analysis, according to an exemplary embodiment. See also... Figure 1 As shown, the method includes: Step S11: Obtain the multi-source planting dataset for each round of planting under the multi-round planting situation corresponding to the jujube fruit to be planted, and use it as the multi-source planting sample dataset called in this iteration of jujube planting data analysis; The multi-source planting dataset contains data from various collections related to jujube cultivation, potentially including soil, climate, and planting technology data for each planting cycle. Iterative calls refer to datasets that are used and updated multiple times during data analysis or model training.

[0026] In this embodiment of the disclosure, various data related to jujube cultivation are collected from multiple planting rounds. This data will serve as the basis for subsequent analysis and model training. For example, a jujube farm may have conducted multiple planting rounds over the past few years, recording data such as soil moisture, temperature, fertilizer application, and irrigation amount for each round. This data is integrated to form a multi-source planting dataset for subsequent analysis.

[0027] Step S12: Analyze the multi-source planting sample dataset of each round using the jujube planting optimization model generated by the last round of jujube planting analysis iteration, generate planting analysis supervision data, and determine the target jujube planting optimization model from the jujube planting optimization models generated by the corresponding jujube planting analysis iteration in each round based on the planting analysis supervision data. Among these, the planting optimization model is a model built based on data and algorithms, used to optimize various parameters and decisions during the planting process. Examples include neural network models and convolutional neural network models. Supervised data is data used to evaluate and optimize model performance.

[0028] In this embodiment, the planting optimization model generated in the previous iteration is used to analyze the multi-source planting sample dataset of the current round, generating supervisory data. Based on this data, the best-performing target planting optimization model is determined from multiple iterations. For example, suppose that in the third round of planting analysis, the current data is analyzed using the optimization model generated in the second round, and it is found that adjusting certain planting parameters can significantly increase jujube yield. These analysis results serve as supervisory data to help determine which round of optimization model performed best.

[0029] Step S13: Based on the planting analysis supervision data, determine the target planting dataset used to configure the target jujube planting optimization model analysis iteration from the multi-source planting sample dataset corresponding to each round; The supervised planting data is the data label used to guide model training, and it typically corresponds to the model's prediction target. The target planting dataset is the dataset used to configure and optimize the target planting optimization model.

[0030] In this embodiment of the disclosure, based on the supervised data, the most suitable dataset for configuring and optimizing the target planting optimization model is selected from the multi-source datasets of multiple rounds of planting. The multi-source planting sample datasets of each round are evaluated to determine which data samples best match the planting analysis supervised data, i.e., those with the highest accuracy and completeness. By comparing the consistency between the data samples and the supervised data, the data that best reflects the actual planting situation can be selected.

[0031] Furthermore, after selecting highly accurate data, it is necessary to analyze its representativeness. That is, whether this data can comprehensively reflect the growth of jujubes under different planting conditions. Selecting representative data ensures that the model has greater generalization ability in subsequent analyses and can adapt to various planting environments.

[0032] Based on the analysis results of the above two steps, a target dataset can be identified. This dataset not only has high accuracy but also good representativeness. The selection of the target dataset is crucial for subsequent model training and optimization, as it directly affects the quality of model learning and the accuracy of predictions.

[0033] For example, suppose a jujube cultivation research project has conducted multiple rounds of planting experiments and collected data on various aspects, including soil conditions, climate conditions, and planting techniques. Now, researchers hope to use this data to optimize a planting optimization model in order to more accurately predict the growth and yield of jujubes.

[0034] This process begins by using existing planting analysis and monitoring data (such as historical yield data and fruit quality data) to assess the accuracy and representativeness of the planting sample data from each round. Through comparison and analysis, they found that the data from certain rounds were highly consistent with the monitoring data and could comprehensively reflect the growth of jujubes under different planting conditions. These highly accurate and representative data are then aggregated to form a "target planting dataset." This dataset will be used for subsequent model training and optimization to improve the model's ability to predict and guide future planting activities. In this way, planting strategies can be adjusted more precisely, potentially improving the yield and quality of jujubes.

[0035] Step S14: Filter the planting data in the multi-source planting sample dataset corresponding to each round of planting to generate a global planting supervision dataset; Among them, the global planting supervision dataset is a dataset selected from multiple rounds of planting data for global supervision and model optimization.

[0036] In this embodiment, multi-round planting data is filtered to generate a global supervised dataset, which will be used for subsequent model comparison and optimization. After selecting the best-performing data from the multi-round planting data, a global planting supervised dataset is formed. This dataset contains best practices under various planting conditions and will be used to guide future planting activities.

[0037] Step S15: Compare the global planting supervision dataset with the multi-source planting dataset corresponding to the target jujube of the same variety as the jujube to be planted, and generate an extended planting supervision dataset. The target jujube is a jujube whose planting conditions are similar to those of the jujube to be planted. The expanded planting supervision dataset is a dataset generated by comparing the global supervision dataset with other jujube datasets of the same variety to be planted, and is used to further expand and optimize the supervision data.

[0038] In this embodiment of the disclosure, a broader and more representative supervised dataset can be generated by comparing the global supervised dataset with other jujube datasets under similar planting conditions. For example, suppose there is a dataset from other jujube farms with the same jujube variety to be planted and similar planting conditions. Comparing this dataset with the global supervised dataset can generate an expanded planting supervised dataset, thereby providing a more comprehensive understanding of best practices under different planting conditions.

[0039] Step S16: Input the extended planting supervision dataset and the target planting dataset as input data into the target jujube planting optimization model to obtain the planting analysis results output by the target jujube planting optimization model, and obtain the jujube planting optimization model generated by the iteration of this round of jujube planting analysis. The planting analysis results include the fertilization time node, irrigation time node, and pesticide spraying time node of this round of planting, as well as the amount of fertilizer applied, the amount of water applied, and the amount of pesticide sprayed for each fertilization, so that fertilization can be carried out according to the fertilization time node and the amount of fertilizer applied, irrigation can be carried out according to the irrigation time node and the amount of water applied, and pesticide spraying can be carried out according to the pesticide spraying time node and the amount of pesticide sprayed to complete the jujube planting.

[0040] In this embodiment, an extended supervised dataset and a target planting dataset are input into a target planting optimization model to obtain the optimal planting plan. This plan will guide future planting activities to maximize the yield and quality of jujube fruit. After inputting the extended supervised dataset and the target planting dataset, the planting optimization model outputs the optimal fertilization, irrigation, and pesticide application plan for this round of planting. Planting activities can be carried out according to this plan, which is expected to achieve higher yields and better jujube fruit quality.

[0041] The method described above acquires multi-source planting datasets corresponding to multiple planting cycles of jujubes to be planted, serving as the multi-source sample set used in this iteration of jujube planting data analysis. This effectively utilizes information from historical planting data, improving the accuracy of model construction and optimization. The jujube planting optimization model generated in the last iteration of jujube planting analysis analyzes the multi-source planting datasets from each cycle, generating planting analysis supervision data. This approach not only considers the characteristics of the current cycle's data but also improves the model's generalization ability and accuracy by comparing the differences and correlations between data from multiple cycles.

[0042] Furthermore, based on the supervised planting analysis data, a target jujube planting optimization model was determined from the jujube planting optimization models generated in each round of corresponding jujube planting analysis iterations. This ensured the optimal match between the selected model and the current planting conditions, thereby improving the accuracy and applicability of the planting recommendations. The target planting dataset used for configuring the target jujube planting optimization model analysis iterations was determined from the multi-source planting datasets corresponding to each round based on the supervised planting analysis data. This refined the dataset selection process, enabling the model to be trained and optimized on more representative data.

[0043] Furthermore, by loading the planting data from the multi-source planting datasets corresponding to each round of planting into the jujube planting optimization model generated by the last round of jujube planting analysis iteration, a global planting supervision dataset is generated. This dataset is then compared with the multi-source planting datasets corresponding to the target jujube variety to be planted to generate an extended planting supervision dataset, thereby enhancing the model's generalization ability and adaptability to similar planting conditions.

[0044] Finally, by inputting the extended planting supervision dataset and the target planting dataset into the target jujube planting optimization model, the planting analysis results were obtained. These results include key information such as the timing of fertilization and irrigation in this planting cycle, as well as the amount of fertilizer applied and the amount of water used for each irrigation. This information provides scientific and precise planting recommendations, which help improve the yield and quality of jujubes while reducing the risk of resource waste and environmental pollution.

[0045] In one possible implementation, see Figure 2 As shown, step S12, determining the target jujube planting optimization model from the jujube planting optimization models generated by the jujube planting analysis iterations corresponding to each round based on the planting analysis supervision data, includes: Step S121: Connect the jujube planting optimization models generated by the iterative analysis of jujube planting in each round according to the time series order to obtain the model sequence; The model sequence refers to a chain of models formed by connecting the jujube planting optimization models generated from each round of jujube planting analysis iterations in their corresponding time series order. This model chain can reflect the improvement and optimization process of the models in different rounds of planting iterations.

[0046] In this embodiment, all iterative jujube planting optimization models are linked together to form a complete model sequence. This sequence reflects the change process of the model from the initial state to the final optimized state, providing a basis for subsequent target model selection. For example, suppose there are 5 rounds of planting analysis iterations, and each iteration generates a jujube planting optimization model (denoted as Model1, Model2, ..., Model5). Step S121 connects these 5 models in chronological order (i.e., Model1→Model2→...→Model5) to form a model sequence.

[0047] Step S122: Pre-set the sample data weights for each data point in the planting analysis and monitoring data; Here, sample data weight is a parameter used to measure the importance of each data sample during model training. In this embodiment of the disclosure, each data sample in the planting analysis supervision data is assigned a weight, which will affect the degree of importance the model attaches to the data during training.

[0048] Different data samples may have varying degrees of impact on model training. By setting weights for the data samples, the model can focus more on those samples that have a greater impact on model performance during training. For example, in supervised planting analysis data, some data samples may contain very important information (such as planting data under extreme weather conditions), and these samples should have a greater impact on model training. Therefore, higher weights can be assigned to these important data samples.

[0049] Step S123: Based on the pre-set sample data weights, the planting analysis supervision data, and the planting labeling results for the planting analysis supervision data, perform weak learning training in the order of each jujube planting optimization model in the model sequence to obtain the corresponding predicted planting results, weak learning error rate, and updated sample data weights. Weak learning, a concept within ensemble learning, refers to training data using a simple learning algorithm (such as decision stumps or linear classifiers). The resulting model is called a weak classifier or weak learner. Weak learners typically have lower performance, but combining multiple weak learners through specific strategies can yield a powerful strong learner.

[0050] Predicting planting outcomes refers to the predicted values ​​about planting results given by the model after analyzing the planting dataset through the jujube planting optimization model, such as the timing of fertilization, the timing of irrigation, the amount of fertilizer, and the amount of irrigation.

[0051] The weak learning error rate refers to the degree of difference between the prediction results of the weak learner on the training dataset and the actual results during the weak learning training process. The smaller the error rate, the better the performance of the weak learner.

[0052] In this embodiment, weak learning training can yield prediction results and error rates for multiple weak learners (i.e., jujube planting optimization models). Simultaneously, based on the prediction results and error rates of each weak learner, the sample data weights can be updated, enabling subsequent weak learners to focus more on samples misclassified by previous learners. For example, Model1 is used to perform weak learning training on supervised planting analysis data to obtain predicted planting results and a weak learning error rate. Then, the sample data weights are updated based on the prediction results and error rate of Model1. Next, weak learning training is performed using the updated sample data weights and Model2, and the above process is repeated until Model5.

[0053] Step S124: The jujube planting optimization model with the smallest weak learning error rate is determined as the target jujube planting optimization model.

[0054] Among these methods, weak learners with better performance (i.e., lower error rates) are selected for combination. Therefore, the jujube planting optimization model with the lowest weak learning error rate is chosen as the target model. For example, assuming that Model3 achieves the lowest error rate during weak learning training, then Model3 is determined as the target jujube planting optimization model.

[0055] In this way, by connecting multiple rounds of planting optimization models through time series analysis, setting sample data weights, performing weak learning training and updating the weights, and finally selecting the model with the smallest error rate as the target model, the planting plan is precisely optimized, improving planting efficiency and fruit quality.

[0056] In one possible implementation, see Figure 3 As shown, step S123, which involves performing weak learning training according to the pre-set sample data weights, the planting analysis supervision data, and the planting labeling results for the planting analysis supervision data, in the order of each jujube planting optimization model in the model sequence, to obtain the corresponding predicted planting results, weak learning error rate, and updated sample data weights, includes: Step S1231: Input the planting analysis supervision data into the first jujube planting optimization model of the model sequence, so that the first jujube planting optimization model can be weakly trained according to the pre-set sample data weights to obtain the corresponding predicted planting results; In this embodiment of the disclosure, at the beginning of the weak learning training phase, the supervised planting analysis data (including historical planting data, weather data, soil data, etc.) is first input into the first jujube planting optimization model in the model sequence. The model is trained according to the pre-set sample data weights and outputs the corresponding predicted planting results. For example, suppose there is a dataset containing planting data over many years, and the corresponding planting labeling results (i.e., actual planting conditions). This data is input into the first model (Model1) in the model sequence. Model1 is trained according to the weights of each data sample and outputs a prediction of the planting results.

[0057] Step S1232: Compare the predicted planting results with the planting labeling results to obtain the weak learning error rate corresponding to the first jujube planting optimization model; In this embodiment of the disclosure, the predicted planting results output by Model1 are compared with the actual planting labeling results, and the degree of difference between the predicted results and the actual results is calculated to obtain the weak learning error rate of Model1. The predicted planting results (such as fertilization time, irrigation amount, etc.) output by Model1 are compared with the actual planting labeling results, and the difference between the two (such as error percentage) is calculated to obtain the weak learning error rate of Model1.

[0058] Step S1233: Update the preset sample data weights according to the weak learning error rate corresponding to the first jujube planting optimization model, and obtain the updated sample data weights corresponding to the next jujube planting optimization model connected to the first jujube planting optimization model in the model sequence. Specifically, the pre-set sample data weights are adjusted based on the weak learning error rate of Model1. Typically, the weights of samples misclassified by Model1 are increased so that subsequent models pay more attention to these samples. If Model1 makes a large error in predicting certain data samples, the weights of these samples are increased during subsequent training so that Model2 (the next model connected to Model1) can learn these samples better.

[0059] Step S1234: Repeat the following steps: input the planting analysis supervision data, the planting labeling results, and the updated sample data weights corresponding to the previous jujube planting optimization model into the next jujube planting optimization model in sequence for weak learning training, to obtain the corresponding predicted planting results, weak learning error rate, and updated sample data weights, until all jujube planting optimization models have completed weak learning training.

[0060] In this embodiment, the planting analysis supervision data, planting annotation results, and updated sample data weights are sequentially input into the subsequent jujube planting optimization model for training, according to the model sequence. Each model outputs a predicted planting result and a weak learning error rate, and updates the sample data weights based on the error rate. This process continues until all models have completed training.

[0061] For example, after Model 1 completes training and updates the sample data weights, the updated data is fed into Model 2 for training. Model 2 outputs the prediction results and error rate, and updates the sample data weights again. Then, this data is fed into Model 3 for training, and so on, until all models have completed training.

[0062] In this way, by using a sequential jujube cultivation optimization model for weak learning training, combined with sample data weight adjustments, the model can gradually focus on and optimize key cultivation factors. By iteratively updating sample weights and comparing prediction results, precise optimization of the cultivation plan is achieved, ultimately obtaining the best-performing jujube cultivation optimization model, significantly improving cultivation efficiency and fruit quality.

[0063] In one possible implementation, step S14, which involves filtering the planting data in the multi-source planting sample dataset corresponding to each round of planting to generate a global planting supervision dataset, includes: Expert knowledge labeling is performed on the planting data in the multi-source planting sample dataset corresponding to each round of planting to generate expert labels for the planting data in the multi-source planting sample dataset corresponding to each round of planting. Expert knowledge labeling refers to the process by which experts in the field evaluate and label data in a planting sample dataset based on their professional knowledge and experience, generating expert labels. Expert labels are identifiers or classifications that experts assign to data with specific meaning based on their understanding and analysis of the data, such as "appropriate fertilization amount" or "high risk of pests and diseases".

[0064] For example, based on the expertise and experience of domain experts, the sample data for each round of planting is evaluated and labeled with expert tags. For instance, the amount of fertilizer applied in one round might be labeled as "appropriate" by experts, while in another round it might be labeled as "excessive." Another example is examining data from multi-source planting sample datasets, such as soil pH, fertilizer application rate, and irrigation frequency, and labeling each data point with expert tags based on their experience, such as "appropriate soil pH" or "excessive fertilizer application."

[0065] Based on the expert labels, determine the conflict and consistency among the planting data in the multi-source planting sample dataset corresponding to each round of planting; Conflict refers to the phenomenon in multi-source datasets where differences in data source, collection method, or data quality may lead to contradictions or inconsistencies among data under the same or similar conditions. Consistency, on the other hand, refers to the degree of similarity or matching between data items within a dataset under the same or similar conditions, indicating the information or conclusions they provide.

[0066] The system analyzes the conflicts and consistency between data based on these expert labels. For example, under the same geographical and climatic conditions, the fertilizer application rates for two rounds of planting were labeled "appropriate" and "excessive" by experts, respectively, which presents a conflict. Simultaneously, the system also checks the consistency of these two rounds of data with other data to see if there is any other data supporting or contradicting these two labels.

[0067] If the conflict between any two planting data sets meets the preset conflict conditions, one of the two planting data sets is deleted based on the consistency between the two planting data sets and each of the other planting data sets. When a conflict is detected that meets preset conditions (such as conflicting data exceeding 10% of the total), the system will decide whether to delete or retain one of the data based on its consistency with other data. For example, if the label "over-fertilization" is supported by only a few data points, while the label "appropriate fertilization" is supported by the majority of data points, the system may delete the data for "over-fertilization".

[0068] A global planting supervision dataset is generated based on all the planting data that has not been deleted.

[0069] All data that was not deleted will be integrated into a global planting supervision dataset, which will serve as an important basis for subsequent planting analysis, model training, and planting optimization.

[0070] In this way, the accuracy of planting data is ensured through expert knowledge labeling, and conflicts and consistency between data are analyzed based on expert labels. When conflicts meet preset conditions, data consistency is assessed to select which data to retain or delete, ultimately generating a global planting supervision dataset. This effectively improves data quality, providing a reliable data foundation for subsequent planting analysis and optimization, thereby enhancing the scientific nature and accuracy of planting decisions.

[0071] In one possible implementation, step S15, which involves matching the global planting supervision dataset with the multi-source planting dataset corresponding to the target jujube of the same variety to be planted, to generate an extended planting supervision dataset, includes: The expert labels of each planting data in the global planting supervision dataset are matched with the expert labels of each planting data in the multi-source planting dataset corresponding to the target jujube of the same variety to be planted. The planting data in the multi-source planting dataset corresponding to the target jujube that do not achieve a match are used as the extended planting supervision dataset.

[0072] In this embodiment of the disclosure, the data in the global planting supervision dataset is compared with the multi-source planting dataset corresponding to the target jujube of the same variety to be planted by means of label matching, so as to find data items that can be expanded to the global planting supervision dataset.

[0073] For example, suppose there is a global planting supervision dataset containing planting data for various jujube varieties under different conditions, and all of this data has been calibrated by experts. Now we want to expand this dataset to better understand and optimize the planting conditions for the jujubes to be planted (let's say "jujube variety A").

[0074] There is a multi-source dataset containing planting data for "Jujube Variety A," originating from different planting environments, years, and techniques. To match these two datasets and generate an expanded planting supervision dataset, the following steps are performed: First, extract the expert labels (such as "suitable irrigation amount", "high disease risk", etc.) for all planting data related to "A variety of jujube" in the global planting supervision dataset. Then, traverse the multi-source planting datasets corresponding to the target jujube and compare the expert labels of the planting data therein with the labels in the global dataset.

[0075] Then, during the comparison process, it was found that some target jujube planting data did not have corresponding labels in the global dataset (i.e., no match was achieved). These unmatched data may be because they represent new planting conditions, new planting techniques, or previously unrecorded planting environments.

[0076] Finally, these unmatched data points were added to the expanded planting supervision dataset. Since these data belong to the same variety of "A-type jujube" to be planted, they are invaluable for understanding the growing conditions of this variety. Furthermore, because these data have no corresponding entries in the global dataset, their inclusion enriches the dataset's diversity and improves the accuracy of subsequent data analysis and model training.

[0077] In one possible implementation, step S12 involves analyzing the multi-source planting sample dataset from each round of planting using the jujube planting optimization model generated through the last round of jujube planting analysis iterations, to generate planting analysis supervision data, including: Set the initial sample data weights for each round of the multi-source planting sample dataset; In this embodiment of the disclosure, an initial sample data weight is set for each round of multi-source planting sample dataset. This weight is typically based on a preset criterion, such as all samples having equal weight, or being initially assigned according to the importance of the samples. The initial sample data weights are easy to adjust based on model performance during iteration.

[0078] Based on the initial sample data weights, the planting data in each round of the multi-source planting sample dataset is input into the jujube planting optimization model generated by the last round of jujube planting analysis iteration, and the first predicted planting analysis result of the jujube planting optimization model generated by the last round of jujube planting analysis iteration predicts the multi-source planting dataset corresponding to each round of planting. In this embodiment, based on the set initial sample data weights, the planting data from each round of multi-source planting sample dataset is input into the jujube planting optimization model generated by the final round of jujube planting analysis iterations. The model is then used to predict the loaded planting data, yielding preliminary prediction results.

[0079] The first predicted planting analysis result is compared with the data in the actual multi-source planting dataset during the last round of planting to generate the first error rate for each round of planting; In this embodiment, the first predicted planting analysis result obtained by the model is compared with the data in the actual multi-source planting dataset during the last round of planting. The difference between the two is calculated to generate a first error rate for each round of planting. This allows for the evaluation of the accuracy of the model's predictions.

[0080] Based on the first error rate, update the initial sample data weights of the multi-source planting sample dataset for each round; In this embodiment of the disclosure, the initial sample data weights of the multi-source planting sample dataset in each round are updated based on the calculated first error rate. Typically, the weights of sample data with higher error rates are decreased, while the weights of sample data with lower error rates are increased. By adjusting the sample data weights, the model focuses more on samples with smaller prediction errors in subsequent iterations, thereby improving the model's prediction performance.

[0081] Repeat the following steps: Based on the updated sample data weights, input the planting data in the multi-source planting sample dataset of each round into the jujube planting optimization model generated by the last round of jujube planting analysis iteration, and obtain the new predicted planting analysis results of the jujube planting optimization model generated by the last round of jujube planting analysis iteration for predicting the multi-source planting dataset corresponding to each round of planting. The new predictive planting analysis results are compared with the data in the actual multi-source planting dataset during the last round of planting to generate a new error rate for each round of planting. Based on the new error rate, update the sample data weights of the multi-source planting sample dataset in each round until the preset number of repetitions is reached; In this embodiment of the disclosure, before reaching a preset number of repetitions, the following steps are repeated: planting data is input into the model according to the updated sample data weights to obtain new prediction results; the new prediction results are compared with the actual data to generate a new error rate; and the sample data weights are updated again based on the new error rate. Through multiple iterations, the predictive performance of the model is gradually optimized, and sample data with smaller prediction errors and higher stability are selected.

[0082] Data from the multi-source planting sample dataset whose average change in sample data weights does not exceed a preset change rate threshold are used as the planting analysis supervision data.

[0083] After reaching the preset number of repetitions, the average change in the weights of the sample data is checked. If the average change does not exceed the preset rate of change threshold, the model is considered stable, and a batch of sample data with high stability and small prediction errors is selected. The planting analysis monitoring data not only has small prediction errors but also good stability, and can more accurately reflect the actual planting situation.

[0084] In one possible implementation, updating the initial sample data weights of the multi-source planting sample dataset for each round based on the first error rate includes: Based on the first error rate, the initial sample data weights of the multi-source planting sample dataset in each round are updated using a strategy gradient method that maximizes expected returns. In this embodiment of the disclosure, a gradient is calculated that indicates how the initial sample data weights should be adjusted to maximize the expected return (i.e., minimize the error rate). This gradient can be obtained by estimating the derivative of the error rate with respect to the weights. The direction of the gradient indicates how the weights should be adjusted to reduce the error rate. Specifically, for samples that cause large errors, their weights are reduced; while for samples that are predicted accurately, their weights are increased.

[0085] The learning rate is then chosen, which determines the step size for weight updates. A large learning rate may cause weight updates to be too fast, thus skipping the optimal solution; while a small learning rate may result in slow learning. Using the calculated gradient and learning rate, the weights of the initial sample data can be updated. This process is usually accomplished by adding the weights to the product of the gradient and the learning rate.

[0086] The step of updating the sample data weights of the multi-source planting sample dataset in each round according to the new error rate includes: Calculate the weighted error rate of the new error rate, and compare the magnitude of the weighted error rates after two consecutive updates of the sample data weights; In this embodiment of the disclosure, to more comprehensively evaluate the overall performance of the model, a weighted error rate can be calculated, where the error rate of each sample is weighted according to its weight. Thus, samples with larger weights contribute more to the weighted error rate.

[0087] Based on the relationship between the weighted error rates after two consecutive updates of the sample data weights, adjust the gradient direction and gradient amount in the gradient method for maximizing expected return; In this embodiment of the disclosure, a new weighted error rate is calculated after each update of the sample data weights. By comparing the weighted error rates after two consecutive updates, it can be determined whether the model performance is improving.

[0088] Based on the adjusted gradient change direction and gradient change amount in the strategy gradient method for maximizing expected return, the sample data weights of the multi-source planting sample dataset in each round are updated according to the sample data weights of the last round.

[0089] In this embodiment of the disclosure, if the new weighted error rate is lower than the previous one, it indicates that the model's performance is improving, and we can continue to adjust the weights along the current gradient direction; if the new weighted error rate is higher, it indicates that the current direction may not be optimal, and we need to adjust the direction and / or amount of gradient change.

[0090] Based on the adjusted gradient direction and magnitude, we can update the sample weights of each round of multi-source planting datasets using the weights from the last iteration. This allows us to adjust the sample weights more precisely in subsequent iterations, aiming to further improve the model's predictive performance.

[0091] In one possible implementation, step S12 involves analyzing the multi-source planting dataset for each round of planting using the jujube planting optimization model generated through the last round of jujube planting analysis iterations, generating planting analysis supervision data, including: The data from the multi-source planting dataset corresponding to each round of planting is loaded as input data into the jujube planting optimization model generated by the last round of jujube planting analysis iteration, and the second prediction planting analysis result is obtained by the jujube planting optimization model generated by the last round of jujube planting analysis iteration predicting the multi-source planting dataset corresponding to each round of planting. In this embodiment, data from the multi-source planting dataset corresponding to each round of planting is loaded as input data into the jujube planting optimization model generated by the last round of jujube planting analysis iterations. This step is to use the model, which has undergone multiple rounds of iterative optimization, to evaluate the predictive accuracy of historical planting data. The loading process may involve data preprocessing steps, such as data cleaning and format conversion, to ensure that the data meets the model's input requirements. Then, the data is input into the model according to the required format and structure.

[0092] After the data is loaded, the model makes predictions on the multi-source planting dataset corresponding to each round of planting, generating a second predictive planting analysis result. These prediction results reflect the model's judgment on the current round of planting. Based on the input planting data, the model uses its learned knowledge and patterns to predict the planting results. The prediction results may include multiple aspects such as jujube yield, quality, and growth cycle.

[0093] The second predicted planting analysis result corresponding to each round of planting is compared with the data in the actual multi-source planting dataset during that round of planting to generate the second error rate corresponding to each round of planting. In this embodiment, the second predicted planting analysis result corresponding to each round of planting is compared with the data in the actual multi-source planting dataset during that round of planting to evaluate the model's predictive accuracy. The comparison process may involve the calculation of multiple indicators, such as error rate and accuracy. By comparing the predicted results with the actual data, we can understand the model's performance in each round of planting.

[0094] Furthermore, based on the comparison results, a second error rate is calculated for each round of planting. This error rate reflects the magnitude of the model's prediction error in that round of planting. The calculation of the error rate may involve various methods, such as mean squared error and absolute error. By calculating the second error rate for each round of planting, we can understand the model's predictive performance in different rounds of planting.

[0095] Planting analysis and supervision data are generated by removing the multi-source planting datasets of the rounds to be removed from the multi-source planting datasets whose second error rate exceeds the preset error rate threshold.

[0096] In this embodiment, datasets from a single planting round whose second error rate exceeds a preset error rate threshold are removed from the multi-source planting dataset. This step ensures the accuracy and reliability of subsequent analysis and avoids adverse effects from data with large errors. First, a reasonable error rate threshold is set. Then, the second error rate for each planting round is iterated, and datasets exceeding the threshold are marked. Finally, these marked datasets are removed from the multi-source planting dataset to generate the final planting analysis supervision data.

[0097] Through the above steps, the jujube planting optimization model generated in the last round of jujube planting analysis iteration can be effectively used to analyze multi-source planting datasets and generate accurate planting analysis and supervision data.

[0098] In one possible implementation, the multi-source planting dataset includes, but is not limited to, data in the following dimensions: soil data, climate data, pest and disease data, fertilization data, irrigation data, pesticide application data, and jujube fruit growth data.

[0099] Soil data includes information about various properties, components, and conditions of the soil. This data can include soil type (such as sandy soil, clay, loam, etc.), soil fertility (such as the content of elements like nitrogen, phosphorus, and potassium), soil pH, soil moisture, and soil structure. This information is crucial for understanding which crops are suitable for planting in the soil and for determining fertilization and irrigation strategies.

[0100] Climate data describes the meteorological conditions of a region or specific location. This data typically includes temperature, humidity, rainfall, wind speed, wind direction, and sunshine duration. Climate data is crucial for predicting crop growth cycles and yields, and for developing appropriate agricultural management practices.

[0101] Pest and disease data refers to information about the occurrence of pests and diseases affecting jujubes. This data may include the types of pests and diseases, the time of occurrence, the severity of damage, and the effectiveness of control measures. By collecting and analyzing this data, farmers and agricultural experts can understand the trends in pest and disease outbreaks and thus develop effective control measures.

[0102] Fertilization data refers to information about fertilization practices for jujubes. This data may include fertilization timing, amount, type of fertilizer (such as nitrogen, phosphorus, and potassium), and application method (such as broadcasting, furrow application, and hole application). Collecting and analyzing fertilization data helps optimize fertilization strategies and improve crop yield and quality.

[0103] Irrigation data refers to information about crop irrigation. This data may include irrigation time, irrigation volume, irrigation method (such as drip irrigation, sprinkler irrigation, furrow irrigation, etc.), soil moisture, etc. By collecting and analyzing irrigation data, farmers can understand the water requirements of their crops, thereby developing reasonable irrigation plans to ensure normal crop growth.

[0104] Pesticide application data refers to information about pesticide use on crops. This data may include pesticide type, application time, application rate, application method (e.g., spraying, smearing), and control effects. Collecting and analyzing pesticide application data helps farmers and agricultural experts understand pesticide use, enabling them to develop reasonable pesticide application plans, reduce pesticide residues, and ensure the quality and safety of agricultural products.

[0105] Jujube fruit growth data refers to information about the growth of jujube trees and the development of their fruit. This data may include the tree's growth rate, tree height, canopy size, leaf color, fruit size, fruit color, and fruit firmness. By analyzing this data, farmers and agricultural experts can understand the growth status of the jujube trees and the development of the fruit, thereby developing appropriate management measures to improve the yield and quality of the jujube fruit.

[0106] This disclosure also provides a jujube planting system based on multi-source data analysis, including: Processor and computer-readable storage medium; The computer-readable storage medium stores machine-executable instructions, which, when executed by the processor, implement the jujube planting method based on multi-source data analysis as described in any of the foregoing embodiments.

[0107] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of this disclosure. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0108] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for cultivating jujube fruit based on multi-source data analysis, characterized in that, include: Obtain the multi-source planting dataset for each round of planting under the multiple planting conditions corresponding to the jujube fruit to be planted, and use it as the multi-source planting sample dataset called in this iteration of jujube planting data analysis; The jujube planting optimization model generated by the last round of jujube planting analysis iteration is used to analyze the multi-source planting sample dataset of each round, generate planting analysis supervision data, and determine the target jujube planting optimization model from the jujube planting optimization models generated by the corresponding jujube planting analysis iteration in each round based on the planting analysis supervision data. Based on the planting analysis and supervision data, the target planting dataset used for configuring the target jujube planting optimization model analysis iteration is determined from the multi-source planting sample dataset corresponding to each round; The planting data in the multi-source planting sample dataset corresponding to each round of planting is filtered to generate a global planting supervision dataset. The global planting supervision dataset is compared with the multi-source planting dataset corresponding to the target jujube of the same variety as the jujube to be planted, and an extended planting supervision dataset is generated. The target jujube is a jujube with planting conditions similar to those of the jujube to be planted. The extended planting supervision dataset and the target planting dataset are input into the target jujube planting optimization model to obtain the planting analysis results output by the target jujube planting optimization model, and the jujube planting optimization model generated by the iteration of this round of jujube planting analysis is obtained. The planting analysis results include the fertilization time node, irrigation time node, and pesticide spraying time node of this round of planting, as well as the amount of fertilizer applied, the amount of water applied, and the amount of pesticide sprayed for each fertilization, so that fertilization can be carried out according to the fertilization time node and the amount of fertilizer applied, irrigation can be carried out according to the irrigation time node and the amount of water applied, and pesticide spraying can be carried out according to the pesticide spraying time node and the amount of pesticide sprayed to complete the jujube planting. The jujube planting optimization model, generated through the final round of jujube planting analysis iterations, analyzes the multi-source planting sample dataset for each round to generate planting analysis supervision data, including: Set the initial sample data weights for each round of the multi-source planting sample dataset; Based on the initial sample data weights, the planting data in each round of the multi-source planting sample dataset is input into the jujube planting optimization model generated by the last round of jujube planting analysis iteration, and the first predicted planting analysis result of the jujube planting optimization model generated by the last round of jujube planting analysis iteration predicts the multi-source planting dataset corresponding to each round of planting. The first predicted planting analysis result is compared with the data in the actual multi-source planting dataset during the last round of planting to generate the first error rate for each round of planting; Based on the first error rate, update the initial sample data weights of the multi-source planting sample dataset for each round; Repeat the following steps: Based on the updated sample data weights, input the planting data in the multi-source planting sample dataset of each round into the jujube planting optimization model generated by the last round of jujube planting analysis iteration, and obtain the new predicted planting analysis results of the jujube planting optimization model generated by the last round of jujube planting analysis iteration for predicting the multi-source planting dataset corresponding to each round of planting. The new predictive planting analysis results are compared with the data in the actual multi-source planting dataset during the last round of planting to generate a new error rate for each round of planting. Based on the new error rate, update the sample data weights of the multi-source planting sample dataset in each round until the preset number of repetitions is reached; Data from the multi-source planting sample dataset whose average change in sample data weights does not exceed a preset change rate threshold are used as the planting analysis supervision data. or The jujube planting optimization model generated through the last round of jujube planting analysis iterations analyzes the multi-source planting dataset for each round to generate planting analysis supervision data, including: The data from the multi-source planting dataset corresponding to each round of planting is loaded as input data into the jujube planting optimization model generated by the last round of jujube planting analysis iteration, and the second prediction planting analysis result is obtained by the jujube planting optimization model generated by the last round of jujube planting analysis iteration predicting the multi-source planting dataset corresponding to each round of planting. The second predicted planting analysis result corresponding to each round of planting is compared with the data in the actual multi-source planting dataset during that round of planting to generate the second error rate corresponding to each round of planting. Planting analysis and supervision data are generated by removing the multi-source planting datasets of the rounds to be removed from the multi-source planting datasets whose second error rate exceeds the preset error rate threshold.

2. The jujube fruit planting method based on multi-source data analysis according to claim 1, characterized in that, The step of determining the target jujube planting optimization model from the jujube planting optimization models generated by the corresponding iterative jujube planting analysis in each round based on the planting analysis and supervision data includes: The jujube planting optimization models generated by the iterative analysis of jujube planting in each round are connected in time sequence to obtain the model sequence; The sample data weights of each data point in the planting analysis and monitoring data are pre-set; Based on the pre-set sample data weights, the planting analysis supervision data, and the planting labeling results for the planting analysis supervision data, weak learning training is performed in the order of each jujube planting optimization model in the model sequence to obtain the corresponding predicted planting results, weak learning error rate, and updated sample data weights. The jujube planting optimization model with the smallest weak learning error rate is determined as the target jujube planting optimization model.

3. The jujube planting method based on multi-source data analysis according to claim 2, characterized in that, The step of performing weak learning training according to the pre-set sample data weights, the planting analysis and supervision data, and the planting labeling results for the planting analysis and supervision data, in the order of each jujube planting optimization model in the model sequence, to obtain the corresponding predicted planting results, weak learning error rate, and updated sample data weights, includes: The planting analysis and supervision data are input into the first jujube planting optimization model in the model sequence, so that the first jujube planting optimization model is trained by weak learning according to the pre-set sample data weights to obtain the corresponding predicted planting results. The predicted planting results are compared with the planting labeling results to obtain the weak learning error rate corresponding to the first jujube planting optimization model; The pre-set sample data weights are updated based on the weak learning error rate corresponding to the first jujube planting optimization model, so as to obtain the updated sample data weights corresponding to the next jujube planting optimization model connected to the first jujube planting optimization model in the model sequence. Repeat the following steps: input the planting analysis supervision data, the planting labeling results, and the updated sample data weights corresponding to the previous jujube planting optimization model into the next jujube planting optimization model in sequence for weak learning training, to obtain the corresponding predicted planting results, weak learning error rate, and updated sample data weights, until all jujube planting optimization models have completed weak learning training.

4. The jujube planting method based on multi-source data analysis according to claim 1, characterized in that, The step of filtering the planting data in the multi-source planting sample dataset corresponding to each round of planting to generate a global planting supervision dataset includes: Expert knowledge labeling is performed on the planting data in the multi-source planting sample dataset corresponding to each round of planting to generate expert labels for the planting data in the multi-source planting sample dataset corresponding to each round of planting. Based on the expert labels, determine the conflict and consistency among the planting data in the multi-source planting sample dataset corresponding to each round of planting; If the conflict between any two planting data sets meets the preset conflict conditions, one of the two planting data sets is deleted based on the consistency between the two planting data sets and each of the other planting data sets. A global planting supervision dataset is generated based on all the planting data that has not been deleted.

5. The jujube planting method based on multi-source data analysis according to claim 4, characterized in that, The step of matching the global planting supervision dataset with the multi-source planting dataset corresponding to the target jujube of the same variety to be planted, to generate an extended planting supervision dataset, includes: The expert labels of each planting data in the global planting supervision dataset are matched with the expert labels of each planting data in the multi-source planting dataset corresponding to the target jujube of the same variety to be planted. The planting data in the multi-source planting dataset corresponding to the target jujube that do not achieve a match are used as the extended planting supervision dataset.

6. The jujube fruit planting method based on multi-source data analysis according to claim 1, characterized in that, The step of updating the initial sample data weights of the multi-source planting sample dataset in each round according to the first error rate includes: Based on the first error rate, the initial sample data weights of the multi-source planting sample dataset in each round are updated using a strategy gradient method that maximizes expected returns. The step of updating the sample data weights of the multi-source planting sample dataset in each round according to the new error rate includes: Calculate the weighted error rate of the new error rate, and compare the magnitude of the weighted error rates after two consecutive updates of the sample data weights; Based on the relationship between the weighted error rates after two consecutive updates of the sample data weights, adjust the gradient direction and gradient amount in the gradient method for maximizing expected return; Based on the adjusted gradient change direction and gradient change amount in the strategy gradient method for maximizing expected return, the sample data weights of the multi-source planting sample dataset in each round are updated according to the sample data weights of the last round.

7. The jujube cultivation method based on multi-source data analysis according to any one of claims 1-6, characterized in that, The multi-source planting dataset includes, but is not limited to, data in the following dimensions: soil data, climate data, pest and disease data, fertilization data, irrigation data, pesticide application data, and jujube growth data.

8. A jujube fruit planting system based on multi-source data analysis, characterized in that, include: Processor and computer-readable storage medium; The computer-readable storage medium stores machine-executable instructions, which, when executed by the processor, implement the jujube fruit planting method based on multi-source data analysis as described in any one of claims 1-7.

Images