Intelligent agricultural data analysis method and system based on machine learning

Abstract

The invention relates to the technical field of data processing, in particular to an intelligent agricultural data analysis method and system based on machine learning. Comprising the following steps: S1, collecting agricultural analysis data, and constructing a data sample of the agricultural analysis data; based on the data sample, dividing data items according to the time sequence of crop growth to obtain a division result; obtaining a complete data sample based on the division result; s2, based on the data items, dividing the data items into similar data items and heterogeneous data items according to functions of the data items; based on the data sample, extracting time interval data of the data item; and based on the time interval data, performing simplification processing on the data item to obtain a simplification processing result. According to growth cycle classification data, a compression strategy is dynamically adjusted, sensor data processing is optimized, redundancy is reduced through comparison and anomaly detection, data management efficiency is improved, and accurate support is provided for agricultural decision making.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular, to a method and system for analyzing smart agriculture data based on machine learning. Background Art

[0002] In recent years, with the in-depth application of Internet of Things and sensor technologies in the agricultural field, time series data (such as soil temperature and humidity, fertilization amount, crop growth indicators) has shown exponential growth. However, existing data processing methods face the following core challenges in agricultural scenarios: Redundant data storage burden: Traditional compression methods rely on fixed thresholds or static rules and cannot dynamically identify the similarity of adjacent data items (such as the same temperature value collected continuously multiple times), resulting in a large accumulation of duplicate data and significant waste of storage resources; Computational efficiency bottleneck: Existing technologies uniformly process all data, ignoring temporal features (such as data fluctuation frequency, interval density differences). For example, when the sensor is normal, the data generation interval is stable, and when it is abnormal, the interruption is chaotic. However, existing methods do not differentiate the processing, resulting in ineffective calculations and response delays; Lack of dynamic adaptability: General compression algorithms (such as Huffman coding) do not combine the functional attributes of data items (such as single-functional parameters and composite-functional parameters), making it difficult to achieve efficient streamlining while retaining key features (such as the correlation between soil humidity and fertilization amount during the fertilization period), resulting in reduced interpretability of the compressed data or loss of key information.

[0003] Although existing research attempts to optimize the data volume through clustering or feature selection, the rule design lacks the combined application of dynamic analysis of time intervals and functional classification (such as separately processing single parameters and multi-parameter combinations), resulting in the disconnection between the compression strategy and the actual needs of agricultural scenarios. For example, it does not distinguish between the normal state of the sensor (high-frequency stable data can be compressed) and the abnormal state (low-frequency noise needs to be simplified), nor does it mine the mutation points at key growth stages through temporal correlation (such as comparing data during the sowing period and the harvesting period). Summary of the Invention

[0004] In order to overcome the shortcoming of low efficiency in processing agricultural temporal data, the present invention provides a method and system for analyzing smart agriculture data based on machine learning.

[0005] The technical implementation solution of the present invention is: A method for analyzing smart agriculture data based on machine learning, comprising the following steps: S1: Collect agricultural analysis data and construct a data sample of the agricultural analysis data; Based on the data sample, divide data items according to the chronological order of crop growth to obtain a division result; Based on the division result, obtain a complete data sample; S2: Based on the data items, divide the data items into homogeneous data items and heterogeneous data items according to the functions of the data items; based on the data sample, extract the time interval data of the data items; based on the time interval data, perform a simplification process on the data items to obtain a simplification result; S3: Based on the simplification result, obtain compressed data items and simplified data items; based on the compressed data items and the simplified data items, obtain a first analysis result of the agricultural analysis data; based on the first analysis result, extract the first data item and the last data item in the data sample; S4: Based on the first data item and the last data item in the data sample, extract the first data item and the last data item in all data samples; based on the change situation of the first data item and the last data item, then obtain a second analysis result of the agricultural analysis data.

[0006] Preferably, the collecting of the agricultural analysis data and the constructing of the data sample of the agricultural analysis data include: The agricultural analysis data at least includes crop growth monitoring data, crop fertilization data, crop fertilization feedback data, and crop yield data during the crop growth process.

[0007] Preferably, the dividing of the data items according to the time sequence of crop growth based on the data sample to obtain a division result; based on the division result, obtaining a complete data sample includes: Divide the data sample in the time sequence of the crop growth process; Based on the division result, obtain data items; Based on the data items, obtain a data sample composed of N data items.

[0008] Preferably, the dividing of the data items into homogeneous data items and heterogeneous data items according to the functions of the data items based on the data items includes: Divide the data items that implement specific functions and combined functions in a data sample, take the data items that implement specific functions as homogeneous data items, and take the data items that implement combined functions as heterogeneous data items.

[0009] Preferably, the extracting of the time interval data of the data items based on the data sample; and the performing of a simplification process on the data items based on the time interval data to obtain a simplification result includes: Perform data standardization processing on the homogeneous data items and the heterogeneous data items in the data sample; Extract the time interval data between the homogeneous data items and the heterogeneous data items, the time interval data between the homogeneous data items and the homogeneous data items, and the time interval data between the heterogeneous data items and the heterogeneous data items; Use the length of the time interval data as the processing standard for the data items; If the time interval is less than or equal to the preset time interval threshold, compress the homogeneous data items or the heterogeneous data items; If the time interval is greater than the preset time interval threshold, simplify the homogeneous data items or the heterogeneous data items; Based on the compression and the simplification processing, obtain the simplified processing result.

[0010] Preferably, the obtaining of the compressed data items and the simplified data items based on the simplified processing result includes: If the data item is processed by compression, delete the duplicate data in the data item, only retain the unique value, and retain the non-duplicate data intact, and define the data item formed by combining the data after the deletion processing and the retention processing as the compressed data item; If the data item is processed by simplification, retain the same data in the data item, mark the importance according to the quantity distribution of the same data, and at the same time delete the data in the data item with the quantity less than the preset quantity threshold, and define the data item formed by combining the data after the deletion processing and the retention processing as the simplified data item.

[0011] Preferably, the obtaining of the first analysis result of the agricultural analysis data based on the compressed data items and the simplified data items; extracting the first data item and the last data item in the data sample based on the first analysis result includes: Based on the compressed data items and the simplified data items, obtain the quantities of the compressed data items and the simplified data items; Use the quantity of the compressed data items and the quantity of all data items as the first ratio; Use the quantity of the simplified data items and the quantity of all data items as the second ratio; Based on the first ratio and the second ratio, obtain the first analysis result; The first data item includes at least one specific parameter in the crop growth monitoring data; The last data item includes at least one type of data in the crop yield data.

[0012] Preferably, the extracting of the first data item and the last data item in all data samples based on the first data item and the last data item in the data sample includes: Construct a new data sample with all the first data items and define it as the first data sample; A new data sample is constructed with all the last data items and is defined as the second data sample.

[0013] Preferably, the obtaining of a second analysis result of the agricultural analysis data based on the change between the first data item and the last data item comprises: Obtain data values ​​of all data items in the second data sample, and take the first data item in the second data sample as a starting point, take the abnormal value of the data value as a split point, traverse the entire second data sample, take the data items from the first data item to the first split point as a first category of data items, take the data items from the first split point to the second split point as a second category of data items, and so on, to obtain several categories of data items; Based on the position of the split point data items in the second data sample, the first data sample is divided in the same way; the compressed data items and simplified data items in the first data sample are extracted, and the ratio of the number of compressed data items to the number of data items in the corresponding category data items, and the ratio of the number of simplified data items to the number of data items in the corresponding category data items are calculated, and then a second analysis result of the agricultural analysis data is obtained.

[0014] Preferably, a smart agricultural data analysis system based on machine learning comprises: Data collection and sample construction module: collect crop growth monitoring data, fertilization data, fertilization feedback data and yield data from multiple sources through IoT sensors, historical planting records and manual inspection logs; cut the original data into data items for continuous periods according to the time axis of the crop growth cycle, and generate a structured data sample consisting of N data items; Data classification and simplification processing module: completes classification according to the functional attributes of data items; standardizes and cleans similar and heterogeneous data items, calculates the time interval between adjacent data items, and performs compression or simplification operations based on preset thresholds; Feature compression and preliminary analysis module: perform deduplication on compressed data items, mark the importance of simplified data items based on data distribution density and remove low-frequency data; count the proportion of compressed and simplified data items, calculate the first ratio and the second ratio as key analysis indicators; extract the first and last items of the data sample, and construct the first and last data indexes; Time series comparison and in-depth analysis module: Aggregate the first items of all samples to construct the first data sample set, and the last items to construct the second data sample set; detect production anomalies in the second data sample set as segmentation points, and reversely map them to the first data sample set for synchronous segmentation; calculate the difference in the proportion of compressed and simplified data items in each segment to generate the second analysis result.

[0015] Beneficial effects: The present invention divides data items based on the chronological order of the crop growth cycle, and classifies data items into homogeneous data items and heterogeneous data items through functional attributes; after extracting time interval data, it dynamically adjusts the compression and simplification strategies, and optimizes the data processing flow according to the sensor state differences. Through the head and tail data comparison mechanism, combined with anomaly detection and segment analysis, it effectively mines the temporal correlation, reduces redundant storage while retaining key features, significantly improves the data management efficiency and decision-making reliability, and provides an accurate basis for agricultural operation optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flowchart of the intelligent agricultural data analysis method based on machine learning of the present invention; Figure 2 is a structural diagram of the intelligent agricultural data analysis system based on machine learning of the present invention; Figure 3 is a schematic diagram of the data sample construction process of the present invention; Figure 4 is a schematic diagram of the compression and simplification process of the present invention; Figure 5 is a schematic diagram of the construction processes of the first data sample and the second data sample of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Embodiment 1: An intelligent agricultural data analysis method based on machine learning, as Figure 1 and Figures 3 - 5 shown, includes the following steps: S1: Collect agricultural analysis data and construct a data sample of the agricultural analysis data; based on the data sample, divide data items according to the chronological order of crop growth to obtain a division result; based on the division result, obtain a complete data sample; S2: Based on the data items, divide the data items into homogeneous data items and heterogeneous data items according to the functions of the data items; based on the data sample, extract the time interval data of the data items; based on the time interval data, perform a simplification process on the data items to obtain a simplification result; S3: Based on the simplified processing result, obtain compressed data items and simplified data items; based on the compressed data items and the simplified data items, obtain the first analysis result of the agricultural analysis data; based on the first analysis result, extract the first data item and the last data item in the data sample; S4: Based on the first data item and the last data item in the data sample, extract the first data item and the last data item in all data samples; based on the change situation of the first data item and the last data item, then obtain the second analysis result of the agricultural analysis data.

[0019] Collect agricultural analysis data and construct a data sample of the agricultural analysis data, including: The agricultural analysis data at least includes crop growth monitoring data, crop fertilization data, crop fertilization feedback data, and crop yield data during the crop growth process.

[0020] It should be noted that modern agricultural planting relies on multi-source sensors to monitor crop growth and environmental parameters, and drives full-process automated operations, generating high-dimensional time-series data. In existing analysis methods, the long short-term memory network (LSTM) model has the characteristics of a black box, manual experience is easily subject to subjective limitations, and the yield is affected by the coupling of multiple factors. It is necessary to systematically integrate growth parameters, environmental variables, and farm operation data to improve the reliability of decision-making.

[0021] Specifically, it is first necessary to collect agricultural analysis data, and the specific process is as follows: Before sowing, obtain the basic parameters of the plot through soil temperature and humidity sensors and light sensors; during the growth period: monitor growth indicators such as transpiration and leaf area index (LAI) in real time through crop physiological sensors (such as stem sap flow meters and leaf area meters); during the farm operation period: obtain fertilization amount, fertilization time, and feedback data based on the operation logs of Internet of Things devices (intelligent fertilization machines, drones) and manual inspection records; during the harvest period: collect the final yield data through yield sensors or manual entry into the system.

[0022] So far, the agricultural analysis data is obtained through the above method.

[0023] Based on the data sample, divide the data items according to the time sequence of crop growth to obtain a division result; based on the division result, obtain a complete data sample, including: Divide the data sample according to the time sequence of the crop growth process; Based on the division result, obtain data items; Based on the data items, obtain a data sample composed of N data items.

[0024] It should be noted that, such as Figure 3As shown, the data sample constructed from agricultural analysis data includes various data throughout the entire growth cycle of crops, including soil temperature and humidity, light intensity, fertilization amount, fertilization time, crop growth indicators (such as transpiration amount, leaf area index), and final yield. The data is segmented by growth stage (sowing stage, growth stage, fertilization stage, maturity stage), and within each stage, the relevant data is combined into "data items". For example: Sowing stage: Combine "soil humidity" and "light intensity" into one data item; Fertilization stage: Combine "soil humidity" and "fertilization amount and fertilization time" into one data item; Maturity stage: Combine "leaf area index" and "transpiration amount" into one data item. The data items of all stages are concatenated in chronological order to form a new data sample. The original data is chaotic (for example, soil humidity, fertilization amount, and transpiration amount are scattered in different tables), and after segmentation, it is sorted by stage to clarify the key parameter combinations for each stage. When analyzing the fertilization effect, directly look at the soil humidity and fertilization amount in the "fertilization stage data item" without having to search for other data; just process the corresponding data items by stage to reduce ineffective calculations.

[0025] Based on the above data items, the data items are divided into similar data items and dissimilar data items according to the functions of the data items, including: The data items that achieve specific functions and combined functions in a data sample are divided. The data items that achieve specific functions are used as similar data items, and the data items that achieve combined functions are used as dissimilar data items.

[0026] It should be noted that the classification method of data items is as follows. Similar data items only record a single function or parameter. For example: Only record "soil humidity", or only record "light intensity". Dissimilar data items: Record data of multiple related functions or parameters at the same time. For example: Record "soil humidity and fertilization amount" at the same time, or "light intensity and transpiration amount". Purpose of division: Similar data items: Used to quickly view the changes of a single parameter (such as only checking whether the soil humidity meets the standard); Dissimilar data items: Used to analyze how multiple parameters jointly affect the result (such as whether the fertilization amount needs to be adjusted when the soil humidity is low). Example, the use of similar data items: Quickly check whether a single parameter is abnormal (for example, trigger an irrigation alarm when the soil humidity is too low). The use of dissimilar data items: Analyze the relationship between parameters (for example, whether the transpiration amount increases when the light is strong and whether the irrigation strategy needs to be adjusted).

[0027] It should be noted that a data sample consists of several data items, and the data items contain similar data items and dissimilar data items. Among them, the quantities of similar data items and dissimilar data items are determined according to the actual situation.

[0028] Based on the above data sample, extract the time interval data of the data items; based on the time interval data, perform a simplification process on the data items to obtain a simplification result, including: Perform data standardization processing on the homogeneous data items and heterogeneous data items in the data sample; Extract the time interval data between the homogeneous data items and the heterogeneous data items, the time interval data between the homogeneous data items and the homogeneous data items, and the time interval data between the heterogeneous data items and the heterogeneous data items; Use the length of the time interval data as the processing standard for the data items; If the time interval is less than or equal to the preset time interval threshold, compress the homogeneous data items or the heterogeneous data items; If the time interval is greater than the preset time interval threshold, simplify the homogeneous data items or the heterogeneous data items; Based on the compression and the simplification processing, obtain the simplification processing result.

[0029] It should be noted that data standardization processing: converts the data collected by different sensors into a unified format, including: unit unification: converting the temperature unit to °C and the humidity to a percentage; data cleaning: removing missing values and data outside the physical range (such as humidity > 100%); time alignment: unifying the timestamps to millisecond-level accuracy.

[0030] Time interval extraction and threshold setting: Sensor response time: Record the time taken from detecting an anomaly (such as too high temperature) to triggering an operation (such as cooling down), and calculate the threshold ; Data generation interval: Record the interval between two data transmissions by the sensor (such as once every 10 minutes), and calculate the threshold ; Adjustment coefficient : Select according to business requirements (for example, in a loose scenario = 1, in a strict scenario = 2).

[0031] Status judgment and processing: If the response time ≤ and the generation interval ≤ , it is determined to be in a normal state, indicating that the sensor is working properly, the data is reliable, and duplicate data can be compressed (such as only retaining one of the consecutive same temperature values); if any interval exceeds the threshold, it is determined to be in an abnormal state, indicating that the sensor is faulty or has low efficiency, and the data needs to be simplified (such as deleting abnormal values with an appearance frequency < 2 times), and trigger the alarm mechanism.

[0032] Among them, (mean response time): The average time taken by the sensor from detecting an anomaly (such as too high temperature) to triggering an operation (such as starting cooling down), (standard deviation of response time): Measures the volatility of the response time. The larger the standard deviation, the more unstable the response time, (Generated Interval Mean): The average interval time between two data transmissions by the sensor. (Generated Interval Standard Deviation): Measures the stability of the data generation interval. A small standard deviation indicates regular intervals, while a large one indicates interruptions or delays.

[0033] Based on the simplified processing result, compressed data items and simplified data items are obtained, including: If the data item is for compression processing, duplicate data in the data item is deleted, and only unique values are retained. Non-duplicate data is retained in its entirety, and the data item formed by combining the data after deletion processing and retention processing is defined as the compressed data item. If the data item is for simplified processing, the same data in the data item is retained, and importance markings are made according to the quantity distribution of the same data. At the same time, data with a quantity less than the preset quantity threshold in the data item is deleted, and the data item formed by combining the data after deletion processing and retention processing is defined as the simplified data item.

[0034] It should be noted that as Figure 4 shown, the compression processing steps: Delete duplicate data: If there are multiple identical data in the data item, only one is retained. Retain different data: All different data is retained. Form the compressed data item: Combine the retained data into a new data item. Purpose: Reduce redundant data and save storage space. Example: Original data item: 12°C, 12°C, 13°C, 13°C, 15°C, after compression: 12°C, 13°C, 15°C (delete the duplicate 12°C and 13°C). Example: When the sensor is working normally, the same temperature is continuously collected multiple times (such as in a constant temperature greenhouse), and only one valid data needs to be stored after compression.

[0035] The simplified processing steps: Mark importance: Count the number of times the same data appears, and the data with more occurrences is marked as important data. The weight of the influence of different parameters in the simplified data item on the yield is evaluated through a random forest to achieve intelligent marking. Machine learning can adapt to complex agricultural scenarios and improve the accuracy and generalization ability of data analysis. Delete low-frequency data: Delete data with a number of occurrences less than the preset quantity threshold (such as the number of occurrences ≤ 2 times). Form the simplified data item: Combine the retained important data and the different data that have not been deleted into a new data item. Purpose: Eliminate noise data (such as occasional error values caused by sensor abnormalities) and retain key information. Example: Original data item: 12°C, 12°C, 13°C, 13°C, 15°C (assuming the preset quantity threshold = 2 times), after simplification: 12°C, 12°C, 13°C, 13°C (delete 15°C). Application scenario: When the sensor is abnormal, an abnormal temperature value (such as 15°C) is occasionally collected, and the noise is eliminated after simplification, and the main data is retained.

[0036] Based on the compressed data items and the simplified data items, obtain a first analysis result of the agricultural analysis data; based on the first analysis result, extract the first data item and the last data item in the data sample, including: Based on the compressed data items and the simplified data items, obtain the quantities of the compressed data items and the simplified data items; Use the quantity of the compressed data items and the quantity of all data items as the first ratio; Use the quantity of the simplified data items and the quantity of all data items as the second ratio; Based on the first ratio and the second ratio, obtain the first analysis result; The first data item includes at least one specific parameter in the crop growth monitoring data; The last data item includes at least one type of data in the crop yield data.

[0037] It should be noted that the definition of the first ratio (proportion of compressed data items): The proportion of the quantity of compressed data items in all data items. The larger the first ratio: It indicates that there is more duplicate content in the data, the sensor works stably but the amount of redundant data is large. For example: The temperature sensor sends the same data every hour (such as 25°C continuously 10 times), and only 1 time is retained after compression, and the proportion of compressed data items is high. Practical significance: Reflects the degree of data redundancy. When the ratio is high, it is necessary to optimize the acquisition frequency or storage strategy.

[0038] The definition of the second ratio (proportion of simplified data items): The proportion of the quantity of simplified data items in all data items. The larger the second ratio: It indicates that there is more abnormal or noisy data in the data, the sensor has a fault or the environmental interference is large. For example: The humidity sensor frequently collects abnormal values (such as 40%, 80%, 45%), and the low-frequency abnormal values (such as 40%) are deleted after simplification, and the proportion of simplified data items is high. Practical significance: Reflects the data noise level. When the ratio is high, it is necessary to repair the equipment or optimize the anti-interference algorithm. Practical application analysis scenario 1: The first ratio is 80% and the second ratio is 5%. Conclusion: The data redundancy is serious but the noise is small. It is recommended to reduce the sensor sampling frequency (such as from every minute to every 10 minutes). Scenario 2: The first ratio is 20% and the second ratio is 50%. Conclusion: The data redundancy is small but the noise is large. It is necessary to check whether the sensor is damaged or the environmental interference source (such as electromagnetic interference). The significance of associating the first and last data items. The first data item (such as soil humidity at the sowing stage): represents the initial state. If its compression / simplification ratio is abnormal, it will affect subsequent operations; the last data item (such as yield at the harvest stage): represents the final result. If the associated simplification ratio is high, it indicates that there is noise in the data during the critical growth period, resulting in a deviation in yield prediction.

[0039] Based on the first and last data items in the data samples, extract the first and last data items from all data samples, including: Construct a new data sample with all the first data items and define it as the first data sample; Construct a new data sample with all the last data items and define it as the second data sample.

[0040] It should be noted that, as Figure 5 shown, the first data sample: Take out the first data items in all data samples (such as soil humidity and light intensity at sowing time) and form a new data set. The second data sample: Take out the last data items in all data samples (such as yield data at harvest time) and form another new data set. Purpose: The first data sample: Analyze the impact of the initial conditions of crops (such as soil conditions) on overall growth; The second data sample: Analyze the correlation between the final yield and operations at each stage (such as fertilization and irrigation). Similarly, the correlation between the data samples constructed from all the second data items and the data samples constructed from all the last data items can be analyzed, and so on, which will not be elaborated here.

[0041] Based on the change situations of the first and last data items, then obtain the second analysis result of agricultural analysis data, including: Obtain the data values of all data items in the second data sample, and starting from the first data item in the second data sample, using the outliers of the data values as the segmentation points, traverse the entire second data sample. Take the data items from the first data item to the first segmentation point as one type of data items, take the data items from the first segmentation point to the second segmentation point as the second type of data items, and so on, to obtain several types of data items; Based on the positions of the segmentation point data items in the second data sample, make the same division for the first data sample; Extract the compressed data items and simplified data items in the first data sample, and use the ratio of the number of compressed data items to the number of data items in the corresponding type of data items and the ratio of the number of simplified data items to the number of data items in the corresponding type of data items, then obtain the second analysis result of agricultural analysis data.

[0042] It should be noted that to find the yield outliers: in the second data sample (yield data), find the yield values that are higher or lower than the preset yield threshold as the splitting points. For example, if the normal yield is 400 - 600 kg / mu, and the yield of a certain plot is 800 kg or 200 kg, it is marked as an abnormal splitting point. Split the yield data: divide the yield data into several segments according to the splitting points. Example: Splitting point 1: 800 kg; Splitting point 2: 200 kg; Segmentation result: high-yield segment (800 kg / mu), normal segment (400 - 600 kg / mu), low-yield segment (200 kg / mu). Synchronously split the sowing data: according to the position of the yield splitting points, also divide the first data sample (sowing date data, such as soil humidity, light) into corresponding segments. Example: Sowing data corresponding to the high-yield segment: soil humidity = 70%, light = 2200 Lx; Sowing data corresponding to the low-yield segment: soil humidity = 50%, light = 1500 Lx. Analyze the compression and simplification ratios: count the proportions of compressed data items (after duplicate data is deleted) and simplified data items (after noise data is deleted) in each segment. For example: High-yield segment: compression ratio 80% (data is stable), simplification ratio 5% (less noise); Low-yield segment: compression ratio 30% (data is unstable), simplification ratio 60% (more noise).

[0043] The first analysis result directly guides the data storage strategy by quantifying the proportions of compressed and simplified data; the second analysis result mines the correlation between abnormal growth stages and yield through head-to-tail data comparison. The two form a technical closed-loop of "data optimization - precise decision-making - feedback correction", which not only improves the storage efficiency but also enhances the pertinence of agricultural operations. For example, if the soil humidity during the sowing period of a certain plot is abnormal (the first analysis result shows a high simplification ratio), resulting in a low yield at the harvest stage (the second analysis result locates the low-yield segment), an irrigation adjustment suggestion is automatically triggered, and at the same time, a sensor maintenance task is marked.

[0044] Embodiment 2: On the basis of Embodiment 1, a smart agriculture data analysis system based on machine learning, as Figure 2 shown, includes: Data acquisition and sample construction module: Multisource collects crop growth monitoring data, fertilization data, fertilization feedback data, and yield data through Internet of Things sensors, historical planting records, and manual inspection logs; cuts the original data into continuous time-period data items according to the time axis of the crop growth cycle, and generates a structured data sample composed of N data items; Data classification and simplification processing module: Completes classification according to the functional attributes of the data items; performs standardized cleaning on the same-type and different-type data items, calculates the time interval between adjacent data items, and performs compression or simplification operations based on a preset threshold; Feature Compression and Preliminary Analysis Module: Perform deduplication on compressed data items, mark the importance of simplified data items based on data distribution density, and eliminate low-frequency data; count the proportions of compressed and simplified data items, calculate the first ratio and the second ratio as key analysis indicators; extract the first and last items of the data sample to construct the first and last data indexes. Time Series Comparison and In-depth Analysis Module: Aggregate the first items of all samples to construct the first data sample set, and the last items to construct the second data sample set; detect production outliers in the second data sample set as segmentation points, and inversely map them to the first data sample set for synchronous segmentation; calculate the proportion differences of compressed and simplified data items within each segment to generate the second analysis result.

[0045] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A smart agricultural data analysis method based on machine learning, characterized in that: The following steps are involved: S1: Collect agricultural analysis data and construct a data sample of the agricultural analysis data; Based on the data sample, the data items are divided according to the time sequence of crop growth to obtain a division result; based on the division result, a complete data sample is obtained; S2: Based on the data items, the data items are divided into similar data items and heterogeneous data items according to their functions; based on the data samples, the time interval data of the data items are extracted; based on the time interval data, the data items are simplified to obtain simplified processing results; S3: Based on the simplification processing result, a compressed data item and a simplified data item are obtained; Based on the compressed data item and the simplified data item, obtaining a first analysis result of the agricultural analysis data; based on the first analysis result, extracting a first data item and a last data item in the data sample; S4: Based on the first data item and the last data item in the data sample, extract the first data item and the last data item in all data samples; and then obtain a second analysis result of the agricultural analysis data based on the changes in the first data item and the last data item.

2. The method for analyzing smart agricultural data based on machine learning according to claim 1, characterized in that: The collecting of agricultural analysis data and constructing a data sample of the agricultural analysis data includes: The agricultural analysis data at least includes crop growth monitoring data, crop fertilization data, crop fertilization feedback data, and crop yield data during the crop growth process.

3. The method for analyzing smart agricultural data based on machine learning according to claim 1, characterized in that: Based on the data sample, the data items are divided according to the time sequence of crop growth to obtain a division result; Based on the division results, a complete data sample is obtained, including: Dividing the data samples according to the time sequence of the crop growth process; Based on the division result, obtaining data items; Based on the data items, a data sample consisting of N data items is obtained.

4. The method for analyzing smart agricultural data based on machine learning according to claim 1, characterized in that: The dividing the data items into similar data items and heterogeneous data items based on the data items and according to their functions includes: The data items that realize specific functions and combined functions in a data sample are divided into data items that realize specific functions as similar data items, and data items that realize combined functions as heterogeneous data items.

5. The method for analyzing smart agricultural data based on machine learning according to claim 1, characterized in that: extracting time interval data of the data item based on the data sample; Based on the time interval data, the data item is simplified to obtain a simplified processing result, including: Performing data standardization processing on the homogeneous data items and the heterogeneous data items in the data sample; Extracting the time interval data between the same type of data items and the different type of data items, the time interval data between the same type of data items and the time interval data between different type of data items; The length of the time interval data is used as a processing standard for the data item; If the time interval is less than or equal to a preset time interval threshold, compressing the same type of data items or the different type of data items; If the time interval is greater than a preset time interval threshold, simplifying the same type of data items or the different type of data items; Based on the compression and the simplification processing, a simplification processing result is obtained.

6. The method for analyzing smart agricultural data based on machine learning according to claim 1, characterized in that: The step of obtaining compressed data items and simplified data items based on the simplified processing result includes: If the data item is compressed, the duplicate data in the data item is deleted, only unique values ​​are retained, and non-duplicate data is completely retained, and the data item composed of the deleted and retained data is defined as a compressed data item; If the data item is simplified, the same data in the data item will be retained and marked according to the quantity distribution of the same data. At the same time, the data in the data item whose quantity is less than the preset quantity threshold will be deleted, and the data item composed of the deleted and retained data will be defined as a simplified data item.

7. The method for analyzing smart agricultural data based on machine learning according to claim 1, characterized in that: said obtaining a first analysis result of the agricultural analysis data based on the compressed data item and the simplified data item; Extracting a first data item and a last data item in the data sample based on the first analysis result includes: Based on the compressed data items and the simplified data items, obtaining the quantities of the compressed data items and the simplified data items; Taking the number of the compressed data items and the number of all data items as a first ratio; Taking the number of the simplified data items and the number of all data items as a second ratio; Obtaining a first analysis result based on the first ratio and the second ratio; The first data item includes at least one type of specific parameter in the crop growth monitoring data; The last data item includes at least one type of data of crop yield data.

8. The method for analyzing smart agricultural data based on machine learning according to claim 1, characterized in that: The extracting the first data item and the last data item in all data samples based on the first data item and the last data item in the data sample comprises: Constructing a new data sample with all the first data items and defining it as the first data sample; A new data sample is constructed with all the last data items and is defined as the second data sample.

9. The method for analyzing smart agricultural data based on machine learning according to claim 8 is characterized in that: The method of obtaining a second analysis result of the agricultural analysis data based on the change between the first data item and the last data item includes: Obtain data values ​​of all data items in the second data sample, and take the first data item in the second data sample as a starting point, take the abnormal value of the data value as a split point, traverse the entire second data sample, take the data items from the first data item to the first split point as a first category of data items, take the data items from the first split point to the second split point as a second category of data items, and so on, to obtain several categories of data items; Based on the position of the split point data items in the second data sample, the first data sample is divided in the same way; the compressed data items and simplified data items in the first data sample are extracted, and the ratio of the number of compressed data items to the number of data items in the corresponding category data items, and the ratio of the number of simplified data items to the number of data items in the corresponding category data items are calculated, and then a second analysis result of the agricultural analysis data is obtained.

10. A smart agricultural data analysis system based on machine learning, used to implement a smart agricultural data analysis method based on machine learning as described in any one of claims 1 to 9, characterized in that: include: Data collection and sample construction module: collect crop growth monitoring data, fertilization data, fertilization feedback data and yield data from multiple sources through IoT sensors, historical planting records and manual inspection logs; cut the original data into data items for continuous periods according to the time axis of the crop growth cycle, and generate a structured data sample consisting of N data items; Data classification and simplification processing module: completes classification according to the functional attributes of data items; standardizes and cleans similar and heterogeneous data items, calculates the time interval between adjacent data items, and performs compression or simplification operations based on preset thresholds; Feature compression and preliminary analysis module: perform deduplication on compressed data items, mark the importance of simplified data items based on data distribution density and remove low-frequency data; count the proportion of compressed and simplified data items, calculate the first ratio and the second ratio as key analysis indicators; extract the first and last items of the data sample, and construct the first and last data indexes; Time series comparison and in-depth analysis module: Aggregate the first items of all samples to construct the first data sample set, and the last items to construct the second data sample set; detect production anomalies in the second data sample set as segmentation points, and reversely map them to the first data sample set for synchronous segmentation; calculate the difference in the proportion of compressed and simplified data items in each segment to generate the second analysis result.

Images