Pest prediction device using field data and method using same for pest prediction
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- EPINET CORP
- Filing Date
- 2026-01-09
- Publication Date
- 2026-07-30
Smart Images
Figure KR2026000500_30072026_PF_FP_ABST
Abstract
Description
Pest prediction device utilizing field data and pest prediction method using the same
[0001] The present invention relates to a pest prediction device and a pest prediction method using the same, and more particularly to a pest prediction device utilizing field data that predicts pest occurrence by applying a model based on the annual generation number of insects, and a pest prediction method using the same.
[0002] Insects are arthropods that inhabit diverse environments on Earth and are an important component of the ecosystem. Among these insects, those that cause harm to human activities and daily life are called pests, and these pests cause serious problems by inflicting damage across various fields, including agriculture, forestry, and public health.
[0003] Previously, statistical analysis techniques based on meteorological data or simple regression models were used to predict the occurrence of these pests. However, these methods failed to adequately reflect the physiological characteristics of pests and complex environmental factors, resulting in low prediction accuracy and limitations in considering regional or pest-specific characteristics. Furthermore, existing technologies faced difficulties in automatically processing large volumes of data or analyzing and utilizing them in real time.
[0004] Therefore, more sophisticated prediction technologies utilizing weather and pest data are required to enable accurate prediction and efficient management of pest outbreaks. In particular, there is a need to develop technologies capable of rapidly and accurately predicting pest outbreaks by region and time period through automated data processing and analysis.
[0005] The objective of the present invention is to provide a pest prediction device utilizing field data that can easily and accurately predict the occurrence of pests, and a pest prediction method using the same.
[0006] The objectives of the present invention are not limited to those mentioned above, and other objectives and advantages of the present invention not mentioned may be understood from the following description and will be more clearly understood by the embodiments of the present invention. Furthermore, it will be readily apparent that the objectives and advantages of the present invention can be realized by the means and combinations thereof set forth in the claims.
[0007] A pest prediction device utilizing field data according to the present invention comprises: a data collection unit that collects weather data and pest data of a target area; a data preprocessing unit that standardizes the amount of pest occurrence using the weather data and pest data collected by the data collection unit, calculates the temperature required for pest growth to obtain the accumulated heat amount, and processes missing values; a model training unit that analyzes pest occurrence patterns and trains a model based on the data processed by the data preprocessing unit; and a prediction unit that predicts pest occurrence in a target area based on the final model trained by the model training unit.
[0008] The preprocessing unit includes a data converter that calculates the cumulative heat amount based on the lower and upper temperature limits of pest development for each target region and converts pest occurrence information into cumulative pest ratio data; a model selector that selects a model based on the annual number of pest generations based on the cumulative heat amount and cumulative ratio data output from the data converter; a training dataset for selecting variables of the model selected by the model selector; and a dataset builder that constructs a validation dataset for measuring the prediction accuracy of the model trained on the training dataset.
[0009] The model training unit includes a Waybull function selector that applies a model selected in the data preprocessing unit, an initial value range setter that specifies the range of initial values of the model applied in the Waybull function selector, trains with arbitrary values within the interval, and evaluates using the coefficient of determination, a model variable estimator that estimates model variables such that the deviation between the model applied in the Waybull function selector and the pest occurrence amount information of the pest data is minimized, and an outlier remover that removes outliers based on the distance between the model and the pest occurrence amount information.
[0010] A pest prediction method utilizing field data according to the present invention comprises: a step in which a data collection unit collects weather data and pest data of a target area; a step in which a data preprocessing unit standardizes the amount of pest occurrence and processes missing values using the weather data and pest data collected by the data collection unit, and then calculates the temperature required for pest growth to obtain an accumulated amount of heat; a step in which a model learning unit analyzes the pest occurrence pattern and trains a model based on the data processed by the data preprocessing unit; and a step in which a prediction unit predicts the occurrence of pests in a target area based on the final model trained by the model learning unit.
[0011] The step of obtaining cumulative heat amount by calculating the temperature required for pest growth and the data preprocessing unit standardizing the pest occurrence amount and processing missing values using weather data and pest data collected by the data collection unit; comprises: a step in which a data converter calculates the cumulative heat amount based on the lower and upper developmental temperature limits of pests for each target region and converts the pest occurrence amount information into cumulative ratio data of pests; a step in which a model selector selects a model according to the annual number of generations of pests based on the cumulative heat amount and cumulative ratio data output from the data converter; and a step in which a dataset builder constructs a training dataset for selecting variables of the model selected by the model selector and a verification dataset for measuring the prediction accuracy of the model trained on the training dataset.
[0012] The step of a model learning unit analyzing pest occurrence patterns and training a model based on data processed by a data preprocessing unit; comprises: a step in which a Waybull function selector applies a model selected by the data preprocessing unit; a step in which an initial value range specifier specifies the range of initial values of the model applied by the Waybull function selector, trains the model with arbitrary values within the interval, and evaluates it using a coefficient of determination; a step in which a model variable estimator estimates model variables such that the deviation between the model applied by the Waybull function selector and the pest occurrence amount information of the pest data is minimized; and a step in which an outlier remover removes outliers based on the distance between the model and the pest occurrence amount information.
[0013] The step in which the model training unit analyzes pest occurrence patterns and trains the model based on data processed by the data preprocessing unit; comprises a fit validator calculating SSE (sum of squares residual of error), SSR (sum of squared due to regression), and SST (total sum of squares); calculating the mean values MSE (mean squared error) and MSR (mean squares due to regression) using the calculated SSE, SSR, SST, and degrees of freedom; calculating the F value and P value using the calculated MSR and MSE; checking whether the P value is less than 0.05; calculating the variable value and the SEM (standard error of mean) of the variable value; and checking whether the magnitude of the variable value is at least one order of magnitude larger than the SEM, while using SSE and SST, r 2The method may further include a step of verifying the model's fit by calculating a value; and a step of converting the 50% occurrence time for each peak period of pest occurrence in the model into a Julian day when the predictive power verifier identifies the cumulative heat amount, converting the 50% occurrence time for each peak period of pest occurrence in the verification data into a Julian day to calculate the difference between the predicted time and the actual occurrence time for each peak period of pest occurrence and automatically displaying it in a table.
[0014] The model selected in the model selector is,
[0015] is,
[0016] n is the annual number of generations of pests, which is 0 if only overwintering generations exist, and x is the cumulative daily temperature, representing the accumulated amount of heat during weather conditions where the temperature exceeds a certain standard during a specific period, and α i is the height of the i-th peak curve, a constant that controls the maximum value of the peak curve, and β i is the point at which 50% of the i-th peak curve occurs based on cumulative day temperature, and γ i is a parameter that controls the steepness of the i-th curve.
[0017] The pest prediction device utilizing field data and the pest prediction method using the same according to the present invention can easily and accurately predict the occurrence of pests by applying a model based on the annual number of generations of insects based on weather data and pest data.
[0018] In addition to the effects described above, the specific effects of the present invention are described together with the specific details for implementing the invention below.
[0019] Figure 1 is a conceptual diagram of a pest prediction device utilizing field data according to the present invention.
[0020] Figure 2 is a flowchart of a pest prediction method using field data according to the present invention.
[0021] The aforementioned objectives, features, and advantages are described in detail below with reference to the attached drawings, thereby enabling those skilled in the art to easily implement the technical concept of the present invention. In describing the present invention, detailed descriptions of known technologies related to the present invention are omitted if it is determined that such descriptions would unnecessarily obscure the essence of the invention. Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the attached drawings. In the drawings, the same reference numerals are used to indicate the same or similar components.
[0022] Although terms such as "first," "second," etc., are used to describe various components, it goes without saying that these components are not limited by these terms. These terms are used merely to distinguish one component from another, and unless specifically stated otherwise, the first component may also be the second component.
[0023] Throughout the specification, unless specifically stated otherwise, each component may be singular or plural.
[0024] In the following, the statement that any configuration is placed on the "upper (or lower)" of a component or on the "upper (or lower)" of a component may mean not only that any configuration is placed in contact with the upper (or lower) surface of said component, but also that another configuration may be interposed between said component and any configuration placed on (or below) said component.
[0025] In addition, where it is stated that one component is "connected," "combined," or "connected" to another component, it should be understood that while the components may be directly connected or connected to each other, another component may be "interposed" between each component, or each component may be "connected," "combined," or "connected" through another component.
[0026] Additionally, singular expressions used in this specification include plural expressions unless the context clearly indicates otherwise. In this application, terms such as "composed of" or "comprising" should not be interpreted as necessarily including all of the various components or steps described in the specification, and should be interpreted as meaning that some of the components or steps may not be included, or that additional components or steps may be included.
[0027] Throughout the specification, "A and / or B" means A, B, or A and B unless specifically stated otherwise, and "C to D" means C or more and D or less unless specifically stated otherwise.
[0028] Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings.
[0029] Figure 1 is a conceptual diagram of a pest prediction device utilizing field data according to the present invention.
[0030] As shown in FIG. 1, the pest prediction device utilizing field data according to the present invention includes a data collection unit (100) that collects weather data and pest data, a data preprocessing unit (200) that preprocesses weather data and pest data, a model learning unit (300) for the preprocessed weather data and pest data, and a prediction unit (400).
[0031] The data collection unit (100) collects data for predicting pests that may occur in a target area. Here, the data for pest prediction includes weather data and pest data, and accordingly, the data collection unit (100) includes a weather data collector (110) for collecting weather data and a pest data collector (120) for collecting pest data.
[0032] The weather data collector (110) collects weather data of the target area. Here, the collected weather data includes the daily maximum temperature and daily minimum temperature of the target area, observation location, and observation time information.
[0033] The pest data collector (120) collects data on pests whose occurrence is to be predicted, i.e., pest data. Here, the necessary pest data includes the density of specific insects surveyed during the year, the survey date, and the survey location (address, latitude / longitude, etc.). Accordingly, the pest data collector (120) collects survey location information, survey date information, and pest occurrence amount information.
[0034] The data preprocessing unit (200) standardizes the amount of pest occurrence and processes missing values, then calculates the temperature required for pest growth to obtain the accumulated heat amount. Here, the data preprocessing unit (200) may treat cases as missing values, for example, when pest data exists but weather data does not. In addition, to this end, the data preprocessing unit (200) includes a data converter (210), a model selector (220), and a dataset builder (230).
[0035] The data converter (210) calculates the cumulative heat amount based on the lower and upper developmental temperatures of pests in each target area and converts the pest occurrence information into cumulative ratio data of pests. Additionally, to this end, the data converter (210) includes a cumulative heat amount calculator (211) and a cumulative ratio data converter (212). First, the data converter (210) converts the annual pest occurrence information for each target area into cumulative ratio information to accommodate variations in occurrence amounts by region. This can be performed by converting the density per sampling unit into a cumulative ratio (cumulative density / total annual occurrence). This embodiment exemplifies converting the pest occurrence information generated over one year in each target area into ratio data ranging from 0 to 1. Additionally, the data converter (210) calculates the insect-specific cumulative heat amount for each survey year at each survey point using an arbitrary potential lower developmental threshold temperature and daily minimum / maximum temperatures. It is exemplified that an arbitrary developmental zero temperature is used if research data for the insect being developed is available, and 10°C is used otherwise. Here, the calculation of cumulative heat is expressed by predicting the change in daily temperature data using a sine function based on the daily minimum and maximum temperatures. Subsequently, the sine function is segmented by the developmental minimum and upper limit temperatures to calculate the area under the sine function between the developmental zero temperature and the upper limit temperature. The calculated area under the sine function represents the daily accumulated degree-day, and summing these daily accumulated degrees-days yields the cumulative degree-day for a specific period. Since insects are poikilothermic animals, the cumulative heat required during a specific event period (e.g., the cumulative heat needed for the next adult to emerge) has a fixed value. The calculated cumulative degree-day variable is combined with the pest occurrence rate data, and this is utilized to create a pest occurrence prediction model.At this time, the sum of the squares of the deviations from the model of the generated model and each data is calculated, which is called the mean squared error (MSE), and the value that is minimized is determined as the developmental zero temperature.
[0036] The model selector (220) selects and applies a model based on the annual generation of pests, i.e., a pest prediction model, based on the cumulative heat amount and cumulative ratio data generated by the data converter (210). This can be done by applying a model based on the annual generation of pests, i.e., a Waybull function. Here, the selection of the model is based on the annual generation of a specific insect, i.e., a pest, and is determined by adding 1 to the known generation number, as overwintering generations are generally not included in the annual generation number. However, if there is no known generation number, it is estimated by applying the Waybull function. The applied model based on the annual generation number is as follows.
[0037] First, the model applied when only the overwintering generation exists is the same as the mathematical formula 1 (1 Peak Weibull function) below.
[0038]
[0039] In Equation 1, x represents the cumulative temperature, which signifies the accumulated heat of weather conditions where the temperature exceeds a certain threshold during a specific period. Additionally, α represents the height of the first peak curve and is a constant that controls the maximum value of the peak curve. This embodiment exemplifies that α is 1, which means that the maximum height of the first peak is 1. β represents the point at which 50% of the first peak curve occurs based on the cumulative temperature. This means that when the cumulative temperature is β, the value of the curve reaches 50% of the maximum value. γ is a parameter that controls the steepness of the first curve. If the value of γ is small, the curve becomes flatter, and if the value is large, the curve rises or falls more steeply.
[0040] Next, for insects that have one generation per year, i.e., overwintering generation + one generation (2 Peak Weibull function), it is as shown in Equation 2 below.
[0041]
[0042] In Equation 2, x represents the cumulative temperature, and α1 and α2 represent the heights of the first and second peak curves, respectively. The sum of α1 and α2 is set to 1, so that the sum of the heights of the first and second peak curves is always 1. Additionally, β1 is the point at which the first peak curve reaches 50% based on the cumulative temperature, and γ1 and γ2 are parameters that control the steepness of the first and second peak curves, respectively; a larger value indicates a steeper rise or fall of the curve, while a smaller value indicates a gentler slope. Δβ1 is the time difference between the point at which the first and second peak curves reach 50% based on the cumulative temperature, representing the time interval between the point at which the first curve reaches 50% and the point at which the second curve reaches 50%.
[0043] In addition, for insects that have two generations per year, i.e., overwintering generation + 2 generations (3 Peak Weibull function), it is as shown in Equation 3 below.
[0044]
[0045] In Equation 3, x is the cumulative temperature, and α1, α2, and α3 are the heights of the first, second, and third curves, respectively, with the sum of α1, α2, and α3 set to 1. Additionally, β1 is the point at which the first peak curve reaches 50% of its height based on the cumulative temperature, and γ1, γ2, and γ3 are parameters that control the steepness of the first, second, and third peak curves, respectively; larger values cause the curve to rise or fall more steeply, while smaller values cause it to become flatter. Δβ1 is the time difference between the point at which the first and second peak curves reach 50% of their height based on the cumulative temperature. Δβ2 is the time difference between the point at which the second and third peak curves reach 50% of their height based on the cumulative temperature.
[0046] For insects that have three generations per year, i.e., overwintering generation + three generations (4 Peak Weibull function), it is as shown in Equation 4 below.
[0047]
[0048] In Equation 4, x is the cumulative temperature. α1, α2, α3, and α4 are the heights of the first, second, third, and fourth peak curves, respectively, and the sum of α1, α2, α3, and α4 is set to 1. β1 is the point at which the first peak curve reaches 50% of its cumulative temperature. γ1, γ2, γ3, and γ4 are parameters that control the steepness of the first, second, third, and fourth peak curves, respectively; the larger the value, the steeper the curve, and the smaller the value, the flatter it becomes. Δβ1 is the time difference between the point at which the first and second peak curves reach 50% of their cumulative temperature, and Δβ2 is the time difference between the point at which the second and third peak curves reach 50% of their cumulative temperature. In addition, Δβ3 is the time difference between the 50% occurrence point of the third peak curve and the fourth peak curve based on the cumulative temperature.
[0049] In addition, for insects that have 4 generations per year, i.e., overwintering generation + 4 generations (5 Peak Weibull function), it is as shown in Equation 3 below.
[0050]
[0051] In Equation 5, x is the cumulative temperature. α1, α2, α3, α4, and α5 are the heights of the first, second, third, fourth, and fifth peak curves, respectively, and the sum of α1, α2, α3, α4, and α5 is set to 1. β1 is the point at which the first peak curve reaches 50% based on the cumulative temperature. γ1, γ2, γ3, γ4, and γ5 are parameters that control the steepness of the first, second, third, fourth, and fifth peak curves, respectively; the larger the value, the steeper the curve, and the smaller the value, the flatter it becomes. Δβ1 is the time difference between the point at which the first and second peak curves reach 50% based on the cumulative temperature, and Δβ2 is the time difference between the point at which the second and third peak curves reach 50% based on the cumulative temperature. In addition, Δβ3 is the time difference between the 50% occurrence point of the third peak curve and the fourth peak curve based on the cumulative temperature of day, and Δβ4 is the time difference between the 50% occurrence point of the fourth peak curve and the fifth peak curve based on the cumulative temperature of day.
[0052] Meanwhile, the aforementioned model can increase the number of generations to infinity. This can be implemented by increasing α and Δβ to match the number of generations of insects, and when generalized to insects with n generations per year, it can be expressed as Equation 6 below.
[0053]
[0054] In mathematical equation 6, n represents the annual number of insect generations; it is 0 if only overwintering generations exist and 1 if one generation occurs annually. x is the cumulative daily temperature, which refers to the accumulated amount of heat during weather conditions where the temperature exceeds a certain threshold over a specific period. Additionally, α i is the height of the i-th peak curve, which is a constant that controls the maximum value of the peak curve. In this embodiment, α iThis exemplifies that α is 1, meaning that the maximum height of the first peak is 1. β i is the point at which 50% of the i-th peak curve occurs based on the cumulative temperature-to-heat. This means that when the cumulative temperature-to-heat is β, the curve value reaches 50% of the maximum value. γ i is a parameter that controls the steepness of the i-th curve.
[0055] The dataset builder (230) builds a training dataset and a verification dataset. Here, the training dataset refers to a dataset used to select model variables, and the verification dataset refers to a dataset used to measure the prediction accuracy of a model developed from the training dataset. Here, the classification of the datasets can be done arbitrarily or selectively provided by the user. Additionally, the dataset builder (230) may be configured to select the dataset with the highest prediction accuracy as the verification dataset. Since the prediction accuracy must be high, this embodiment exemplifies selecting the dataset with the highest prediction accuracy as the verification dataset.
[0056] Generally, pests have a fixed number of annual peak occurrence periods. For example, the tobacco moth in chili peppers has three peak occurrence periods. In this case, the optimal time for control is determined based on the maximum occurrence period (50% occurrence period) for each peak occurrence period. The difference between the model prediction value (calendar day) and the actual occurrence value (calendar day) for the first peak occurrence, the difference between the model prediction value and the actual occurrence value for the second peak occurrence, and the difference between the model prediction value and the actual occurrence value for the third peak occurrence are all calculated. The year in which the sum of these values is minimized is then determined, and the data from this year is used as validation data, while the data from the year determined for validation is excluded from model development.
[0057] The model training unit (300) trains a model to predict pest occurrence. To this end, the model training unit (300) includes a Waybull function selector (310), an initial value range specifier (320), a model variable estimator (330), an outlier remover (340), a fit verifier (350), a predictive power verifier (360), and a data storage unit (370).
[0058] The Waybull function selector (310) selects and applies the model selected from the model selector (220), that is, the Waybull function.
[0059] The initial value range specifier (320) specifies a range of initial values, learns with arbitrary values within the range, and evaluates using the coefficient of determination. It learns by finding other initial values within the range using Bayesian optimization techniques, and evaluates how well the model explains the data using the coefficient of determination (correlative coefficient, r), which is an indicator. 2The optimal initial value is derived by repeating the act of evaluating with the value. To fit the function “f(x) = ax+b” to the data, initial variable values (a and b) must be input. Generally, when training a model, specific values such as 3 and 100 are input for the initial values of a and b. However, to automate this process, the present invention initially sets a as an array with a constant interval between 1 and 5, and b as an array with a constant logarithmic interval between 10^0 and 10^4. One value is extracted from each variable array to obtain the initial value “a=1, b=10”. The model is trained with this initial value, and the coefficient of determination of the trained model is calculated. Using Bayesian optimization techniques, the next initial value, for example “a=2, b=10”, is found in the array and input to train the model, and the coefficient of determination of the model is calculated. This process is repeated about 100 times to find the initial value for training the model that showed the highest coefficient of determination. That is, the model learning unit (300) of the present embodiment does not determine the initial value randomly but selects it from a certain range to optimize it efficiently, and through Bayesian optimization, reduces unnecessary attempts and finds the initial value that obtains the highest coefficient of determination.
[0060] The model variable estimator (330) estimates the variables of the model such that the deviation between the model and the target pest occurrence amount information is minimized. This can be done by changing the parameters to calculate the mean squared error between the model's predicted value and the pest occurrence amount information, and finding the variable that minimizes the value. At this time, the variables are controlled through constraints, for example, constraints such as ensuring that the difference in the pest occurrence period distance (heat amount) is within 10% except for the last peak period can be applied.
[0061] The outlier remover (340) removes outliers from the data based on the Cox's distance. Here, the Cox's distance refers to the distance between the model and the data. Values that are further than this distance are automatically identified as outliers that do not follow the general trend of the model and only have a negative effect on the model's fit, and are automatically excluded, after which the final model is selected.
[0062] The suitability verifier (350) is a statistical value used to verify the suitability of an insect development model (the model's F value, degree of freedom, P value; each variable value and SEM of the variable value; the model's r 2 It automatically calculates and displays the value. Here, if the model's P-value is greater than 0.05, it returns to the process of distinguishing between the training dataset and the validation dataset and proceeds with the model coefficient estimation again.
[0063] Here, the fit validator (350) first calculates the sum of squares residual of error (SSE), sum of squared due to regression (SSR), and total sum of squares (SST) to obtain the aforementioned statistical figures. SSE, SSR, and SST represent the sum of errors between actual data values and predicted data. The fit validator (350) calculates the mean squared error (MSE) and mean squares due to regression (MSR), which are average values, using SSE, SSR, SST, and the degree of freedom. Additionally, it calculates the F value and P value using the calculated MSR and MSE. Here, it first checks whether the P value is less than 0.05, calculates the variable value and the standard error of mean (SEM) of the variable value, and checks whether the magnitude of the variable value is at least one order of magnitude larger than the SEM. Also, using SSE and SST, r2 The model's fit is verified by calculating . If any of the three verified values do not pass, the process returns to dataset classification. Here, in this embodiment, for example, the aforementioned r 2 The suitability of the model can be verified by checking if this is 0.5 or higher.
[0064] The prediction power verifier (360) automatically converts the 50% occurrence time for each peak period of occurrence in the final model into a Julian day when it finds the cumulative heat amount, and automatically converts the 50% occurrence time for each peak period of occurrence in the verification data into a Julian day to calculate the difference between the predicted time for each peak period of occurrence and the actual time for occurrence and displays it in a table.
[0065] The data storage unit (370) finally stores the final model reflecting the growth zero temperature and growth upper limit temperature, the data with outliers removed, and the corresponding weather data.
[0066] The prediction unit (400) displays model results in the target area by utilizing weather data of the target area, the stored growth zero point temperature, growth upper limit temperature, and the final model. Here, the final model is a verified model, and the prediction unit (400) predicts pest occurrence based on the final model.
[0067] The following describes a pest prediction method utilizing field data according to the present invention with reference to the drawings. Any content described below that overlaps with the description of the pest prediction device according to the present invention mentioned above will be omitted or briefly explained.
[0068] Figure 2 is a flowchart of a pest prediction method using field data according to the present invention.
[0069] As illustrated in FIG. 2, the pest prediction method using field data according to the present invention includes a step of collecting data (S1), a step of preprocessing data (S2), a step of training a model (S3), and a step of predicting pest occurrence (S4).
[0070] The step of collecting data (S1) involves a data collection unit collecting data to be used for predicting pests that may occur in a target area. Here, the step of collecting data (S1) collects weather data and pest data for the target area to predict pest occurrence, and accordingly includes a step of collecting weather data and a step of collecting pest data.
[0071] The data preprocessing step (S2) involves the data preprocessing unit preprocessing the data collected in the data collection step (S1). The data preprocessing step (S2) preprocesses the data to standardize the pest occurrence amount and handle missing values, and then calculates the temperature required for pest growth to obtain the cumulative heat amount. To this end, the data preprocessing step (S2) includes a data transformation step (S2-1), a model selection step (S2-2), and a dataset construction step (S2-3).
[0072] The step of training the model (S3) involves the model training unit training a model to predict pest occurrence. To this end, the step of training the model (S3) includes a step of selecting a Waybull function (S3-1), a step of specifying an initial value range (S3-2), a step of estimating model variables (S3-3), a step of removing outliers (S3-4), a step of verifying goodness of fit (S3-5), and a step of verifying predictive power (S3-6).
[0073] The step of predicting pest occurrence (S4) predicts pest occurrence based on the final model selected and verified in the step of training the prediction unit model (S3).
[0074] As described above, the present invention can easily and accurately predict the occurrence of pests by applying a model based on the annual generation of insects based on weather data and pest data.
[0075] The embodiments described above should be understood as exemplary in all respects and not limiting, and the scope of the invention will be defined by the claims set forth below rather than by the detailed description above. Furthermore, the meaning and scope of the claims set forth below, as well as all modifications and variations derived from equivalents thereof, should be interpreted as being included within the scope of the invention.
[0076] Although the present invention has been described above with reference to the illustrated drawings, the present invention is not limited by the embodiments and drawings disclosed in this specification, and it is obvious that various modifications can be made by a person skilled in the art within the scope of the technical concept of the present invention. Furthermore, even if the effects of the configuration of the present invention were not explicitly described while explaining the embodiments of the present invention above, it is natural to acknowledge that the effects predictable by said configuration should also be recognized.
Claims
1. A data collection unit that collects weather data and pest data of the target area, and A data preprocessing unit that standardizes the amount of pest occurrence using weather data and pest data collected from the above data collection unit, processes missing values, and calculates the temperature required for pest growth to obtain the accumulated heat amount, A model training unit that analyzes pest occurrence patterns and trains a model based on data processed by the above data preprocessing unit, and A pest prediction device utilizing field data, comprising a prediction unit that predicts pest occurrence in the target area based on the final model learned in the above-mentioned model learning unit.
2. In Paragraph 1, The above preprocessing unit is, A data converter that calculates the cumulative heat amount based on the lower and upper temperature limits of pest development for each target area, and converts pest occurrence information into cumulative pest ratio data, and A model selector that selects a model based on the annual number of pest generations based on cumulative heat amount and cumulative ratio data output from the above data converter, and A pest prediction device utilizing field data, comprising a dataset builder for constructing a training dataset for selecting variables of a model selected by the above model selector, and a validation dataset for measuring the prediction accuracy of a model trained with the above training dataset.
3. In Paragraph 2, The model selected in the above model selector is, is, The above n is the annual number of generations of the pest, and is 0 if only overwintering generations exist, and The above x is the cumulative daily temperature, representing the accumulated amount of heat during weather conditions where the temperature is above a certain standard during a specific period, and The above α i is the height of the i-th peak curve, which is a constant controlling the maximum value of the peak curve, and The above β i is the point at which 50% of the i-th peak curve occurs based on the cumulative temperature, and The above γ i A pest prediction device utilizing field data, which is a parameter controlling the steepness of the i-th curve.
4. In Paragraph 3, The above model learning unit is, A Waybull function selector that applies the model selected in the above data preprocessing unit, and An initial value range selector that specifies the range of initial values of the model applied in the above Waybull function selector, trains with arbitrary values within the interval, and evaluates using the coefficient of determination, A model variable estimator that estimates model variables such that the deviation between the model applied in the above-mentioned Waybull function selector and the pest occurrence amount information of the pest data is minimized, and A pest prediction device utilizing field data, comprising an outlier remover that removes outliers based on the distance between the above model and the pest occurrence amount information.
5. In Paragraph 4, The above model learning unit is, Calculate SSE (sum of squares residual of error), SSR (sum of squared due to regression), and SST (total sum of squares); calculate the mean values MSE (mean squared error) and MSR (mean squares due to regression) using the calculated SSE, SSR, SST, and degrees of freedom; calculate the F-value and P-value using the calculated MSR and MSE; verify whether the P-value is less than 0.05; calculate the variable value and its SEM (standard error of mean); and verify whether the magnitude of the variable value is at least one order of magnitude larger than the SEM, while using the SSE and SST, r 2 A conformity validator that calculates a value to verify the conformity of the above model, and A pest prediction device utilizing field data, further comprising a predictive power verifier that, when the 50% occurrence time for each peak period of pest occurrence in the above model is identified as a cumulative heat amount, converts it to a Julian day, and converts the 50% occurrence time for each peak period of pest occurrence in the above verification data to a Julian day to calculate the difference between the predicted time and the actual occurrence time for each peak period of pest occurrence and automatically displays it in a table.
6. A pest prediction method using a pest prediction device according to any one of paragraphs 1 to 5, wherein A step in which a data collection unit collects weather data and pest data of a target area; and, A step in which a data preprocessing unit standardizes the pest occurrence amount and processes missing values using weather data and pest data collected by the above data collection unit, and then calculates the temperature required for pest growth to obtain the accumulated heat amount; A step in which a model training unit analyzes pest occurrence patterns and trains a model based on data processed by the above-mentioned data preprocessing unit; and A pest prediction method utilizing field data comprising: a step in which a prediction unit predicts pest occurrence in the target area based on the final model learned in the model learning unit.
7. In Paragraph 6, The step of the data preprocessing unit standardizing the pest occurrence amount and processing missing values using weather data and pest data collected by the data collection unit, and then calculating the temperature required for pest growth to obtain the accumulated heat amount; A step in which a data converter calculates the cumulative heat amount based on the lower and upper developmental temperature limits of pests by target region, and converts the pest occurrence information into cumulative ratio data of pests; and A step in which a model selector selects a model based on the annual number of pest generations according to the cumulative heat amount and cumulative ratio data output from the above data converter; and A method for predicting pests using field data, comprising the step of a dataset builder constructing a training dataset for selecting variables of a model selected by the above model selector, and a verification dataset for measuring the prediction accuracy of a model trained with the above training dataset.
8. In Paragraph 7, The model selected in the above model selector is, is, The above n is the annual number of generations of the pest, and is 0 if only overwintering generations exist, and The above x is the cumulative daily temperature, representing the accumulated amount of heat during weather conditions where the temperature is above a certain standard during a specific period, and The above α i is the height of the i-th peak curve, which is a constant controlling the maximum value of the peak curve, and The above β i is the point at which 50% of the i-th peak curve occurs based on the cumulative temperature, and The above γ i is a pest prediction method using field data, which is a parameter that controls the steepness of the i-th curve.
9. In Paragraph 8, The step of the model learning unit analyzing pest occurrence patterns and training a model based on the data processed by the above data preprocessing unit; is, A step in which a Waybull function selector applies the model selected in the above data preprocessing unit; and, A step of the initial value range of the model applied in the above-mentioned Waybull function selector being specified by an initial value range specifier, learning with arbitrary values within the interval, and then evaluating using the coefficient of determination; A step in which a model variable estimator estimates model variables such that the deviation between the model applied in the above-mentioned Waybull function selector and the pest occurrence amount information of the pest data is minimized, and A pest prediction method utilizing field data comprising the step of removing outliers using an outlier remover based on the distance between the above model and the pest occurrence amount information.
10. In Paragraph 9, The step of the model learning unit analyzing pest occurrence patterns and training a model based on the data processed by the above data preprocessing unit; is, The validator calculates SSE (sum of squares residual of error), SSR (sum of squared due to regression), and SST (total sum of squares); calculates the mean values MSE (mean squared error) and MSR (mean squares due to regression) using the calculated SSE, SSR, SST, and degrees of freedom; calculates the F value and P value using the calculated MSR and MSE; verifies whether the P value is less than 0.05; calculates the variable value and its SEM (standard error of mean); and verifies whether the magnitude of the variable value is at least one order of magnitude larger than the SEM, while using the SSE and SST, r 2 A step of verifying the suitability of the above model by calculating a value; and A pest prediction method utilizing field data, further comprising the step of: when a prediction power verifier identifies the 50% occurrence time for each peak period of pest occurrence in the above model as a cumulative heat amount, converting it to a Julian day; and converting the 50% occurrence time for each peak period of pest occurrence in the verification data to a Julian day to calculate the difference between the predicted time and the actual occurrence time for each peak period of pest occurrence and automatically displaying it in a table.