CN111985728A - Method for establishing organic sorghum yield prediction model - Google Patents

Abstract

The invention particularly relates to a method for establishing an organic sorghum yield prediction model. The method for establishing an organic sorghum yield prediction model firstly divides the growth period of organic sorghum into three periods: the initial planting period, the growing period and the mature period, and constructs training data sets respectively; Filter out the most important feature variables and train the model; adjust the model variables and evaluate the feasibility of the model until it meets the requirements of the modeler and the reference coefficient for evaluation; finally, save the trained three models of different growth stages separately To the background big data cluster, it is convenient for users to call later. The method for establishing an organic sorghum yield prediction model can only predict the organic sorghum yield according to different growth stages, which improves the prediction speed and prediction accuracy, and can also provide users with early warning prompts, timely remind users to take rescue measures, and provide organic sorghum production. Sorghum provides benchmarks and references for research and judgment.

Description

A kind of method to build organic sorghum yield prediction model

technical field

The invention relates to the technical field of machine learning algorithms, in particular to a method for establishing an organic sorghum yield prediction model.

Background technique

Organic sorghum is one of the raw materials that wineries often use to make wine, and its yield is critical to the operation of the winery. However, the existing prediction of organic sorghum mainly relies on experts to estimate the yield of different plots according to the area. The artificial prediction results are not very accurate, and the error and deviation are large.

Based on this, the present invention proposes a method for establishing an organic sorghum yield prediction model.

SUMMARY OF THE INVENTION

In order to make up for the defects of the prior art, the present invention provides a simple and efficient method for establishing an organic sorghum yield prediction model.

The present invention is achieved through the following technical solutions:

A method for establishing an organic sorghum yield prediction model, comprising the following steps:

The first step is to build a training dataset

The growth cycle of organic sorghum is divided into three periods: the initial planting period, the growing period and the mature period, and the basic attributes of planting, the natural environment data during the planting period and the historical data of field management data are used as the initial explanatory variables of the model. The historical yield records of the block were used as the response variable of the model, and were aggregated into three model training data sets according to the growth stage of organic sorghum;

The second step is data preprocessing

Preprocess the three training data sets, including observing abnormal data points in the data set, and eliminating suspicious human error input information;

The third step, variable filtering

In order to improve the stability and prediction accuracy of the model, the preprocessed data is filtered, and the 15 most important feature variables are screened out, and the feature variables are tested for collinearity;

The fourth step is to train the model

The organic sorghum yield number was set as the response variable, and other characteristic variables that affected the organic sorghum yield were used as the explanatory variables of the model, and the training data of three organic sorghums in different growth stages were used as the training data set to train the model;

The fifth step, model diagnosis

Observe and adjust the model variables and evaluate the feasibility of the model, and continuously adjust the cycle until the requirements of the modeler and the evaluation reference coefficient are met;

The sixth step, model application

The three models of the trained organic sorghum yield in the early planting, growing and mature stages are saved to the backend big data cluster, which is convenient for users to call later.

In the first step, the training dataset consists of the following three parts:

The basic attributes of planting, including land area, soil temperature, soil moisture, soil pH and content of special chemical substances, are obtained through manual records and soil detector testing;

The natural environment data during the planting period, including recent temperature changes, lighting time, and the degree of influence of special natural disasters, are collected through the local meteorological observation station and the base station where the plot is located;

Field management data, records of the amount of fertilizer used on the land, the amount of seeds used, the number of pests and weeds, and the number of on-site support by agricultural technicians were obtained through on-site collection.

In the second step, when there is vacancy in a certain column of feature data in the training data set, if the vacant column accounts for less than 5% of the total columns, select the average value of the corresponding column to be filled in the vacant column, otherwise delete the column of feature data or Fill empty columns manually.

In the third step, the concept of variance expansion coefficient in mathematical statistics is used to measure the degree of collinearity of variables in regression modeling, and variables with a variance expansion coefficient exceeding 10 are excluded to obtain three training data of organic sorghum in different growth stages. As a training dataset for organic sorghum yield at different times.

In the described fourth step, the organic sorghum yield prediction model is written using the Statsmodels professional regression modeling extension tool language package in the Python language, and the training data of three organic sorghum in different growth stages are used as the training data set for the respective models. Training was performed to obtain three organic sorghum yield prediction models at different growth stages.

In the fifth step, the variable adjustment and evaluation of the model includes the following steps:

S1. Use the residual diagnostic chart to observe the fitting degree between the constructed model and the data, observe the outlier abnormal points and the sample points with strong influence, and eliminate the outlier abnormal points with strong influence in the process of training the model ;

S2. Evaluate whether the fitting power of the variable is reasonable by drawing a GAM (Generalized Additive Model) graph in statistics for the important variable indicators in the model and add an alternative model scheme;

S3. Eliminate the characteristic variables with the T-test P value greater than 0.2 in the statistical diagnosis results of the model, because the T-test shows that there is no significant linear relationship between higher characteristic variables and yield;

S4. Calculate the adjusted coefficient of determination (Adjusted R-Square) of the candidate model, AIC (Akaikeinformation criterion, Akaike information criterion) Akaike information criterion and the sum of the mean square errors obtained by the model in the test data set, and finally compare and select the interpretation of the original data The model with the highest strength and fit.

In the sixth step, the basic attributes of planting, the natural environment data and field management data during the planting period are continuously monitored during the growth stage of sorghum, and the monitoring data are input into the characteristic variables of the model, and the growth stage input by the user is used. Call the organic sorghum yield prediction model of the corresponding growth stage to make regular predictions on the yield of organic sorghum;

If the predicted value is less than the user's demand or experience value, an early warning message will be issued to remind the user to make timely remediation for the abnormal situation that occurs.

In the sixth step, if the purchase volume is more than 15% higher than the predicted output during the grain harvesting stage, it is considered that the farmer may have borrowed or shoddy goods, and the model will issue a system reminder.

The beneficial effects of the invention are as follows: the method for establishing the organic sorghum yield prediction model can only predict the organic sorghum yield according to different growth stages, which improves the prediction speed and prediction accuracy, and can also provide users with early warning prompts and timely reminders Users took rescue measures to provide benchmarks and references for organic sorghum.

Description of drawings

In order to illustrate the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are For some embodiments of the present invention, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative efforts.

Accompanying drawing 1 is the schematic diagram of the method for establishing the organic sorghum yield prediction model of the present invention.

Detailed ways

In order to make those skilled in the art better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

The method for establishing an organic sorghum yield prediction model includes the following steps:

The first step is to build a training dataset

According to the growth characteristics of organic sorghum, this model invention divides the growth cycle of organic sorghum into three periods: the initial stage of planting, the There are three periods of growth and maturity, and the basic attributes of planting, the natural environment data and the historical data of field management data during the planting period are used as the initial explanatory variables of the model, and the historical yield records of the planting plots are used as the response variables of the model. According to the growth stage of organic sorghum, they were aggregated into three model training data sets;

The second step is data preprocessing

Preprocess the three training data sets, including observing abnormal data points in the data set, and eliminating suspicious human error input information;

The third step, variable filtering

In order to improve the stability and prediction accuracy of the model, the preprocessed data is filtered, and the 15 most important feature variables are screened out, and the feature variables are tested for collinearity;

The fourth step is to train the model

The organic sorghum yield number was set as the response variable, and other characteristic variables that affected the organic sorghum yield were used as the explanatory variables of the model, and the training data of three organic sorghums in different growth stages were used as the training data set to train the model;

The fifth step, model diagnosis

Observe and adjust the model variables and evaluate the feasibility of the model, and continuously adjust the cycle until the requirements of the modeler and the evaluation reference coefficient are met;

The sixth step, model application

The three prediction models of the trained organic sorghum yield in the early planting, growing and mature stages are saved to the background big data cluster, which is convenient for users to call later.

In the first step, the training dataset consists of the following three parts:

The basic attributes of planting, including land area, soil temperature, soil moisture, soil pH and content of special chemical substances, are obtained through manual records and soil detector testing;

The natural environment data during the planting period, including recent temperature changes, lighting time, and the degree of influence of special natural disasters, are collected through the local meteorological observation station and the base station where the plot is located;

Field management data, records of the amount of fertilizer used on the land, the amount of seeds used, the number of pests and weeds, and the number of on-site support by agricultural technicians were obtained through on-site collection.

In the second step, when there is vacancy in a certain column of feature data in the training data set, if the vacant column accounts for less than 5% of the total columns, select the average value of the corresponding column to be filled in the vacant column, otherwise delete the column of feature data or Fill empty columns manually.

In the third step, the concept of variance expansion coefficient in mathematical statistics is used to measure the degree of collinearity of variables in regression modeling, and variables with a variance expansion coefficient exceeding 10 are excluded to obtain three training data of organic sorghum in different growth stages. As a training dataset for organic sorghum yield at different times.

In the described fourth step, the organic sorghum yield prediction model is written using the Statsmodels professional regression modeling extension tool language package in the Python language, and the training data of three organic sorghum in different growth stages are used as the training data set for the respective models. Training was performed to obtain three organic sorghum yield prediction models at different growth stages.

In the fifth step, the variable adjustment and evaluation of the model includes the following steps:

S1. Use the residual diagnostic chart to observe the fitting degree between the constructed model and the data, observe the outlier abnormal points and the sample points with strong influence, and eliminate the outlier abnormal points with strong influence in the process of training the model ;

S2. Evaluate whether the fitting power of the variable is reasonable by drawing a GAM (Generalized Additive Model) graph in statistics for the important variable indicators in the model and add an alternative model scheme;

S3. Eliminate the characteristic variables with higher T test (P value greater than 0.2) in the statistical diagnosis results of the model, because the T test shows that there is no significant linear relationship between the higher characteristic variables and the yield;

S4. Calculate the adjusted coefficient of determination (Adjusted R-Square) of the candidate model, AIC (Akaikeinformation criterion, Akaike information criterion) Akaike information criterion and the sum of the mean square errors obtained by the model in the test data set, and finally compare and select the interpretation of the original data The model with the highest strength and fit.

Based on the next year's brewing plan, the user infers the demand for organic sorghum in the next year to make a planting plan. In order to ensure that the dosage requirements are met, in the sixth step, the basic attributes of planting, the natural environment data and field management data during the planting period are continuously monitored in the growth stage of sorghum, and variables are entered on the page according to the characteristic variables in the model. and add the growth stage of the organic sorghum under the current calculation plot; after reading the necessary parameters, the background task calls the organic sorghum yield prediction model of the corresponding growth stage to regularly predict the yield of organic sorghum;

If the predicted value is less than the user's demand or experience value, an early warning message will be issued to remind the user to make timely remediation for the abnormal situation that occurs.

In the sixth step, if the purchase volume is more than 15% higher than the predicted output during the grain harvesting stage, it is considered that the farmer may have borrowed or shoddy goods, and the model will issue a system reminder.

Compared with the current prior art, the method for establishing an organic sorghum yield prediction model has the following characteristics:

First, considering the growth stage of organic sorghum, three training data sets were constructed, respectively, and models were constructed for the yield of organic sorghum in three different periods, so that the model could learn the effects of organic sorghum in different periods in a targeted manner. The degree of influence of various indicators.

Second, the function of result comparison is added, that is, when the user predicts the future yield of organic sorghum through the constructed model, if the result is lower than the user's demand or experience value, the user will be provided with an early warning prompt, which can timely remind the user to Rescue measures should be taken for the insufficient yield of the corresponding plots.

Third, through machine learning, the regression modeling method in the field of statistics provides customers with a set of rules for judging the yield of organic sorghum, that is, to determine the main factors affecting the yield of organic sorghum. After rigorous screening, these main factors have the characteristics of being observable, adjustable, and simulating, which provide benchmarks and references for subsequent research on organic sorghum and other crops.

The above-mentioned embodiment is only one of the specific embodiments of the present invention, and the usual changes and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included in the protection scope of the present invention.

Claims (8)

1. a method for establishing an organic sorghum yield prediction model, is characterized in that, comprises the following steps: The first step is to build a training dataset The growth cycle of organic sorghum is divided into three periods: the initial planting period, the growing period and the mature period, and the basic attributes of planting, the natural environment data during the planting period and the historical data of field management data are used as the initial explanatory variables of the model. The historical yield records of the block were used as the response variable of the model, and were aggregated into three model training data sets according to the growth stage of organic sorghum; The second step is data preprocessing Preprocess the three training data sets, including observing abnormal data points in the data set, and eliminating suspicious human error input information; The third step, variable filtering In order to improve the stability and prediction accuracy of the model, the preprocessed data is filtered, and the 15 most important feature variables are screened out, and the feature variables are tested for collinearity; The fourth step is to train the model The organic sorghum yield number was set as the response variable, and other characteristic variables that affected the organic sorghum yield were used as the explanatory variables of the model, and the training data of three organic sorghums in different growth stages were used as the training data set to train the model; The fifth step, model diagnosis Observe and adjust the model variables and evaluate the feasibility of the model, and continuously adjust the cycle until the requirements of the modeler and the evaluation reference coefficient are met; The sixth step, model application The three models of the trained organic sorghum yield in the early planting, growing and mature stages are saved to the backend big data cluster, which is convenient for users to call later. 2. the method for establishing organic sorghum yield prediction model according to claim 1, is characterized in that: in the described first step, training data set is made up of following three parts: The basic attributes of planting, including land area, soil temperature, soil moisture, soil pH and content of special chemical substances, are obtained through manual records and soil detector testing; The natural environment data during the planting period, including recent temperature changes, lighting time, and the degree of influence of special natural disasters, are collected through the local meteorological observation station and the base station where the plot is located; Field management data, records of the amount of fertilizer used on the land, the amount of seeds used, the number of pests and weeds, and the number of on-site support by agricultural technicians were obtained through on-site collection. 3. the method for establishing organic sorghum yield prediction model according to claim 1 and 2, is characterized in that: in the described second step, when a certain column of characteristic data in the training data set has a vacancy situation, if the vacant column accounts for the total column 5% or less, choose to fill the average value of the corresponding column in the vacant column, otherwise delete the characteristic data of the column or manually fill the vacant column. 4. the method for establishing organic sorghum yield prediction model according to claim 1 and 2, is characterized in that: in the described 3rd step, use the variance expansion coefficient concept in mathematical statistics to measure the collinearity of variable in regression modeling Severity, excluding variables with a variance expansion coefficient exceeding 10, three training data of organic sorghum in different growth stages were obtained as the training data set of organic sorghum yield in different periods. 5. the method for establishing organic sorghum yield prediction model according to claim 4, is characterized in that: in the described 4th step, adopt the Statsmodels professional regression modeling expansion tool language package in the Python language to carry out the organic sorghum yield prediction model. Write and use the training data of three organic sorghum in different growth stages as training data sets to train the respective models, thereby obtaining three organic sorghum yield prediction models in different growth stages. 6. the method for setting up organic sorghum yield prediction model according to claim 1 and 4, is characterized in that: in the described 5th step, to the variable adjustment of model and evaluation, comprise the following steps: S1. Use the residual diagnostic chart to observe the fitting degree between the constructed model and the data, observe the outlier abnormal points and the sample points with strong influence, and eliminate the outlier abnormal points with strong influence in the process of training the model ; S2. Evaluate whether the fitting power of the variable is reasonable by drawing the GAM diagram in statistics for the important variable indicators in the model and add an alternative model scheme; S3. Eliminate the characteristic variables with the T-test P value greater than 0.2 in the statistical diagnosis results of the model, because the T-test shows that there is no significant linear relationship between higher characteristic variables and yield; S4. Calculate the correction coefficient of determination of the candidate model, the AIC Akaike information criterion and the sum of the mean square errors obtained by the model in the test data set, and finally compare and select the model with the highest degree of interpretation and fit to the original data. 7. the method for setting up organic sorghum yield prediction model according to claim 2 or 6, is characterized in that: in the described 6th step, in the growth stage of sorghum to the basic attribute of planting, the natural environment data of planting period and field The management data is continuously monitored, and the monitoring data is input into the characteristic variables of the model, and the organic sorghum yield prediction model of the corresponding growth stage is called according to the growth stage input by the user to regularly predict the yield of organic sorghum; If the predicted value is less than the user's demand or experience value, an early warning message will be issued to remind the user to make timely remediation for the abnormal situation that occurs. 8. the method for establishing an organic sorghum yield prediction model according to claim 7, is characterized in that: in the described 6th step, if in the grain harvest stage, the purchase amount is higher than the predicted yield by more than 15%, then think that this farmer may be possible. If there is borrowing or shoddy charging, the model will issue a system reminder.

Images