CN113762394A - Blasting block size prediction method - Google Patents
Blasting block size prediction method Download PDFInfo
- Publication number
- CN113762394A CN113762394A CN202111054833.2A CN202111054833A CN113762394A CN 113762394 A CN113762394 A CN 113762394A CN 202111054833 A CN202111054833 A CN 202111054833A CN 113762394 A CN113762394 A CN 113762394A
- Authority
- CN
- China
- Prior art keywords
- random forest
- layer
- blasting
- training
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000005422 blasting Methods 0.000 title claims abstract description 66
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000007637 random forest analysis Methods 0.000 claims abstract description 63
- 238000012549 training Methods 0.000 claims abstract description 56
- 238000003066 decision tree Methods 0.000 claims abstract description 31
- 238000012952 Resampling Methods 0.000 claims abstract description 4
- 238000012360 testing method Methods 0.000 claims description 20
- 230000011218 segmentation Effects 0.000 claims description 12
- 239000011435 rock Substances 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 238000005553 drilling Methods 0.000 claims description 7
- 238000012935 Averaging Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 230000000694 effects Effects 0.000 claims description 5
- 238000004088 simulation Methods 0.000 claims description 5
- 230000000903 blocking effect Effects 0.000 claims description 4
- 238000011065 in-situ storage Methods 0.000 claims description 3
- 238000013138 pruning Methods 0.000 claims description 3
- 239000000463 material Substances 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 230000010365 information processing Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- -1 noise Substances 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000035939 shock Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/02—Agriculture; Fishing; Forestry; Mining
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Tourism & Hospitality (AREA)
- Artificial Intelligence (AREA)
- Marketing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Agronomy & Crop Science (AREA)
- Health & Medical Sciences (AREA)
- Mining & Mineral Resources (AREA)
- Primary Health Care (AREA)
- Marine Sciences & Fisheries (AREA)
- Animal Husbandry (AREA)
- General Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a blasting block degree prediction method, and belongs to the technical field of blasting block degree prediction. Firstly, constructing a CART decision tree model; then resampling a training sample set; writing codes by using MATLAB scripting language to calculate the relation between different decision tree trees (Ntree) and model Mean Square Error (MSE); then, a GINI coefficient is adopted as a standard of attribute splitting for the decision tree, and a node is selected based on the standard; obtaining different decision trees based on a training set of nodes, and finally generating a first layer of random forest; generating a second layer of random forest according to the training residual error of the first layer of random forest as a training sample set of the second layer of random forest; and (3) sequentially superposing the outputs of the first layer random forest model and the second layer random forest model to obtain the final output of the double-layer random forest prediction model, and predicting the blasting blockiness according to the final output. Compared with the traditional prediction method, the method has higher reliability and stability.
Description
Technical Field
The invention relates to a blasting block degree prediction method, and belongs to the technical field of blasting block degree prediction.
Background
The drilling and blasting method is one of rock breaking methods frequently adopted in mining and tunnel construction, and the blasting effect brings many negative effects such as blasting vibration, air shock waves, flying rocks, noise, dust and the like while the rock breaking is completed, wherein the blasting vibration damage is particularly obvious. The blasting regulation in China adopts speed to measure the vibration intensity, and the damage of blasting vibration can be effectively controlled by accurately predicting the particle vibration speed caused by blasting.
The accuracy and reliability of the blasting block degree prediction method in the prior art are low.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method for predicting the blasting block degree is provided, the problems of inaccurate and fuzzy information processing needing to consider a plurality of factors and conditions at the same time are solved, and the accuracy and the reliability of the blasting block degree prediction are improved.
The technical scheme adopted by the invention is as follows: a blasting blockiness prediction method comprises the following steps:
1) forming a training data set;
2) selecting a split node;
3) generating a first layer of random forest;
4) generating a second layer of random forest;
5) and the outputs of the first layer of random forest and the second layer of random forest are sequentially superposed to obtain the final output of the double-layer random forest prediction model.
Specifically, the step 1) includes the steps of:
step1.1: the method comprises the steps of taking the blasting average blockiness value actually measured by a blasting test as an output variable (label value) of a model, acquiring indexes of resisting line distance (B), drilling row spacing (S), step height (H), drilling diameter (d), blocking length (L), rock mass elasticity modulus (E), in-situ rock blockiness (x) and blasting unit consumption (q) as input variables (attribute values) of the model according to corresponding blasting test fields, and forming a training data set by the output variable and the input variables
Step1.2: resampling a training sample set by using a Bootstrap method for the data set obtained from Step1.1, extracting m data samples, and randomly generating n training data setsD, D ═ xi1,xi2,…,xin,yi}(i∈[1,m])。
On the basis of a Bootstrap data set, a Bagging method is adopted to select training data which are randomly replaced, then a classifier is constructed, and finally the overall acquisition effect is increased by combining the learned models.
Step1.3: using a decision tree algorithm in each subset, selecting an optimal mode to split nodes according to a 'Kernian coefficient minimum criterion', not pruning in the splitting process, and setting a single decision tree predictor f (x, theta)k) Has a prediction result of fi(x) The final prediction result of the random forest regression model is expressed as that formula (1) defines a random forest regression algorithm modeling process parameter set RFP; and (3) the formula (2) is used for predicting the blasting blockiness based on random forest regression.
RFP={Ntree,Mtry} (1)
Where x represents the input vector, θkIs a vector representing the growth path for each tree generated, and f (x) represents the predicted shot blockiness. The Ntree is the number of decision tree trees in the model, and the Ntree value influences the training degree and the accuracy of the random forest model. In order to obtain the best evaluation result, a MATLAB script language is used for writing codes and carrying out simulation calculation on the relationship between different decision tree (Ntree) trees and the Mean Square Error (MSE) of the model. Mtry is the number of features randomly drawn from the features. The Mtry value controls the disturbance degree of the random forest model attribute and is an important parameter in the model. The Mtry value directly affects the accuracy of the model. According to the following experienceThe equation calculates the Mtry value:
Mtry&=[log2M] (3)
Mtry&=[M/3 (4)
in the formula: m is the number of model input parameters; [] Indicating a rounding down operation.
Specifically, the step 2) includes the following steps:
step2.1: a decision tree in the form of a binary tree is used, and a binary recursion is utilized to continuously divide a data space into different subsets. When classifying, assuming that there are K classes, the probability distribution has a kini index of:
where p denotes the probability of a sample point, pkIndicating the probability that the sample point belongs to class K.
Step2.2: the decision tree uses the GINI coefficient as a criterion for attribute splitting. And selecting the feature with the lowest kini coefficient as a root node. And in the same way, selecting the leaf node with the smallest residual characteristic kini coefficient.
Step2.3: and (3) recursively performing the step3 on each node from the root node according to the training data set to construct a binary decision tree.
Specifically, the step 3) includes the steps of:
step3.1: the subset D is divided into D according to whether the characteristic A takes a certain possible value a1And D2Two moieties, D1={(x,y)∈D|A(x)=a},D2=D-D1Then, under the condition of the feature A, the Gini index of the set D is defined as:
wherein D1Representing a training set of nodes, and calculating the kini index of the existing features to the data set.
Step3.2: for each feature a, a is taken for each possible value thereof, according to the test of the sample point pair a ═ a"Yes" or "No" partition D into D1And D2In two parts, the kini index when a ═ a is calculated. Among all possible features a and all possible cut points a, the cut point with the smallest kini index is selected as the cut point. And generating two child nodes from the current segmentation point according to the optimal feature and the optimal segmentation point, and distributing the training data set into two leaf nodes according to the feature.
Step3.3: step2.3 to step3.2 are recursively called for the two leaf nodes.
Step3.4: and (3) repeating the steps 3.1 to 3.3 to obtain different decision trees, wherein each decision tree recursively branches from top to bottom to grow, the regression trees stop growing after the segmentation termination condition is met, and finally all the regression trees are combined together to form a random forest.
Step3.5: inputting the input vector X in the prediction set X into the first layer of random forest, and performing regression on a single regression tree TiPredicted value y ofiArithmetic averaging using the "simple averaging method
Specifically, the step 4) includes the following steps:
step4.1: computing output of layer 1 random forest on training sample setUsing actual values y of training samplesiSubtracting the output valueThe training residual error can be obtainedWill train residual errorSubstituting the data set into the original training sample to construct a new data set as the training sample of the layer 2 random forest, and inputting the new data set as xiThe desired output is
Step4.2: because the value of the training residual error obtained by Step4.1 is too small, the training residual error needs to be normalized:
in the formula: linIs original input data; lminThe minimum of the same kind of data in the original input data; lmaxThe maximum value of the same kind of data in the original input data is obtained; and l is input data after the normalization processing. And forming a training data set of the layer 2 random forest after normalization processing.
Step4.3: and establishing a layer 2 random forest model according to Step1.1 to Step4.1 to obtain a prediction result of the layer 2 random forest model.
The invention has the beneficial effects that: the invention is particularly suitable for processing inaccurate and fuzzy information processing problems which need to consider many factors and conditions simultaneously, and has higher reliability and stability compared with the traditional prediction method.
Drawings
FIG. 1 is a flow chart of a calculation of a prediction value of blasting blockiness based on random forest regression;
FIG. 2 is a schematic diagram of bench blasting parameters;
FIG. 3 is a graph of a scatter point matrix of the raw data set;
FIG. 4 is a comparison of measured and predicted block sizes for a blasting test;
fig. 5 shows the relative error between the measured block size and the predicted block size in the blasting test.
Detailed Description
The invention is further described below with reference to the accompanying drawings and specific embodiments.
Example 1: the random forest is used as an advanced algorithm tool, is particularly suitable for processing the problems of inaccurate and fuzzy information processing which need to consider a plurality of factors and conditions at the same time, introduces a random forest regression method into blasting vibration speed prediction, and compares the result predicted by a random forest model with the result predicted by a traditional method so as to provide an optimal prediction method for blasting blockiness monitoring.
As shown in fig. 1, a method for predicting blasting block size includes the following steps:
1) forming a training data set;
2) selecting a split node;
3) generating a first layer of random forest;
4) generating a second layer of random forest;
5) and the outputs of the first layer of random forest and the second layer of random forest are sequentially superposed to obtain the final output of the double-layer random forest prediction model.
The specific implementation process is as follows: the method aims to solve the problems of low model prediction precision, poor model generalization capability and the like in the current research on blasting material block size prediction, is difficult to accurately control the rockfill material block size, and meets the dam-up condition of blasting mining rockfill materials. Aiming at the defects of the current blasting prediction model, the blasting block degree of the rockfill dam material is effectively controlled, and the blasting block degree prediction model is established based on a random forest regression method.
Further, the step 1) comprises the following steps:
step1.1: the method comprises the steps of taking the blasting average blockiness value actually measured by a blasting test as an output variable (label value) of a model, acquiring indexes of resisting line distance (B), drilling row spacing (S), step height (H), drilling diameter (d), blocking length (L), rock mass elasticity modulus (E), in-situ rock blockiness (x) and blasting unit consumption (q) as input variables (attribute values) of the model according to corresponding blasting test fields, and forming a training data set by the output variable and the input variablesThe schematic diagram of the step blasting parameters is shown in figure 2.
Step1.2: resampling a training sample set by using a Bootstrap method for the data set obtained from Step1.1, extracting m data samples, and randomly generating n training data setsD, D ═ xi1,xi2,…,xin,yi}(i∈[1,m])。
On the basis of a Bootstrap data set, a Bagging method is adopted to select training data which are randomly replaced, then a classifier is constructed, and finally the overall acquisition effect is increased by combining the learned models.
Step1.3: using a decision tree algorithm in each subset, selecting an optimal mode to split nodes according to a 'Kernian coefficient minimum criterion', not pruning in the splitting process, and setting a single decision tree predictor f (x, theta)k) Has a prediction result of fi(x) The final prediction result of the random forest regression model is expressed as that formula (1) defines a random forest regression algorithm modeling process parameter set RFP; and (3) the formula (2) is used for predicting the blasting blockiness based on random forest regression.
RFP={Ntree,Mtry} (1)
Where x represents the input vector, θkIs a vector representing the growth path for each tree generated, and f (x) represents the predicted shot blockiness. The Ntree is the number of decision tree trees in the model, and the Ntree value influences the training degree and the accuracy of the random forest model. In order to obtain the best evaluation result, a MATLAB script language is used for writing codes and carrying out simulation calculation on the relationship between different decision tree (Ntree) trees and the Mean Square Error (MSE) of the model. Mtry is the number of features randomly drawn from the features. The Mtry value controls the disturbance degree of the random forest model attribute and is an important parameter in the model. The Mtry value directly affects the accuracy of the model. The Mtry value is calculated according to the following empirical formula:
Mtry&=[log2M] (3)
Mtry&=[M/3 (4)
in the formula: m is the number of model input parameters; [] Indicating a rounding down operation.
Further, the step 2) comprises the following steps:
step2.1: a decision tree in the form of a binary tree is used, and a binary recursion is utilized to continuously divide a data space into different subsets. When classifying, assuming that there are K classes, the probability distribution has a kini index of:
where p denotes the probability of a sample point, pkIndicating the probability that the sample point belongs to class K.
Step2.2: the decision tree uses the GINI coefficient as a criterion for attribute splitting. And selecting the feature with the lowest kini coefficient as a root node. And in the same way, selecting the leaf node with the smallest residual characteristic kini coefficient.
Step2.3: and (3) recursively performing the step3 on each node from the root node according to the training data set to construct a binary decision tree.
Further, the step 3) comprises the following steps:
step3.1: the subset D is divided into D according to whether the characteristic A takes a certain possible value a1And D2Two moieties, D1={(x,y)∈D|A(x)=a},D2=D-D1Then, under the condition of the feature A, the Gini index of the set D is defined as:
wherein D1Representing a training set of nodes, and calculating the kini index of the existing features to the data set.
Step3.2: for each feature A, a value is taken for each possible feature A, and D is divided into D according to the test yes or no of the sample point pair A ═ a1And D2In two parts, the kini index when a ═ a is calculated. Among all possible features a and all possible cut points a, the cut point with the smallest kini index is selected as the cut point. And generating two child nodes from the current segmentation point according to the optimal feature and the optimal segmentation point, and distributing the training data set into two leaf nodes according to the feature.
Step3.3: step2.3 to step3.2 are recursively called for the two leaf nodes.
Step3.4: and (3) repeating the steps 3.1 to 3.3 to obtain different decision trees, wherein each decision tree recursively branches from top to bottom to grow, the regression trees stop growing after the segmentation termination condition is met, and finally all the regression trees are combined together to form a random forest.
Step3.5: inputting an input vector X in the prediction set X into the first-layer random forest, and predicting a single regression tree by using a prediction value yiArithmetic averaging using the "simple averaging method
Further, the step 4) comprises the following steps:
step4.1: computing output of layer 1 random forest on training sample setUsing actual values y of training samplesiSubtracting the output valueThe training residual error can be obtainedWill train residual errorSubstituting the data set into the original training sample to construct a new data set as the training sample of the layer 2 random forest, and inputting the new data set as xiThe desired output is
Step4.2: because the value of the training residual error obtained by Step4.1 is too small, the training residual error needs to be normalized:
in the formula: linIs original input data; lminThe minimum of the same kind of data in the original input data; lmaxThe maximum value of the same kind of data in the original input data is obtained; and l is input data after the normalization processing. And forming a training data set of the layer 2 random forest after normalization processing.
Step4.3: and establishing a layer 2 random forest model according to Step1.1 to Step4.1 to obtain a prediction result of the layer 2 random forest model.
Further, based on the statistical analysis results of the training sample set, table 1 shows the Range (Range), Mean (Mean) and standard deviation (std.dev) indexes of each modeling parameter, respectively. As can be seen from fig. 2, the significant correlation between the parameters is not shown, which indicates that the selected parameters have independence.
TABLE 1 random forest regression model training sample input and output parameter descriptive statistics
Case (2):
in order to further illustrate the accuracy and reliability of the method of the present invention, a bench blasting experiment is taken as an example for illustration. Obtaining median bulkiness of each field of 0.200m, 0.135m, 0.105m, 0.115m and 0.122m through a screening test after a field blasting test and a grading curve of actually measured blasting bulkiness of each field; and calculating to obtain the input parameters of the field blasting test model according to the blasting design parameters, the rock information parameters and the explosive information parameters adopted by the field blasting test, as shown in table 2.
TABLE 2 field blasting test model input parameters
According to the established blasting block degree prediction system based on the RF-R method, the invention respectively carries out 10 times of simulation prediction calculation on 5 groups of blasting tests of the experiment, and finally obtains the prediction block degree average value of each test field, and the result is shown in Table 3.
TABLE 3 simulation experiment prediction of blockiness results
Through calculation, the relative errors of the prediction block value and the actual block value of each blasting test field are respectively as follows: 0.60%, 1.04%, 12.76%, 2.63%, 2.80%, with a maximum error of 12.76%, a minimum error of 0.60%, and an average error of 3.96%.
For example, fig. 4 shows a prediction block value and an actual block value of each shot test session, and fig. 5 shows a relative error of the block values.
In conclusion, the blasting block degree prediction accuracy based on the RF-R model is higher than that of other prediction models, the maximum relative error aiming at the experiment is not more than 15%, and the engineering requirements are met. Therefore, through the quantitative relation between the blasting parameters and the blasting block degree, the blasting parameters can be properly adjusted according to the blasting block degree value predicted by the RF-R model and the system and the actual situation so as to meet the requirement of the blocking degree of the dam.
While the present invention has been described in detail with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, and various changes can be made without departing from the spirit and scope of the present invention.
Claims (5)
1. A blasting block size prediction method is characterized in that: the method comprises the following steps:
1) forming a training data set;
2) selecting a split node;
3) generating a first layer of random forest;
4) generating a second layer of random forest;
5) and the outputs of the first layer of random forest and the second layer of random forest are sequentially superposed to obtain the final output of the double-layer random forest prediction model.
2. The method of claim 1, wherein the method comprises: the step 1) comprises the following steps:
step1.1: taking the blasting average blockiness value actually measured by the blasting test as an output variable of the model, acquiring indexes of resisting line distance B, drilling row spacing S, step height H, drilling diameter d, blocking length L, rock mass elastic modulus E, in-situ rock blockiness x and blasting unit consumption q as input variables of the model according to corresponding blasting test fields, and forming a training data set by the output variable and the input variables
Step1.2: resampling a training sample set by using a Bootstrap method for the data set obtained from Step1.1, extracting m data samples, and randomly generating n training data setsD, D ═ xi1,xi2,…,xin,yi}(i∈[1,m]);
On the basis of a Bootstrap data set, randomly and repeatedly selecting training data by adopting a Bagging method, then constructing a classifier, and finally increasing the overall acquisition effect by combining the learned models;
step1.3: using a decision tree algorithm in each subset, selecting an optimal mode to split nodes according to a 'Kernian coefficient minimum criterion', not pruning in the splitting process, and setting a single decision tree predictor f (x, theta)k) Has a prediction result of fi(x) The final prediction result of the random forest regression model is expressed as that formula (1) defines a random forest regression algorithm modeling process parameter set RFP; the formula (2) is used for predicting the blasting blockiness based on random forest regression;
RFP={Ntree,Mtry} (1)
where x represents the input vector, θkThe method comprises the following steps of representing a vector for generating a growth path of each tree, F (x) representing a predicted blasting block degree, taking Ntree as the number of decision tree trees in a model, writing codes by using MATLAB script language, carrying out simulation calculation on the relationship between the Ntree of different decision tree trees and the mean square error MSE of the model, taking Mtry as the number of features randomly extracted from the features, controlling the disturbance degree of the random forest model attribute by using the Mtry value, and calculating the Mtry value according to the following empirical formula:
Mtry&=[log2M] (3)
Mtry&=[M/3] (4)
in the formula: m is the number of model input parameters; [] Indicating a rounding down operation.
3. The method of claim 2, wherein the step of predicting the blasting block size comprises: the step 2) comprises the following steps:
step2.1: adopting a binary tree form decision tree, continuously dividing a data space into different subsets by utilizing binary recursion, and assuming that K classes exist during classification, the probability distribution has the following kiney indexes:
where p denotes the probability of a sample point, pkRepresenting the probability that the sample point belongs to class K;
step2.2: the decision tree adopts the GINI coefficient as the standard of attribute splitting, selects the feature with the lowest basic coefficient as a root node, and so on, selects the feature with the lowest basic coefficient of the other features as a leaf node;
step2.3: and 3) recursively carrying out the step 3) on each node from the root node according to the training data set to construct a binary decision tree.
4. The method of claim 3, wherein the step of predicting the blasting block size comprises: the step 3) comprises the following steps:
step3.1: the subset D is divided into D according to whether the characteristic A takes a certain possible value a1And D2Two moieties, D1={(x,y)∈D|A(x)=a},D2=D-D1Then, under the condition of feature a, the kini index of the subset D is defined as:
wherein D1Representing a training set of nodes, and calculating the kini index of the existing characteristics to the data set;
step3.2: for each feature A, a value is taken for each possible feature A, and D is divided into D according to the test yes or no of the sample point pair A ═ a1And D2Calculating the King index when A is a; selecting the segmentation point with the minimum Gini index as the segmentation point from all possible characteristics A and all possible segmentation points a, generating two child nodes from the current segmentation point according to the optimal characteristics and the optimal segmentation point, and distributing the training data set into two leaf nodes according to the characteristics;
step3.3: recursively invoke step2.3 through step3.2 for the two leaf nodes;
step3.4: repeating the steps 3.1 to 3.3 to obtain different decision trees, wherein each decision tree recursively branches from top to bottom to grow, the regression trees stop growing after the segmentation termination condition is met, and finally all the regression trees are combined together to form a random forest;
5. The method of claim 4, wherein the step of predicting the blasting block size comprises: the step 4) comprises the following steps:
step4.1: computing output of layer 1 random forest on training sample setUsing actual values y of training samplesiSubtracting the output valueThe training residual error can be obtainedWill train residual errorSubstituting the data set into the original training sample to construct a new data set as the training sample of the layer 2 random forest, and inputting the new data set as xiThe desired output is
Step4.2: because the value of the training residual error obtained by Step4.1 is too small, the training residual error needs to be normalized:
in the formula: linIs original input data; lminThe minimum of the same kind of data in the original input data; lmaxThe maximum value of the same kind of data in the original input data is obtained; l is input data after normalization processing, and a training data set of the layer 2 random forest is formed after normalization processing;
step4.3: and establishing a layer 2 random forest model according to Step1.1 to Step4.1 to obtain a prediction result of the layer 2 random forest model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111054833.2A CN113762394B (en) | 2021-09-09 | 2021-09-09 | Blasting block prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111054833.2A CN113762394B (en) | 2021-09-09 | 2021-09-09 | Blasting block prediction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113762394A true CN113762394A (en) | 2021-12-07 |
CN113762394B CN113762394B (en) | 2024-04-26 |
Family
ID=78794212
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111054833.2A Active CN113762394B (en) | 2021-09-09 | 2021-09-09 | Blasting block prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113762394B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115659263A (en) * | 2022-10-14 | 2023-01-31 | 长江三峡通航管理局 | Ship control behavior risk assessment system and assessment method based on big data |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105574544A (en) * | 2015-12-16 | 2016-05-11 | 平安科技(深圳)有限公司 | Data processing method and device |
CN109146195A (en) * | 2018-09-06 | 2019-01-04 | 北方爆破科技有限公司 | A kind of blast fragmentation size prediction technique based on cart tree regression algorithm |
US20200004921A1 (en) * | 2019-08-30 | 2020-01-02 | Bikram Baidya | Random forest model for prediction of chip layout attributes |
-
2021
- 2021-09-09 CN CN202111054833.2A patent/CN113762394B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105574544A (en) * | 2015-12-16 | 2016-05-11 | 平安科技(深圳)有限公司 | Data processing method and device |
CN109146195A (en) * | 2018-09-06 | 2019-01-04 | 北方爆破科技有限公司 | A kind of blast fragmentation size prediction technique based on cart tree regression algorithm |
US20200004921A1 (en) * | 2019-08-30 | 2020-01-02 | Bikram Baidya | Random forest model for prediction of chip layout attributes |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115659263A (en) * | 2022-10-14 | 2023-01-31 | 长江三峡通航管理局 | Ship control behavior risk assessment system and assessment method based on big data |
CN115659263B (en) * | 2022-10-14 | 2023-08-08 | 长江三峡通航管理局 | Ship control behavior risk assessment system and method based on big data |
Also Published As
Publication number | Publication date |
---|---|
CN113762394B (en) | 2024-04-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111814871B (en) | Image classification method based on reliable weight optimal transmission | |
CN111047182B (en) | Airspace complexity evaluation method based on deep unsupervised learning | |
CN109635461B (en) | Method and system for automatically identifying surrounding rock grade by using while-drilling parameters | |
CN110782658B (en) | Traffic prediction method based on LightGBM algorithm | |
CN112529341A (en) | Drilling well leakage probability prediction method based on naive Bayesian algorithm | |
CN110225055A (en) | A kind of network flow abnormal detecting method and system based on KNN semi-supervised learning model | |
CN110751101B (en) | Fatigue driving judgment method based on multiple clustering algorithm of unsupervised extreme learning machine | |
CN112749840B (en) | Method for acquiring energy efficiency characteristic index reference value of thermal power generating unit | |
CN110344824B (en) | Acoustic curve generation method based on random forest regression | |
CN106934417A (en) | A kind of data stream self-adapting clustering method towards mixed attributes | |
US20220092359A1 (en) | Image data classification method, device and system | |
CN113705110A (en) | Blasting vibration speed prediction method based on dual random forest regression method | |
CN111523728A (en) | Four-stage hybrid short-term wind direction prediction method | |
CN113762394A (en) | Blasting block size prediction method | |
CN105740521A (en) | Small grid removing method and apparatus in oil reservoir numerical simulation system solution process | |
CN113641733B (en) | Real-time intelligent estimation method for river cross section flow | |
CN105049286A (en) | Cloud platform speed measurement data judging method based on hierarchical clustering | |
Sharma et al. | A semi-supervised generalized vae framework for abnormality detection using one-class classification | |
CN113868943A (en) | Data enhancement-based blasting block degree prediction method for neural network | |
CN105956318A (en) | Improved splitting H-K clustering method-based wind power plant fleet division method | |
CN113379256A (en) | Reservoir evaluation prediction method for Kmeans-random forest | |
CN108376266A (en) | One-class support vector machines Optimization Method of Kernel Parameter based on sample edge point internal point | |
CN108537249A (en) | A kind of industrial process data clustering method of density peaks cluster | |
CN111985782A (en) | Automatic tramcar driving risk assessment method based on environment perception | |
CN113111588B (en) | NO of gas turbine X Emission concentration prediction method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |