CN117951579A - Mineral type recognition model training method, system, equipment and medium - Google Patents
Mineral type recognition model training method, system, equipment and medium Download PDFInfo
- Publication number
- CN117951579A CN117951579A CN202410137560.5A CN202410137560A CN117951579A CN 117951579 A CN117951579 A CN 117951579A CN 202410137560 A CN202410137560 A CN 202410137560A CN 117951579 A CN117951579 A CN 117951579A
- Authority
- CN
- China
- Prior art keywords
- mineral
- data
- training
- training data
- 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.)
- Pending
Links
- 229910052500 inorganic mineral Inorganic materials 0.000 title claims abstract description 240
- 239000011707 mineral Substances 0.000 title claims abstract description 240
- 238000012549 training Methods 0.000 title claims abstract description 183
- 238000000034 method Methods 0.000 title claims abstract description 53
- 230000002159 abnormal effect Effects 0.000 claims description 30
- 238000012360 testing method Methods 0.000 claims description 24
- 238000004590 computer program Methods 0.000 claims description 10
- 238000006467 substitution reaction Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000007637 random forest analysis Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 abstract description 5
- 238000004458 analytical method Methods 0.000 description 26
- 230000015572 biosynthetic process Effects 0.000 description 12
- 238000005755 formation reaction Methods 0.000 description 12
- 230000008569 process Effects 0.000 description 8
- 238000012545 processing Methods 0.000 description 8
- 238000005259 measurement Methods 0.000 description 7
- 239000011435 rock Substances 0.000 description 7
- VYPSYNLAJGMNEJ-UHFFFAOYSA-N silicon dioxide Inorganic materials O=[Si]=O VYPSYNLAJGMNEJ-UHFFFAOYSA-N 0.000 description 7
- 239000004927 clay Substances 0.000 description 6
- 239000010453 quartz Substances 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 229910000514 dolomite Inorganic materials 0.000 description 5
- 239000010459 dolomite Substances 0.000 description 5
- 230000003287 optical effect Effects 0.000 description 5
- 239000000126 substance Substances 0.000 description 5
- 238000011161 development Methods 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000011002 quantification Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 229910052570 clay Inorganic materials 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000005389 magnetism Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000010310 metallurgical process Methods 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 238000001604 Rao's score test Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000013078 crystal Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000011010 flushing procedure Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000001095 inductively coupled plasma mass spectrometry Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 238000004846 x-ray emission Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The application relates to a mineral type recognition model training method, a system, equipment and a medium, belonging to the technical field of mineral type prediction, wherein the method comprises the steps of acquiring mineral training data, wherein the mineral training data comprises mineral types and training data corresponding to the mineral types; classifying mineral training data according to mineral types; judging the data quantity of mineral training data of each class, and determining a minority class; expanding the mineral training data of a few classes to obtain target training data; and carrying out model training on the mineral training data and the target training data of each class to obtain a mineral type identification model. The application has the effect of improving the accuracy of mineral type prediction.
Description
Technical Field
The application relates to the technical field of mineral type prediction, in particular to a mineral type recognition model training method, system, equipment and medium.
Background
Mineral resources are an important material basis for economic and social development, and mineral resource exploration and development matters are national and national security. Analysis of mineral types is critical for geology and mineral resource exploration. Mineral type analysis provides important information about formation properties and composition, helps determine mineral resource potential, optimizes ore processing and metallurgical processes, and evaluates feasibility and risk of geological engineering and architectural design. Current methods of mineral type identification in formations mainly include two approaches.
One conventional analysis method mainly comprises rock slice observation, chemical analysis, mineral quantification instrument analysis, mineral magnetism measurement and the like. In the conventional analysis method, measurement data are obtained by using a measuring instrument, and an expert analyzes the obtained measurement data to obtain a final mineral type result. However, the conventional analysis method relies on expertise and experience of a professional in the relevant research field, and the analysis takes a long time.
Another approach builds models of various mineral types in a predictable formation through deep learning or other machine learning techniques. However, the accuracy of the machine learning or deep learning model depends on the data amount of the real stratum rock mineral type data, but the real stratum rock mineral data is rare at present, so that the prediction error of the model is larger.
The related technical scheme has the following defects: the prediction model has low prediction accuracy for mineral types.
Disclosure of Invention
In order to improve the accuracy of prediction of a mineral type by a prediction model, the application provides a mineral type recognition model training method, a system, equipment and a medium.
In a first aspect of the application, a mineral type recognition model training method is provided. The method comprises the following steps:
acquiring mineral training data, wherein the mineral training data comprises mineral types and training data corresponding to the mineral types;
Classifying mineral training data according to mineral types;
Judging the data quantity of mineral training data of each class, and determining a minority class;
Expanding the mineral training data of a few classes to obtain target training data;
and carrying out model training on the mineral training data and the target training data of each class to obtain a mineral type identification model.
According to the technical scheme, the mineral types in the mineral training data are classified, the data quantity of each type is judged, the minority class is determined, the data in the minority class is expanded, finally, model training is carried out on all the data which are expanded to obtain a mineral type identification model, the balance of the data quantity of each mineral type is realized by expanding the data of the mineral type with smaller data quantity, the problem that the identification accuracy of part of the mineral types is low due to insufficient data quantity is avoided, and the effect of improving the prediction accuracy of the prediction model is achieved.
In one possible implementation, before acquiring the mineral training data, the method further comprises:
acquiring initial historical data, wherein the initial historical data represents actual historical data acquired by technicians, and the initial historical data is ordered according to the acquisition positions of the data;
Identifying outliers in the initial history data;
And calculating an abnormal substitution value according to a plurality of normal values before and/or after the abnormal value, wherein the abnormal substitution value is used for substituting the abnormal value, so as to obtain mineral training data.
According to the technical scheme, the abnormal value is identified and corrected through the initial historical data, so that on one hand, the negative influence of the abnormal value on the training effect of the mineral type identification model can be reduced, and on the other hand, the data size of the mineral training data is guaranteed through the correction of the abnormal value, and the identification accuracy of the mineral type identification model is further improved.
In one possible implementation, determining the minority class by determining the data amount of the mineral training data of each class includes:
Acquiring a data volume ratio of the data volume of the mineral training data of each class to the total data volume of the mineral training data;
When the data volume ratio is smaller than the ratio preset value, the corresponding class of the data volume ratio is a minority class.
In one possible implementation, the expanding the mineral training data of the minority class to obtain the target training data includes:
Carrying out data interpolation on the mineral training data of a few classes, and calculating a data quantity ratio;
and when the data quantity ratio is greater than or equal to a ratio preset value, determining target training data.
In one possible implementation, model training is performed on each type of mineral training data and target training data to obtain a mineral type recognition model, including:
Inputting mineral training data and target training data of each class into a plurality of training models to obtain a plurality of target models;
testing a plurality of target models according to the test data set to obtain a test accuracy;
determining the model weight of each target model according to the test accuracy;
a mineral type recognition model is determined based on the model weights and the plurality of target models.
According to the technical scheme, a plurality of training models are trained to obtain a plurality of target models, the dependence of the prediction accuracy on a single training model is reduced, then the model weight of each target model is determined according to the test accuracy of each target model, and the mineral type recognition model is obtained, so that the prediction accuracy is improved.
In one possible implementation, the training model includes a random forest regression model, an extreme gradient lifting regression model, and a neural network regression model.
In a second aspect of the application, a method of mineral type identification is provided. The method comprises the following steps:
Acquiring data to be predicted;
inputting data to be predicted into the mineral type recognition model obtained according to the first aspect of the application to obtain the mineral type.
In a third aspect of the application, a mineral type recognition model training system is provided. The system comprises:
the data acquisition module is used for acquiring mineral training data, wherein the mineral training data comprises mineral types and training data corresponding to the mineral types;
the data classification module is used for classifying mineral training data according to mineral types;
The data quantity judging module is used for judging the data quantity of the mineral training data of each class and determining a minority class;
the data expansion module is used for expanding the mineral training data of a few classes to obtain target training data;
The model training module is used for carrying out model training on the mineral training data and the target training data of each class to obtain a mineral type identification model.
In a fourth aspect of the application, an electronic device is provided. The electronic device includes: a memory and a processor, the memory having stored thereon a computer program, the processor implementing the method as described above when executing the program.
In a fifth aspect of the application, there is provided a computer readable storage medium having stored thereon a computer program which when executed by a processor implements a method as according to the first and/or second aspects of the application.
In summary, the present application includes at least one of the following beneficial technical effects:
Classifying the mineral types in the mineral training data, judging the data quantity of each type, determining a minority class, expanding the data in the minority class, and finally performing model training on all the data which are expanded to obtain a mineral type identification model.
Drawings
Fig. 1 is a schematic flow chart of a mineral type recognition model training method according to an embodiment of the present application.
Fig. 2 is a schematic representation of mineral training data provided in an embodiment of the present application.
Fig. 3 is a schematic structural diagram of a mineral type recognition model training system according to an embodiment of the present application.
Fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
In the figure, 201, a data acquisition module; 202. a data classification module; 203. a data quantity judging module; 204. a data expansion module; 205. a model training module; 301. a CPU; 302. a ROM; 303. a RAM; 304. an I/O interface; 305. an input section; 306. an output section; 307. a storage section; 308. a communication section; 309. a driver; 310. removable media.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more apparent, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
In addition, the term "and/or" herein is merely an association relationship describing an association object, and means that three relationships may exist, for example, a and/or B may mean: a exists alone, A and B exist together, and B exists alone. In this context, unless otherwise specified, the term "/" generally indicates that the associated object is an "or" relationship.
Mineral resources are an important material basis for economic and social development, and mineral resource exploration and development matters are national and national security. Analysis of mineral type and mineral content is critical for geology and mineral resource exploration. Mineral types can provide important information about formation properties and composition, help determine mineral resource potential, optimize ore processing and metallurgical processes, and can also evaluate feasibility and risk of geological engineering and architectural design.
The existing mineral identification method mainly comprises two modes of traditional analysis and computer analysis. The traditional analysis mode refers to the mode of rock slice observation, chemical analysis, mineral quantification instrument analysis, mineral magnetic measurement and the like, so as to obtain measurement data, and a person skilled in the relevant field or a technical expert analyzes the obtained measurement data.
Rock flake observation is a common method for preliminary determination of mineral type by preparing flakes of formation samples and observing them using an optical microscope to identify the morphology, color and crystal structure of different minerals. Chemical analysis is the determination of chemical composition of different minerals in a formation sample by sampling and chemical laboratory analysis, such as X-ray fluorescence spectroscopy, inductively coupled plasma mass spectrometry. The analysis of the mineral quantification apparatus requires image analysis and energy spectrum analysis of the formation sample by using related apparatuses such as an electron microscope, a scanning electron microscope, an energy spectrum analyzer, etc., to quantitatively measure the distribution of different minerals. The mineral magnetism measurement is to quantitatively measure minerals in stratum samples by utilizing the magnetic difference of minerals in rock and through a magnetic measuring instrument, a magnetic hysteresis instrument and other devices. These traditional analytical methods find wide application in geology and mineral resource exploration. However, the traditional analysis method is seriously dependent on professionals in the related research field, has high requirements on professional knowledge of professionals, and takes a long time in the whole analysis process. Meanwhile, the traditional analysis method cannot fully mine various geological data information when the stratum mineral type is identified, and has limitation on the analysis of the mineral type.
For the computer analysis mode, models of various mineral types in predictable formations can be built by deep learning and other machine learning techniques from large-scale geological data. The method can improve the accuracy and efficiency of mineral identification and provide powerful support for geological research and resource exploration. On the one hand, the current computer analysis mode depends on a single learning prediction model, and the prediction accuracy of the learning prediction model depends on a large amount of geological data, but the data of the stratum rock minerals are scarce at present, and in this case, the accuracy of the learning prediction model cannot be ensured. On the other hand, the learning prediction model is relatively sensitive to abnormal values and noise data in the data set, and the abnormal values and noise data in the real data can also have negative effects on the prediction accuracy of the learning prediction model, so that the prediction accuracy of the learning prediction model is low.
The application provides a mineral type recognition model training method, which improves the data quantity of a mineral type with less data by expanding the data of a certain or a plurality of mineral types so as to achieve the aim of improving the prediction accuracy of a learning prediction model. Meanwhile, the problems that the current model is excessively dependent on a single prediction learner and is influenced by abnormal values and noise are solved by adopting modes of multi-model combination, abnormal data processing and the like.
Embodiments of the application are described in further detail below with reference to the drawings.
The embodiment of the application provides a mineral type recognition model training method, and the main flow of the method is described as follows.
As shown in fig. 1:
step S101: mineral training data is obtained.
Specifically, the mineral training data includes a mineral type and training data corresponding to the mineral type, for example, referring to fig. 2, fig. 2 shows a part of the training data, in which DEPTH represents a logging DEPTH, AC represents a sonic time difference, CAL represents a borehole diameter, i.e., a diameter of a borehole, CNL represents a compensation neutron, DEN represents a compensation density, GR represents a natural gamma value, RT represents a true formation resistivity, RXO represents a formation resistivity of a flushing zone, and Label represents a mineral type. In other embodiments, the training data may also include other data for determining mineral type, not limited herein.
Before the mineral training data is acquired, the mineral type recognition model training method further comprises the following steps:
acquiring initial historical data, wherein the initial historical data represents actual historical data acquired by technicians, and the initial historical data is ordered according to the acquisition positions of the data; identifying an outlier in the initial history data; and calculating an abnormal substitution value according to a plurality of normal values before and/or after the abnormal value, wherein the abnormal substitution value is used for substituting the abnormal value, so as to obtain the mineral training data.
In one specific example, data preparation is first performed, resulting in initial historical data. And then checking the initial historical data by adopting an abnormal value checking method to determine the abnormal value in the initial historical data. In the embodiment provided by the application, the outlier checking method is an isolated forest algorithm, key parameters in the isolated forest, such as the number n_ estimators of isolated trees, the number of samples max_samples sampled in each isolated tree, the expected outlier proportion is continuity, and the maximum feature number max_features considered during splitting of each isolated tree is set by setting the key parameters of the isolated forest, so that the search range is set to optimize the model performance of the isolated forest.
In another embodiment, the abnormal value detection method may be a method capable of detecting an abnormal value, such as a principal component analysis (PRINCIPAL COMPONENTS ANALYSIS, PCA), a local outlier factor (local outlier factor, LOF), or a Z-Score test, and is not limited thereto.
After the abnormal value is detected, if the abnormal value is directly deleted, the data sample size of model training is reduced, so that the detected abnormal value needs to be corrected on the premise of ensuring that the data sample size is not reduced, and the negative influence of the abnormal value on the model training is reduced. Before correcting the outlier, it is necessary to understand the ordering rule of the training data, in fig. 2, each row of data represents a set of data, each set of data is measured at a corresponding logging depth, and the logging depths are measured at intervals of 0.125m, so the ordering among the sets of data is performed according to the logging depths, and may be ascending ordering according to the logging depths, or descending ordering according to the logging depths, which is not limited herein.
For example, for a detected outlier, a selection is made to replace it with a mean of a plurality of data points before and after the outlier, where none of the plurality of data points involved in the mean calculation is an outlier. For example, the first eight data points of the outlier may be averaged to obtain the outlier, and for example, the first four data points and the last four data points of the outlier may be averaged to obtain the outlier, and in other embodiments, the outlier may be calculated by taking ten data points, six data points, and so on. The front and the back of the abnormal value are determined according to the sorting result, the sorting result is different, and the data before and after the abnormal value are different. On the premise of reducing the interference of abnormal values on the model, the correction method reserves the overall trend and the data sample size of the data, ensures the reliability of the prediction result to a certain extent, and improves the prediction accuracy of the learning prediction model.
Step S102: mineral training data is classified according to mineral type.
Specifically, the mineral types comprise a plurality of types such as shale, dolomite, quartz, clay and the like, training data corresponding to different mineral types are different, data sample sizes are also different, and analysis processing is needed to be carried out on the different mineral types respectively. For example, data of Label as gray shale is used as a class, data of Label as dolomite is used as a class, data of Label as quartz is used as a class, and data of Label as clay is used as a class.
Step S103: and judging the data quantity of the mineral training data of each class, and determining a few classes.
Specifically, a data amount ratio of the data amount of the mineral training data of each class to the total data amount of the mineral training data is obtained; when the data volume ratio is smaller than the ratio preset value, the corresponding class of the data volume ratio is a minority class.
In a specific example, the data amount refers to how many sets of data, for example, four different mineral types of data are included in the mineral training data, so there are four types of data in total, and the data amount corresponding to the shale is 100, that is, there are one hundred sets of data. Dolomite corresponds to a data volume of 50, i.e. fifty sets of data exist. The amount of data corresponding to quartz was 30, i.e., there were thirty sets of data. The data amount corresponding to clay is 20, namely twenty groups of data exist. The ratio of the corresponding data amount of the gray shale is 0.5, the ratio of the corresponding data amount of the dolomite is 0.25, the ratio of the corresponding data amount of the quartz is 0.15, and the ratio of the corresponding data amount of the clay is 0.1. If the ratio preset value is set to 0.25, the classes corresponding to quartz and clay are minority classes.
In other embodiments, the judgment of the minority class may be directly performed according to the data amount of each class, for example, the maximum data amount in each class is a preset value, that is, the preset value is 100, the classes corresponding to dolomite, quartz and clay are all minority classes, and when the data amount in the minority classes reaches 100, the expansion of the minority class data is ended.
Step S104: and expanding the mineral training data of a few classes to obtain target training data.
Specifically, carrying out data interpolation on the mineral training data of the minority class, and calculating the data quantity ratio; and when the data quantity ratio is larger than or equal to the ratio preset value, determining the target training data.
In order to reduce the influence of insufficient sample number on the accuracy of mineral type identification, the application realizes the expansion of data in a minority class by carrying out data interpolation on the data in the minority class, thereby realizing the improvement of the accuracy of the prediction of the mineral type identification model. In a specific example, the application chooses to use a synthetic minority class oversampling technique (SYNTHETIC MINORITY OVERSAMPLING TECHNIQUE, SMOTE) for data interpolation, and generates new mineral training data by using a K-nearest neighbor algorithm by considering the distribution of the mineral training data in minority classes in the feature space. The sample number of the minority class is effectively increased by using the SMOTE method, so that the mineral training data is more balanced. The new mineral training data generated by interpolation provides rich training data for the model, so that the model learns the characteristics of few classes more comprehensively, and the overall performance of the model is improved.
Further, for each determined minority class of mineral training data, K nearest neighbors of a certain mineral training data in the minority class in a feature space are selected by using a K-nearest neighbor algorithm, and in the embodiment provided by the application, euclidean distances are adopted to quantify the distance between the mineral training data. A neighbor is randomly selected from the K nearest neighbors and a random proportion is selected on a line segment between the mineral training data and the neighbor. The ratio determines the position of the new mineral training data in the feature space. Through the above process, a new mineral training data is generated. The above process is repeated until the sample expansion requirement for the minority class is satisfied. The random proportion may be generated by a random function or may be specified by a skilled person, and is not limited thereto.
Step S105: and carrying out model training on the mineral training data and the target training data of each class to obtain a mineral type identification model.
Specifically, inputting the mineral training data and the target training data of each class into a plurality of training models to obtain a plurality of target models; testing the target models according to the test data set to obtain a test accuracy; determining the model weight of each target model according to the test accuracy; determining the mineral type recognition model based on the model weights and the plurality of target models.
In one specific example, the training models include a random forest regression model, an extreme gradient lifting (extreme gradient boosting, XGBoost) regression model, and a neural network regression model. And respectively inputting the mineral training data and the target training data of each class into a random forest regression model, a XGBoost regression model and a neural network regression model for training, and obtaining three target models after training. And then testing the three target models by using a test data set, wherein the data type in the test data set is the same as the data type of the mineral training data, namely the mineral training data is the training data set. After the test is completed, the test accuracy corresponding to each target model can be obtained. For example, for a certain target model, there are 100 pieces of test data in total, and the test result and the actual result of 80 pieces of test data are the same, so that the test accuracy of the target model is 0.8. If the test accuracy of the three target models is 0.95, 0.8 and 0.6, the model weights corresponding to the three target models are 0.95/(0.95+0.8+0.6) ≡0.4, 0.8/(0.95+0.8+0.6) ≡ 0.35,0.6/(0.95+0.8+0.6) ≡0.25). The three target models and their corresponding model weights constitute a mineral type recognition model.
The acquisition process of the target training data can know that the target training data refers to data after the data expansion of a minority of classes is completed, and the mineral training data refers to original data which is not expanded, so that the mineral training data and the target training data of each class refer to all data after the data expansion is completed.
In other embodiments, the training model may be a model that can implement mineral type recognition, such as other machine learning models, deep learning models, and the like, and is not limited herein.
In the embodiment provided by the application, the accuracy and the robustness of the identification of the stratum mineral type are improved by fusing the random forest regression model, the XGBoost regression model and the neural network regression model. In the mineral type recognition model, an adaptive weight distribution mechanism is introduced for each training model. By monitoring the performance of each training on the mineral training data, the performance refers to the accuracy of each training model to the test data set, and the weight of each training model is adjusted so that the training model with the best performance can influence the final prediction result to a greater extent.
The embodiment of the application provides a mineral type identification method, which comprises the steps of obtaining data to be predicted; inputting the data to be predicted into a mineral type recognition model obtained by the mineral type recognition model training method to obtain the mineral type.
In a specific example, the data to be predicted includes logging depth, acoustic time difference, diameter of the borehole, compensated neutrons, compensated density, natural gamma value, true formation resistivity, and formation resistivity of the washout zone, and the data to be predicted is input into the mineral type recognition model, where three target models in the mineral type recognition model may respectively obtain the predicted probabilities of the respective mineral types. And carrying out weighted calculation on the prediction probabilities of the same type of all the target models to obtain probability results of all the models on different types. The mineral type with the highest weighted probability is selected as the final prediction result of the data to be predicted, i.e. the mineral type. For example, the first object model predicts a probability of a1 for mineral type A, B1 for mineral type B, and C1 for mineral type C. The prediction result of the second object model is that the probability of the mineral type A is a2, the probability of the mineral type B is B2, and the probability of the mineral type C is C2. The predicted result of the third object model is that the probability of the mineral type A is a3, the probability of the mineral type B is B3, and the probability of the mineral type C is C3. The model weight corresponding to the first target model is 0.4, the model weight corresponding to the second target model is 0.35, and the model weight corresponding to the third target model is 0.25. The probability of a mineral type a is 0.4×a1+0.35×a2+0.25×a3, the probability of B mineral type B is 0.4×b1+0.35×b2+0.25×b3, and the probability of C mineral type C is 0.4×c1+0.35×c2+0.25×c3. Comparing the probabilities corresponding to the three mineral types, if the probability of the mineral type A is the largest, the final mineral type is A, and similarly, if the probability of the mineral type B is the largest, the final mineral type is B, and if the probability of the mineral type C is the largest, the final mineral type is C.
An embodiment of the present application provides a mineral type recognition model training system, referring to fig. 3, the mineral type recognition model training system includes:
The data acquisition module 201 is configured to acquire mineral training data, where the mineral training data includes a mineral type and training data corresponding to the mineral type;
a data classification module 202 for classifying mineral training data according to mineral type;
The data amount judging module 203 is configured to judge the data amount of the mineral training data of each class, and determine a minority class;
The data expansion module 204 is configured to expand the minority class of mineral training data to obtain target training data;
The model training module 205 is configured to perform model training on the mineral training data and the target training data of each class, so as to obtain a mineral type recognition model.
It will be clear to those skilled in the art that, for convenience and brevity of description, reference may be made to the corresponding process in the foregoing method embodiment for the specific working process of the described module, which is not described herein again.
The embodiment of the application discloses electronic equipment. Referring to fig. 4, the electronic apparatus includes a central processing unit (central processing unit, CPU) 301 that can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage portion 307 into a random access memory (random access memory, RAM) 303. In the RAM 303, various programs and data required for the system operation are also stored. The CPU 301, ROM 302, and RAM 303 are connected to each other by a bus. An input/output (I/O) interface 304 is also connected to the bus.
The following components are connected to the I/O interface 304: an input section 305 including a keyboard, a mouse, and the like; an output section 306 including a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker, and the like; a storage portion 307 including a hard disk and the like; and a communication section 308 including a network interface card such as a local area network (local area network, LAN) card, modem, or the like. The communication section 308 performs communication processing via a network such as the internet. A driver 309 is also connected to the I/O interface 304 as needed. A removable medium 310 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is installed on the drive 309 as needed, so that a computer program read out therefrom is installed into the storage section 307 as needed.
In particular, the process described above with reference to flowchart fig. 1 may be implemented as a computer software program according to an embodiment of the application. For example, embodiments of the application include a computer program product comprising a computer program embodied on a machine-readable medium, the computer program comprising program code for performing the method shown in the flowcharts. In such embodiments, the computer program may be downloaded and installed from a network via the communication portion 308, and/or installed from the removable media 310. The above-described functions defined in the apparatus of the present application are performed when the computer program is executed by a Central Processing Unit (CPU) 301.
The computer readable medium shown in the present application may be a computer readable signal medium or a computer readable storage medium, or any combination of the two. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (erasable programmable read only memory, EPROM), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present application, however, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, radio Frequency (RF), and the like, or any suitable combination of the foregoing.
The above description is only illustrative of the preferred embodiments of the present application and of the principles of the technology employed. It will be appreciated by persons skilled in the art that the scope of the application is not limited to the specific combinations of the features described above, but also covers other embodiments which may be formed by any combination of the features described above or their equivalents without departing from the spirit of the application. Such as the above-mentioned features and the technical features having similar functions (but not limited to) applied for in the present application are replaced with each other.
Claims (10)
1. A mineral type recognition model training method, comprising:
Acquiring mineral training data, wherein the mineral training data comprises a mineral type and training data corresponding to the mineral type;
Classifying the mineral training data according to the mineral type;
Judging the data quantity of the mineral training data of each class, and determining a minority class;
expanding the mineral training data of the minority class to obtain target training data;
and carrying out model training on the mineral training data and the target training data of each class to obtain a mineral type identification model.
2. The mineral type recognition model training method of claim 1, wherein prior to the acquiring mineral training data, the method further comprises:
Acquiring initial historical data, wherein the initial historical data represents actual historical data acquired by technicians, and the initial historical data is ordered according to the acquisition positions of the data;
identifying outliers in the initial history data;
and calculating an abnormal substitution value according to a plurality of normal values before and/or after the abnormal value, wherein the abnormal substitution value is used for substituting the abnormal value, so as to obtain the mineral training data.
3. The method of claim 1, wherein determining the minority class by determining the data amount of the mineral training data for each class comprises:
Acquiring a data volume ratio of the data volume of the mineral training data of each class to the total data volume of the mineral training data;
And when the data volume ratio is smaller than the ratio preset value, the corresponding class of the data volume ratio is a minority class.
4. A mineral type recognition model training method in accordance with claim 3, wherein said expanding the minority class of mineral training data to obtain target training data comprises:
performing data interpolation on the mineral training data of the minority class, and calculating the data quantity ratio;
And when the data quantity ratio is greater than or equal to the ratio preset value, determining the target training data.
5. The method for training a mineral type recognition model according to claim 1, wherein the training of the mineral training data and the target training data for each class to obtain a mineral type recognition model comprises:
inputting the mineral training data and the target training data of each class into a plurality of training models to obtain a plurality of target models;
Testing the target models according to the test data set to obtain a test accuracy;
determining the model weight of each target model according to the test accuracy;
The mineral type recognition model is determined from the model weights and the plurality of target models.
6. The mineral type recognition model training method of claim 5, wherein the training model comprises a random forest regression model, an extreme gradient lifting regression model, and a neural network regression model.
7. A method of mineral type identification comprising:
Acquiring data to be predicted;
Inputting the data to be predicted into the mineral type identification model according to any one of claims 1-6 to obtain a mineral type.
8. A mineral type recognition model training system, comprising:
the data acquisition module is used for acquiring mineral training data, wherein the mineral training data comprises mineral types and training data corresponding to the mineral types;
the data classification module is used for classifying the mineral training data according to the mineral types;
The data quantity judging module is used for judging the data quantity of the mineral training data of each class and determining a minority class;
the data expansion module is used for expanding the minority class mineral training data to obtain target training data;
And the model training module is used for carrying out model training on the mineral training data and the target training data of each class to obtain a mineral type identification model.
9. An electronic device comprising a memory and a processor, the memory having stored thereon a computer program capable of being loaded by the processor and performing the method according to any of claims 1 to 6 or 7.
10. A computer readable storage medium, characterized in that a computer program is stored which can be loaded by a processor and which performs the method according to any of claims 1 to 6 or 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410137560.5A CN117951579A (en) | 2024-01-31 | 2024-01-31 | Mineral type recognition model training method, system, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410137560.5A CN117951579A (en) | 2024-01-31 | 2024-01-31 | Mineral type recognition model training method, system, equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117951579A true CN117951579A (en) | 2024-04-30 |
Family
ID=90801234
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410137560.5A Pending CN117951579A (en) | 2024-01-31 | 2024-01-31 | Mineral type recognition model training method, system, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117951579A (en) |
-
2024
- 2024-01-31 CN CN202410137560.5A patent/CN117951579A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210026031A1 (en) | Dolomite reservoir prediction method and system based on well and seismic combination, and storage medium | |
| CN111767674B (en) | Logging lithology recognition method based on active domain adaptation | |
| US11893495B2 (en) | Dual neural network architecture for determining epistemic and aleatoric uncertainties | |
| CN117272181B (en) | Method for discriminating zircon host rock and ore forming environment by machine learning modeling | |
| CN117576335B (en) | Three-dimensional space model data processing method and system for mineral area investigation | |
| CN115049627B (en) | Steel surface defect detection method and system based on domain self-adaptive depth migration network | |
| CN116756679A (en) | Multi-source information fusion-based method for judging geological mode of down-the-hole mountain | |
| CN117056834A (en) | Big data analysis method based on decision tree | |
| CN113065279A (en) | Method, device, equipment and storage medium for predicting total organic carbon content | |
| CN111626377A (en) | Lithofacies identification method, device, equipment and storage medium | |
| CN116012275A (en) | Downhole real-time processing device and method for while-drilling electric imaging data | |
| CN112347776B (en) | Medical data processing method and device, storage medium and electronic equipment | |
| CN115081485B (en) | AI-based magnetic flux leakage internal detection data automatic analysis method | |
| CN117951579A (en) | Mineral type recognition model training method, system, equipment and medium | |
| CN116188266A (en) | Spatial interpolation method for geological investigation based on fuzzy theory and ensemble learning | |
| CN113327070B (en) | Method and device for intelligently surveying coal-based gas and electronic equipment | |
| CN116011234A (en) | Pressure front sleeve change risk level judgment method integrating geomechanics and Bayes | |
| CN115831219A (en) | Quality prediction method, device, equipment and storage medium | |
| CN114818493A (en) | Method for quantitatively evaluating integrity degree of tunnel rock mass | |
| CN117951594A (en) | Reservoir classification model training method, classification method, system, equipment and medium | |
| CN117766068B (en) | Machine learning method for predicting porphyry deposit type and reserve based on zircon component | |
| CN116780524B (en) | Industrial enterprise short-term load prediction method based on LSTM deep learning | |
| CN117934019B (en) | Copper concentrate sample tracing method and system based on deep learning | |
| CN116451812B (en) | Wettability prediction method and device based on multi-granularity cascade forest and super-parameter optimization | |
| CN117172081A (en) | Regional evaluation method for geological engineering economic multi-factor fusion |
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 |