Multi-dimensional data preferential analysis method and device based on gradient boosting decision tree model
By using a gradient boosting decision tree model for multidimensional data optimization, this method solves the problems of accurately identifying key features and poor dynamic adaptability in multidimensional data processing, achieving efficient and accurate strategy optimization and real-time performance improvement.
Patent Information
- Application Number
- CN202510807325.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-17
- Publication Date
- 2025-11-14
AI Technical Summary
Existing technologies struggle to accurately identify key features when processing multidimensional data, resulting in low processing efficiency, poor accuracy, and an inability to guarantee the accuracy and real-time nature of optimal decision-making. Furthermore, they are poorly adaptable to dynamic scenarios.
A multidimensional data optimization analysis method based on gradient boosting decision tree model is adopted. By acquiring multidimensional data of the current time period, the initial features are extracted using the target gradient boosting decision tree model, the importance is ranked based on the target weight, the optimal strategy combination is output, and the model parameters are optimized by gradient descent method to adapt to dynamic scene changes.
It improves the accuracy and real-time performance of strategy optimization in multidimensional data, has high processing efficiency, interpretability and robustness, and can adaptively adjust optimization strategies in changing environments.
Smart Images

Figure CN120950924A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data mining technology, and in particular to a method and apparatus for multidimensional data optimization analysis based on a gradient boosting decision tree model. Background Technology
[0002] In existing technologies, when processing multidimensional data of a target device, the relationships between the multidimensional data are often highly complex and nonlinear. Therefore, random forest algorithms are typically used to mine key features in the multidimensional data and use these key features to provide the optimal decision for the target device.
[0003] However, while the Random Forest algorithm has strong processing capabilities for multidimensional data, it struggles to accurately identify key features, resulting in low processing efficiency and poor accuracy, failing to guarantee the accuracy and real-time performance of optimal decisions. Furthermore, the Random Forest algorithm is poorly adaptable to dynamic scenarios, unable to adjust in real time, and exhibits poor generalization ability. Summary of the Invention
[0004] This invention provides a multidimensional data optimization analysis method and apparatus based on a gradient boosting decision tree model, in order to address the deficiencies in related technologies.
[0005] This invention provides a multidimensional data optimization analysis method based on a gradient boosting decision tree model, comprising: Obtain the first multidimensional data for the target device in the current time period; The first multidimensional data is input into the target gradient boosting decision tree model, which extracts multiple initial features from the first multidimensional data. Based on the target weights corresponding to the multiple initial features, the multiple initial features are ranked by importance to obtain multiple key features. Based on the multiple key features, the optimal strategy combination for the target device is output. The target gradient boosting decision tree model is obtained by training the initial gradient boosting decision tree model of the previous time period based on the second multidimensional data of the target device in the previous time period and the optimal strategy combination label of the target device in the current time period.
[0006] According to the present invention, a multidimensional data optimization analysis method based on a gradient boosting decision tree model is provided. The first multidimensional data and the second multidimensional data both include operation data, equipment status data and environmental data of multiple actions of multiple operators on the target equipment. The optimal strategy combination and the optimal strategy combination label both include the optimal operation data, optimal equipment status data and optimal environmental data for the key actions among the multiple actions.
[0007] According to the present invention, a multidimensional data optimization analysis method based on a gradient boosting decision tree model is provided, wherein the first multidimensional data and the second multidimensional data both include different combinations of multiple parameters of the target device, and the optimal strategy combination and the optimal strategy combination label both include the optimal combination of key parameters among the multiple parameters.
[0008] According to the present invention, a multidimensional data optimization analysis method based on a gradient boosting decision tree model is provided, wherein the training steps of the target gradient boosting decision tree model include: The second multidimensional data is input into the initial gradient boosting decision tree model, which extracts multiple sample features from the second multidimensional data. Based on the initial weights corresponding to the multiple sample key features, the multiple sample features are ranked by importance to obtain multiple sample key features. Based on the multiple sample key features, the output result is obtained. Based on the optimal strategy combination label and the output result, the loss function is calculated, and the reduction of the loss function caused by the feature split of each decision tree in the initial gradient boosting decision tree model is statistically analyzed. Based on the reduction, the gradient descent method is applied to iteratively optimize the model parameters of the initial gradient boosting decision tree model until the loss function converges, thereby obtaining the target weight and the target gradient boosting decision tree model. The model parameters include the initial weights, and the gradient descent method uses a dynamic learning rate for gradient descent.
[0009] According to the present invention, a multidimensional data optimization analysis method based on a gradient boosting decision tree model is provided, wherein each decision tree in the initial gradient boosting decision tree model is generated by a classification regression tree algorithm; the splitting metric for feature splitting includes mean squared error or Gini coefficient.
[0010] According to the present invention, a multidimensional data optimization analysis method based on a gradient boosting decision tree model is provided, wherein the initial gradient boosting decision tree model is constructed based on a gradient boosting decision tree or based on an extreme gradient boosting decision tree.
[0011] According to the present invention, a multidimensional data optimization analysis method based on a gradient boosting decision tree model is provided, wherein the first multidimensional data for the target device in the current time period is obtained, followed by: Principal component analysis was performed on the first multidimensional data to obtain the analysis results; Accordingly, inputting the first multidimensional data into the target gradient boosting decision tree model includes: The analysis results are then input into the target gradient boosting decision tree model.
[0012] According to the present invention, a multidimensional data optimization analysis method based on a gradient boosting decision tree model is provided, wherein the method further includes obtaining the first multidimensional data for the target device in the current time period, and then further comprising: The first multidimensional data is processed for missing values, normalization, and noise reduction to obtain the processing result; Accordingly, inputting the first multidimensional data into the target gradient boosting decision tree model includes: The processing results are then input into the target gradient boosting decision tree model.
[0013] The present invention also provides a multidimensional data optimization analysis device based on a gradient boosting decision tree model, comprising: The data acquisition module is used to acquire the first multidimensional data for the target device in the current time period; The data analysis module is used to input the first multidimensional data into the target gradient boosting decision tree model, extract multiple initial features from the first multidimensional data by the target gradient boosting decision tree model, sort the multiple initial features by importance based on the target weights corresponding to the multiple initial features to obtain multiple key features, and output the optimal strategy combination for the target device based on the multiple key features. The target gradient boosting decision tree model is obtained by training the initial gradient boosting decision tree model based on the second multidimensional data of the target device in the previous time period of the current time period and the optimal strategy combination label of the target device.
[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multidimensional data optimization analysis method based on the gradient boosting decision tree model as described above.
[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multidimensional data optimization analysis method based on the gradient boosting decision tree model as described above.
[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the multidimensional data optimization analysis method based on the gradient boosting decision tree model as described above.
[0017] This invention provides a multidimensional data optimization analysis method and apparatus based on a gradient boosting decision tree model. First, it acquires first multidimensional data for the target device in the current time period. Then, it inputs this first multidimensional data into a target gradient boosting decision tree model, which extracts multiple initial features from the first multidimensional data. Based on the target weights corresponding to these initial features, the initial features are ranked by importance to obtain multiple key features. Finally, based on these key features, the optimal strategy combination for the target device is output. This method uses a target gradient boosting decision tree model to solve the nonlinear modeling problem of strategy optimization in multidimensional data. By using target weights determined during training, key features can be accurately identified, resulting in high processing efficiency and accuracy. Compared with traditional linear regression or single decision tree methods, it not only improves the accuracy and real-time performance of strategy optimization but also has strong interpretability and robustness. It can adaptively adjust the optimization strategy in a changing environment and is widely used in industrial optimization, intelligent decision-making, and other fields. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in this invention or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is one of the flowcharts of the multidimensional data optimization analysis method based on gradient boosting decision tree model provided by the present invention.
[0020] Figure 2 This is a schematic diagram showing the importance ranking of key features in the multidimensional data optimization analysis method based on gradient boosting decision tree model provided by the present invention.
[0021] Figure 3 This is the second flowchart of the multidimensional data optimization analysis method based on gradient boosting decision tree model provided by the present invention.
[0022] Figure 4 This is a schematic diagram of the structure of the multidimensional data optimization analysis device based on the gradient boosting decision tree model provided by the present invention.
[0023] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0025] Figure 1 This is a flowchart illustrating a multidimensional data optimization analysis method based on a gradient boosting decision tree model provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes: S1, Obtain the first multi-dimensional data for the target device in the current time period; S2, the first multidimensional data is input into the target gradient boosting decision tree model, the target gradient boosting decision tree model extracts multiple initial features from the first multidimensional data, the multiple initial features are ranked by importance based on the target weights corresponding to the multiple initial features to obtain multiple key features, and the optimal strategy combination for the target device is output based on the multiple key features. The target gradient boosting decision tree model is obtained by training the initial gradient boosting decision tree model of the previous time period based on the second multidimensional data of the target device in the previous time period and the optimal strategy combination label of the target device in the current time period.
[0026] Specifically, the multidimensional data optimization analysis method based on gradient boosting decision tree model provided in this embodiment of the invention is executed by a multidimensional data optimization analysis device based on gradient boosting decision tree model. This device can be configured in a computer, which can be a local computer or a cloud computer. The local computer can be a computer, tablet, etc., and no specific limitation is made here.
[0027] First, execute step S1 to obtain the first multi-dimensional data for the target device in the current time period. The current time period refers to the update cycle of the target gradient boosting decision tree model, and its length can be set as needed; no specific limitation is made here.
[0028] The target equipment can be a vehicle, an aircraft, or other type of equipment that requires operators to perform target tasks. In this case, the first multidimensional data can be the operation data, equipment status data, and environmental data of each operator's multiple actions on the target equipment. For example, the operation data, equipment status data, and environmental data of each operator's multiple actions when performing target tasks using the target equipment.
[0029] The target mission can be a vehicle tracking mission, an aircraft landing protection mission, an emergency landing mission, etc. Multiple actions can include translation and rotation. Translation can include forward and backward movements, while rotation can include left turn, right turn, flying to the upper right, flying to the lower left, flying to the upper left, flying to the lower right, etc.
[0030] Operational data may include translational distance, rotation angle, etc. Equipment status data may include system fault information, equipment load, attitude angle, speed, acceleration, etc. Environmental data may include air pressure, terrain slope, light intensity, temperature, wind speed, pressure, humidity, etc.
[0031] The target device can also be a second type of device that is controlled by a combination of multiple parameters, such as a large sensor system, where the values of each parameter can be acquired by sensors of the corresponding type. In this case, the first multidimensional data can be the values of multiple parameters acquired by each group of sensors.
[0032] Then, step S2 is executed, where the first multidimensional data is input into the Gradient Boosting Decision Tree (GBDT) model. The GBDT model models the first multidimensional data, capturing the non-linear relationships between features, determining multiple initial features in the first multidimensional data, and determining the target weight corresponding to each initial feature through hierarchical learning. Using the target weights corresponding to each initial feature, the initial features are ranked by importance, resulting in several key features with high importance.
[0033] Here, compared with the AdaBoost model, the GBDT model can directly optimize the gradient direction, making it more suitable for continuous value prediction.
[0034] The initial features refer to the data classification in the first multidimensional data. Each initial feature corresponds to a target weight, which is determined when the target gradient boosting decision tree model is determined, that is, it is obtained by training the initial gradient boosting decision tree model.
[0035] The greater the target weight of each initial feature, the more important that initial feature is.
[0036] The initial features are sorted by importance and visualized, such as... Figure 2 As shown, this allows users to more intuitively and clearly identify the key features. Figure 2 The horizontal axis represents the proportion of the target weight corresponding to each initial feature, and the vertical axis represents each initial feature.
[0037] From the initial features, a preset number of initial features with large target weights can be selected as key features, or initial features with target weights greater than or equal to preset weights can be selected as key features; no specific limitation is made here. Both the preset number and preset weights can be set as needed.
[0038] If the target device is a Class I device, the key feature can be at least one of the following: key action, key device status, and key environmental data.
[0039] If the target device is a type II device, then the key feature can be the key parameter among the parameters.
[0040] After identifying the key features, the optimal strategy combination for the target device can be obtained and output using these features. If the target device is a type 1 device, the optimal strategy combination can be the optimal operational data, optimal device state data, and optimal environmental data for the key actions of the target device. If the target device is a type 2 device, the optimal strategy combination can be the optimal combination of values for the key parameters of the target device.
[0041] It is understandable that the target gradient boosting decision tree model can be obtained by using the second multidimensional data of the target device from the previous time period as training samples and applying the optimal policy combination label of the target device to train the initial gradient boosting decision tree model.
[0042] Here, the second multidimensional data is of the same type as the first multidimensional data, and the optimal strategy combination label is of the same type as the optimal strategy combination. That is, if the target device is a first-type device, both the optimal strategy combination and the optimal strategy combination label include the optimal operation data, optimal device state data, and optimal environment data for the key actions among multiple actions; if the target device is a second-type device, both the optimal strategy combination and the optimal strategy combination label include the optimal combination of values for the key parameters among multiple parameters.
[0043] The training sample can be understood as the multidimensional data increment generated between adjacent time periods. This allows for incremental training of the gradient boosting decision tree model for each time period to adapt to dynamic scene changes (such as sudden changes in environmental parameters) and avoid the problem of low training efficiency caused by full training of the model.
[0044] The multidimensional data optimization analysis method based on a gradient boosting decision tree model provided in this invention first obtains the first multidimensional data for the target device in the current time period. Then, the first multidimensional data is input into the target gradient boosting decision tree model, which extracts multiple initial features from the first multidimensional data. Based on the target weights corresponding to these initial features, the initial features are ranked by importance to obtain multiple key features. Based on these key features, the optimal strategy combination for the target device is output. This method uses a target gradient boosting decision tree model to solve the nonlinear modeling problem of strategy optimization in multidimensional data. By using the target weights determined during training, key features can be accurately identified, resulting in high processing efficiency and accuracy. Compared with traditional linear regression or single decision tree methods, this method not only improves the accuracy and real-time performance of strategy optimization but also has strong interpretability and robustness. It can adaptively adjust the optimization strategy in a changing environment and is widely used in industrial optimization, intelligent decision-making, and other fields.
[0045] Based on the above embodiments, the training steps of the target gradient boosting decision tree model include: The second multidimensional data is input into the initial gradient boosting decision tree model, which extracts multiple sample features from the second multidimensional data. Based on the initial weights corresponding to the multiple sample key features, the multiple sample features are ranked by importance to obtain multiple sample key features. Based on the multiple sample key features, the output result is obtained. Based on the optimal strategy combination label and the output result, the loss function is calculated, and the reduction of the loss function caused by the feature split of each decision tree in the initial gradient boosting decision tree model is statistically analyzed. Based on the reduction, the gradient descent method is applied to iteratively optimize the model parameters of the initial gradient boosting decision tree model until the loss function converges, thereby obtaining the target weight and the target gradient boosting decision tree model. The model parameters include the initial weights, and the gradient descent method uses a dynamic learning rate for gradient descent.
[0046] Specifically, during the training of the initial gradient boosting decision tree model, the second multidimensional data is input into the initial gradient boosting decision tree model. The initial gradient boosting decision tree model extracts multiple sample features from the second multidimensional data. Based on the initial weights corresponding to the key features of the multiple samples, the importance of the multiple sample features is ranked to obtain multiple key features of the samples. Based on the multiple key features of the samples, the output result is obtained. This output result is the prediction result of the optimal strategy combination of the initial target device in the previous time period obtained by the initial gradient boosting decision tree model.
[0047] The optimal strategy combination label is the optimal strategy combination actually obtained for the target device in the previous period, or it is labeled by staff. Using the optimal strategy combination label and the output results, the loss function can be calculated. The type of loss function can be set as needed, such as cross-entropy loss, or other types; no specific limitation is made here.
[0048] Subsequently, the reduction in the loss function caused by the feature splitting of each decision tree in the initial gradient boosting decision tree model is statistically analyzed. Using this reduction, gradient descent is applied to iteratively optimize the model parameters of the initial gradient boosting decision tree model containing the initial weights until the loss function converges, thus obtaining the target weights and the target gradient boosting decision tree model.
[0049] Understandably, gradient descent uses a dynamic learning rate to perform gradient descent, which automatically adjusts the importance of features as the data distribution changes. This can solve the problem of poor scenario adaptability caused by using fixed weights in existing methods.
[0050] During model training, hyperparameters such as tree depth, learning rate, and number of trees can be determined through grid search or Bayesian optimization. The tree depth can be 3-8 layers, the learning rate can be 0.01-0.2, and the number of trees can be 100-500.
[0051] Based on the above embodiments, each decision tree in the initial gradient boosting decision tree model is generated using the Classification and Regression Tree (CART) algorithm. The splitting metric for feature splitting includes mean squared error or Gini coefficient. Mean squared error is applicable to regression tasks, while the Gini coefficient is applicable to classification tasks.
[0052] Based on the above embodiments, the initial gradient boosting decision tree model is constructed based on gradient boosting decision tree (GBDT) or extreme gradient boosting decision tree (XGBoost).
[0053] XGBoost can accelerate training through parallel computing, reducing the consumption of computing resources.
[0054] Based on the above embodiments, the step of obtaining the first multidimensional data for the target device in the current time period, followed by: Principal component analysis was performed on the first multidimensional data to obtain the analysis results; Accordingly, inputting the first multidimensional data into the target gradient boosting decision tree model includes: The analysis results are then input into the target gradient boosting decision tree model.
[0055] Specifically, after obtaining the first multidimensional data, principal component analysis (PCA) can be performed on the first multidimensional data to reduce its dimensionality and obtain the analysis result, namely the dimensionality-reduced first multidimensional data.
[0056] The analysis results can then be input into the target gradient boosting decision tree model (PKBB) for processing. Compared to directly processing the first multidimensional data, this significantly reduces the computational cost of the PKBB model. This PCA process is particularly suitable for situations where the first multidimensional data is high-dimensional and sparse.
[0057] Based on the above embodiments, the step of obtaining the first multidimensional data for the target device in the current time period further includes: The first multidimensional data is processed for missing values, normalization, and noise reduction to obtain the processing result; Accordingly, inputting the first multidimensional data into the target gradient boosting decision tree model includes: The processing results are then input into the target gradient boosting decision tree model.
[0058] Specifically, to further improve the computational efficiency of the target gradient boosting decision tree model, after obtaining the first multidimensional data, preprocessing operations such as missing value handling, normalization, and noise removal can be performed on the first multidimensional data to obtain the processing result.
[0059] Here, the execution order of preprocessing operations such as missing value handling, normalization, and noise removal can be set as needed, and no specific restrictions are imposed here.
[0060] Missing value handling can be achieved by filling missing numerical data with the median of each dimension of the first multidimensional data, and filling categorical data with the mode of each dimension of the first multidimensional data.
[0061] Normalization can be performed by Z-score normalization on the continuous data in the first multidimensional data to eliminate the influence of dimensions.
[0062] Noise processing can be achieved by calculating the variance of each dimension of the first multidimensional data, comparing the calculated variance with a variance threshold, and filtering out data whose variance is lower than the variance threshold in each dimension of the first multidimensional data to reduce noise interference.
[0063] The processed results can then be input into the target gradient boosting decision tree model (DDBMA) for further processing. Compared to directly processing the first multidimensional data, this method significantly improves the computational efficiency of the DDBMA. Furthermore, by handling missing values, normalizing, and reducing noise in the first multidimensional data, this method enhances the input quality of the DDBMA and avoids bias issues caused by uneven data distribution or missing data.
[0064] Based on the above embodiments, the acquisition of the first multidimensional data for the target device in the current time period can be achieved by first processing the first multidimensional data for missing values, normalization, and noise, and then performing principal component analysis on the processed data to obtain the analysis results. Alternatively, principal component analysis can be performed on the first multidimensional data first to obtain the analysis results, and then the analysis results can be processed for missing values, normalization, and noise to obtain the final result.
[0065] In summary, such as Figure 3 As shown, when training the initial gradient boosting decision tree model, the second multidimensional data can be obtained first, and then preprocessing operations such as missing value processing, normalization processing, and noise processing can be performed on the second multidimensional data, followed by principal component analysis to obtain the result data.
[0066] After that, key features can be selected from the results data and labeled to obtain the optimal strategy combination label.
[0067] Next, data splitting is performed, dividing the result data labeled with the optimal policy combination into training and test sets. The initial gradient boosting decision tree model is trained using the training set, followed by model evaluation to determine if the loss function has converged. If converged, the trained model is tested using the test set, and the test results are output. If the test result shows that the trained model passes the test, it is adopted as the target gradient boosting decision tree model. Here, loss function convergence means that the trained model has reached the target accuracy.
[0068] If the loss function does not converge, meaning the trained model does not reach the target accuracy, then update the model parameters of the initial gradient boosting decision tree model and continue training the initial gradient boosting decision tree model until the loss function converges.
[0069] The multidimensional data optimization analysis method based on a gradient boosting decision tree model provided in this invention can accurately capture nonlinear relationships in multidimensional data by utilizing the target gradient boosting decision tree model, improving the strategy optimization accuracy by more than 30%. Moreover, it has high computational efficiency and is suitable for small to medium-sized datasets. Furthermore, it provides interpretability analysis to help users understand the decision-making basis.
[0070] like Figure 4 As shown, based on the above embodiments, this embodiment of the invention provides a multidimensional data optimization analysis device based on a gradient boosting decision tree model, comprising: Data acquisition module 41 is used to acquire the first multidimensional data of the target device for the current time period; Data analysis module 42 is used to input the first multidimensional data into the target gradient boosting decision tree model, extract multiple initial features from the first multidimensional data by the target gradient boosting decision tree model, sort the multiple initial features by importance based on the target weights corresponding to the multiple initial features to obtain multiple key features, and output the optimal strategy combination for the target device based on the multiple key features. The target gradient boosting decision tree model is obtained by training the initial gradient boosting decision tree model based on the second multidimensional data of the target device in the previous time period of the current time period and the optimal strategy combination label of the target device.
[0071] Based on the above embodiments, the multidimensional data optimization analysis device based on the gradient boosting decision tree model provided in this embodiment of the invention includes the first multidimensional data and the second multidimensional data, both of which include operation data, equipment status data and environmental data of multiple actions of multiple operators on the target equipment. The optimal strategy combination and the optimal strategy combination label both include the optimal operation data, optimal equipment status data and optimal environmental data for the key actions among the multiple actions.
[0072] Based on the above embodiments, this embodiment of the invention provides a multidimensional data optimization analysis device based on a gradient boosting decision tree model. The first multidimensional data and the second multidimensional data both include different combinations of multiple parameters of the target device. The optimal strategy combination and the optimal strategy combination label both include the optimal combination of key parameters among the multiple parameters.
[0073] Based on the above embodiments, the multidimensional data optimization analysis device based on the gradient boosting decision tree model provided in this embodiment of the invention further includes a training module, used for: The second multidimensional data is input into the initial gradient boosting decision tree model, which extracts multiple sample features from the second multidimensional data. Based on the initial weights corresponding to the multiple sample key features, the multiple sample features are ranked by importance to obtain multiple sample key features. Based on the multiple sample key features, the output result is obtained. Based on the optimal strategy combination label and the output result, the loss function is calculated, and the reduction of the loss function caused by the feature split of each decision tree in the initial gradient boosting decision tree model is statistically analyzed. Based on the reduction, the gradient descent method is applied to iteratively optimize the model parameters of the initial gradient boosting decision tree model until the loss function converges, thereby obtaining the target weight and the target gradient boosting decision tree model. The model parameters include the initial weights, and the gradient descent method uses a dynamic learning rate for gradient descent.
[0074] Based on the above embodiments, the multidimensional data optimization analysis device based on the gradient boosting decision tree model provided in this embodiment of the invention is wherein each decision tree in the initial gradient boosting decision tree model is generated by a classification regression tree algorithm; the splitting metric for feature splitting includes mean squared error or Gini coefficient.
[0075] Based on the above embodiments, the multidimensional data optimization analysis device based on the gradient boosting decision tree model provided in this embodiment of the invention, wherein the initial gradient boosting decision tree model is constructed based on the gradient boosting decision tree or based on the extreme gradient boosting decision tree.
[0076] Based on the above embodiments, the multidimensional data optimization analysis device based on the gradient boosting decision tree model provided in this embodiment of the invention further includes a principal component analysis module, used for: Principal component analysis was performed on the first multidimensional data to obtain the analysis results; Accordingly, the data analysis module is specifically used for: The analysis results are then input into the target gradient boosting decision tree model.
[0077] Based on the above embodiments, the multidimensional data optimization analysis device based on the gradient boosting decision tree model provided in this embodiment of the invention further includes a preprocessing module, used for: The first multidimensional data is processed for missing values, normalization, and noise reduction to obtain the processing result; Accordingly, the data analysis module is specifically used for: The processing results are then input into the target gradient boosting decision tree model.
[0078] Specifically, the functions of each module in the multidimensional data optimization analysis device based on the gradient boosting decision tree model provided in this embodiment of the invention correspond one-to-one with the operation flow of each step in the above method-like embodiments, and the achieved effects are also the same. For details, please refer to the above embodiments, and this will not be repeated in this embodiment of the invention.
[0079] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include a processor 510, a communications interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communications interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute the multidimensional data optimization analysis method based on the gradient boosting decision tree model provided in the above embodiments.
[0080] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to related technologies, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0081] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the multidimensional data optimization analysis method based on the gradient boosting decision tree model provided in the above embodiments.
[0082] In another aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multidimensional data optimization analysis method based on the gradient boosting decision tree model provided in the above embodiments. This computer-readable storage medium can be either a non-transitory computer-readable storage medium or a transient computer-readable storage medium, and is not specifically limited herein.
[0083] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0084] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of software products. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A multidimensional data optimization analysis method based on a gradient boosting decision tree model, characterized in that, include: Obtain the first multidimensional data for the target device in the current time period; The first multidimensional data is input into the target gradient boosting decision tree model, which extracts multiple initial features from the first multidimensional data. Based on the target weights corresponding to the multiple initial features, the multiple initial features are ranked by importance to obtain multiple key features. Based on the multiple key features, the optimal strategy combination for the target device is output. The target gradient boosting decision tree model is obtained by training the initial gradient boosting decision tree model of the previous time period based on the second multidimensional data of the target device in the previous time period and the optimal strategy combination label of the target device in the current time period.
2. The multidimensional data optimization analysis method based on gradient boosting decision tree model according to claim 1, characterized in that, Both the first multidimensional data and the second multidimensional data include operation data, equipment status data, and environmental data of multiple actions performed by multiple operators on the target device. Both the optimal strategy combination and the optimal strategy combination label include optimal operation data, optimal equipment status data, and optimal environmental data for the key actions among the multiple actions.
3. The multidimensional data optimization analysis method based on gradient boosting decision tree model according to claim 1, characterized in that, Both the first multidimensional data and the second multidimensional data include different combinations of values for multiple parameters of the target device, and both the optimal strategy combination and the optimal strategy combination label include the optimal combination of values for key parameters among the multiple parameters.
4. The multidimensional data optimization analysis method based on gradient boosting decision tree model according to claim 1, characterized in that, The training steps of the target gradient boosting decision tree model include: The second multidimensional data is input into the initial gradient boosting decision tree model, which extracts multiple sample features from the second multidimensional data. Based on the initial weights corresponding to the multiple sample key features, the multiple sample features are ranked by importance to obtain multiple sample key features. Based on the multiple sample key features, the output result is obtained. Based on the optimal strategy combination label and the output result, the loss function is calculated, and the reduction of the loss function caused by the feature split of each decision tree in the initial gradient boosting decision tree model is statistically analyzed. Based on the reduction, the gradient descent method is applied to iteratively optimize the model parameters of the initial gradient boosting decision tree model until the loss function converges, thereby obtaining the target weight and the target gradient boosting decision tree model. The model parameters include the initial weights, and the gradient descent method uses a dynamic learning rate for gradient descent.
5. The multidimensional data optimization analysis method based on gradient boosting decision tree model according to claim 4, characterized in that, Each decision tree in the initial gradient boosting decision tree model is generated using a classification and regression tree algorithm; the splitting metric for feature splitting includes mean squared error or Gini coefficient.
6. The multidimensional data optimization analysis method based on gradient boosting decision tree model according to claim 1, characterized in that, The initial gradient boosting decision tree model is constructed based on a gradient boosting decision tree, or based on an extreme gradient boosting decision tree.
7. The multidimensional data optimization analysis method based on a gradient boosting decision tree model according to any one of claims 1-6, characterized in that, The process of obtaining the first multidimensional data for the target device in the current time period, followed by: Principal component analysis was performed on the first multidimensional data to obtain the analysis results; Accordingly, inputting the first multidimensional data into the target gradient boosting decision tree model includes: The analysis results are then input into the target gradient boosting decision tree model.
8. The multidimensional data optimization analysis method based on a gradient boosting decision tree model according to any one of claims 1-6, characterized in that, The process of obtaining the first multidimensional data for the target device in the current time period also includes: The first multidimensional data is processed for missing values, normalization, and noise reduction to obtain the processing result; Accordingly, inputting the first multidimensional data into the target gradient boosting decision tree model includes: The processing results are then input into the target gradient boosting decision tree model.
9. A multidimensional data optimization analysis device based on a gradient boosting decision tree model, characterized in that, include: The data acquisition module is used to acquire the first multidimensional data for the target device in the current time period; The data analysis module is used to input the first multidimensional data into the target gradient boosting decision tree model, extract multiple initial features from the first multidimensional data by the target gradient boosting decision tree model, sort the multiple initial features by importance based on the target weights corresponding to the multiple initial features to obtain multiple key features, and output the optimal strategy combination for the target device based on the multiple key features. The target gradient boosting decision tree model is obtained by training the initial gradient boosting decision tree model based on the second multidimensional data of the target device in the previous time period of the current time period and the optimal strategy combination label of the target device.
10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the multidimensional data optimization analysis method based on the gradient boosting decision tree model as described in any one of claims 1-8.
Citation Information
Patent Citations
Machine learning-based model training and classification method and device, equipment and medium
CN119669919A
Carbon emission prediction method based on adaptive weight optimization
CN119963224A
Cited By
Heterogeneous computing resource intelligent scheduling method and system for electric charge computing task
CN122044890A
Prediction method and system for formation classification in the field of oil exploration based on explainability diagnosis driving
CN122508359A
