A cigarette filter rod ventilation rate prediction method and device based on local feature enhancement and XGBoost, an electronic device, and a storage medium
Patent Information
- Application Number
- CN202611077538.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-20
- Publication Date
- 2026-10-20
AI Technical Summary
[0007]本发明的目的在于提出一种基于局部特征增强与XGBoost的卷烟滤棒通风率预测方法、装置、电子设备及储存介质，以解决由于卷烟滤棒激光打孔过程中，因孔形态随机波动及卷烟滤棒通风率与孔形态之间非线性因素影响，造成卷烟滤棒通风率预测困难的问题
[0046]1、本发明中的一种基于局部特征增强与XGBoost的卷烟滤棒通风率预测方法，先获取每支烟支滤棒上所打孔后的孔形态数据，并将孔形态数据与烟支滤棒对应通风率数据整合，得到烟支打孔多维数据，然后确定去零处理后烟支打孔多维数据中偏移较大烟支打孔数据，对偏移较大烟支打孔数据进行平滑/降噪处理，将平滑/降噪处理后的烟支打孔多维数据分为烟支打孔测试数据及烟支打孔训练数据，接着基于烟支打孔训练数据及早停机制优化策略，对卷烟滤棒通风率预测模型进行训练，得到训练后的卷烟滤棒通风率预测模型；其中，卷烟滤棒通风率预测模型包括数据局部特征增强模块、特征筛选模块及模型预测模块；最后将烟支打孔测试数据输入训练后的卷烟滤棒通风率预测模型中，输出与烟支打孔所对应的烟支滤棒通风率；如此，不仅可将激光打孔过程中形成的卷烟滤棒两排小孔的形态参数与卷烟滤棒通风率联系起来，还可通过孔形态的波动来反映卷烟滤棒通风率的波动，进而支撑卷烟生产过程中卷烟通风率的预测调控，以解决由于卷烟滤棒激光打孔过程中，因孔形态随机波动及卷烟滤棒通风率与孔形态之间非线性因素影响，造成卷烟滤棒通风率预测困难的问题。
Smart Images
Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
[0001] Electronic devices and storage media Technical Field
[0002] This invention relates to the field of cigarette filter ventilation rate prediction technology, and in particular to a method, device, electronic device and storage medium for predicting cigarette filter ventilation rate based on local feature enhancement and XGBoost. Background Technology
[0003] Cigarette ventilation rate is one of the important physical testing indicators for evaluating the quality of cigarettes. Fluctuations in ventilation rate can lead to imbalances in smoke composition, significant differences in smoking experience, and even affect product compliance and batch consistency in the market. Therefore, accurate prediction and control of ventilation rate are crucial for controlling the release of tar and nicotine from cigarette smoke. Cigarette ventilation rate mainly consists of two parts: cigarette paper ventilation rate and filter rod ventilation rate. To control cigarette ventilation rate, it is necessary to measure the value of the cigarette filter rod ventilation rate.
[0004] At present, domestic and foreign research on cigarette ventilation rate mainly focuses on the influence of cigarette weight, circumference, length, cigarette material, tobacco characteristics, physical indicators and production process on cigarette ventilation rate. There are also studies on the effect of hole morphology in the laser drilling process on ventilation rate, but the research process on the effect of hole morphology in the laser drilling process on ventilation rate still has the following defects: (1) Due to the influence of various nonlinear factors in the laser drilling process, the ventilation rate of cigarette filter rods fluctuates randomly and is difficult to control.
[0005] (2) Since the learning and prediction ability of the nonlinear relationship between various factors is not strong during the laser drilling process, the accuracy of predicting the ventilation rate of cigarettes needs to be improved.
[0006] Therefore, it is necessary to improve and optimize the method for predicting the ventilation rate of cigarette filters in order to solve the problem that the prediction of the ventilation rate of cigarette filters is difficult due to the random fluctuation of hole shape and the nonlinear factors between the ventilation rate of cigarette filters and hole shape during the laser drilling process. Summary of the Invention
[0007] The purpose of this invention is to propose a method, device, electronic device, and storage medium for predicting the ventilation rate of cigarette filter rods based on local feature enhancement and XGBoost, so as to solve the problem that it is difficult to predict the ventilation rate of cigarette filter rods due to the random fluctuation of hole morphology and the influence of nonlinear factors between the ventilation rate of cigarette filter rods and hole morphology during the laser drilling process.
[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:
[0009] This application proposes a method for predicting the ventilation rate of cigarette filter rods based on local feature enhancement and XGBoost in its first aspect, including the following steps:
[0010] The hole morphology data of each cigarette filter rod after drilling is obtained, and the hole morphology data is integrated with the corresponding ventilation rate data of the cigarette filter rod to obtain multi-dimensional data of cigarette perforation; among which, the hole morphology data includes the number of holes, hole length, hole width, hole depth, hole spacing, and the surface area of the upper and inner surfaces of the holes; each cigarette is a cigarette after laser perforation;
[0011] After removing zeros, identify cigarette punch data with large offsets in the multidimensional data of cigarette punching. Perform smoothing / denoising processing on the cigarette punch data with large offsets. Divide the smoothed / denoised multidimensional data of cigarette punching into cigarette punching test data and cigarette punching training data. The smoothing / denoising processing is SG filtering and sliding window moving average method.
[0012] Based on the training data of cigarette punching and the optimization strategy of early stop mechanism, the cigarette filter rod ventilation rate prediction model is trained to obtain the trained cigarette filter rod ventilation rate prediction model; the cigarette filter rod ventilation rate prediction model is constructed based on the XGBoost algorithm; the cigarette filter rod ventilation rate prediction model includes a data local feature enhancement module, a feature selection module and a model prediction module.
[0013] Input the cigarette perforation test data into the trained cigarette filter rod ventilation rate prediction model, and output the cigarette filter rod ventilation rate corresponding to the cigarette perforation.
[0014] Prior to this, the step of acquiring the hole morphology data of each cigarette filter rod after perforation, and integrating the hole morphology data with the corresponding ventilation rate data of the cigarette filter rod to obtain multi-dimensional data of cigarette perforation, includes the following steps:
[0015] The number of holes obtained from the multidimensional data of cigarette perforation on each cigarette is determined; wherein, the number of holes obtained from the multidimensional data of cigarette perforation on each cigarette is 30; each cigarette is a cigarette of the same brand after being perforated by a laser perforation device;
[0016] Based on the number of holes obtained from the multidimensional data of perforation on each cigarette, the hole morphology data and ventilation rate of each hole are obtained; the hole morphology data includes hole length, hole width, hole depth, hole spacing, hole upper surface area, and hole inner surface area; the hole morphology data and ventilation rate of each hole are obtained by measurement.
[0017] By integrating the hole shape data and ventilation rate corresponding to each hole, multi-dimensional data of cigarette perforation is obtained; among which, multi-dimensional data of cigarette perforation is multi-dimensional data of perforation for each cigarette.
[0018] Prioritize the process of identifying cigarette punch data with significant offsets in the multidimensional data after zero-removal processing, performing smoothing / denoising processing on the cigarette punch data with significant offsets, and dividing the smoothed / denoised multidimensional data into cigarette punch test data and cigarette punch training data, including the following steps:
[0019] Zero-value data in the multidimensional data of cigarette punching is removed to obtain the multidimensional data of cigarette punching after removing zero values; where the zero-value data in the multidimensional data of cigarette punching is the zero-value data of hole morphology data and the zero-value data of ventilation rate corresponding to the hole are removed.
[0020] The excessively deviated cigarette perforation data in the multidimensional data of cigarette perforation after removing zero values is identified by the 3σ criterion. The excessively deviated cigarette perforation data is smoothed by SG filtering based on local polynomial least squares fitting to obtain the smoothed multidimensional data of cigarette perforation.
[0021] The noise reduction process of the smoothed cigarette punch multidimensional data was performed by the sliding window moving average method to obtain the noise-reduced cigarette punch multidimensional data.
[0022] The noise-reduced multidimensional data of cigarette punching were divided into cigarette punching test data and cigarette punching training data in an 8:2 ratio.
[0023] Preferably, the cigarette filter rod ventilation rate prediction model includes a data local feature enhancement module, a feature filtering module, and a model prediction module;
[0024] Based on training data of cigarette perforation and optimization strategies of early stop mechanism, a cigarette filter rod ventilation rate prediction model is trained to obtain the trained cigarette filter rod ventilation rate prediction model, including the following steps:
[0025] In the local feature enhancement module for the input data of cigarette punching training data, the local feature enhancement module performs local feature enhancement processing on the cigarette punching training data to obtain the cigarette punching features after local feature enhancement.
[0026] The enhanced local feature features of the cigarette punch are input into the feature filtering module. The feature filtering module performs feature selection processing on the enhanced local feature features of the cigarette punch to obtain the feature-selected cigarette punch features.
[0027] The perforation features of the cigarette after feature selection are input into the model prediction module, and combined with the early stop mechanism optimization strategy, the model prediction module is trained to obtain the trained cigarette filter rod ventilation rate prediction model.
[0028] Preferably, the local feature enhancement module is specifically used for:
[0029] Based on training data of cigarette punching, the sliding mean and sliding standard deviation were calculated with window lengths of 5, 10 and 20, and feature enhancement was performed on the mean of the morphology of the 5 holes in the cigarette filter rod.
[0030] After calculating the moving mean and moving standard deviation, edge padding is used to ensure that the feature length is consistent with the original sample; in this case, 6 features are added to each morphology, for a total of 35 features.
[0031] Preferably, the feature filtering module is specifically used for:
[0032] For all the enhanced local features of the cigarette punch, calculate the mutual information between each enhanced local feature of the cigarette punch and the preset label; there are 35 enhanced local feature features of the cigarette punch.
[0033] Based on the mutual information between the enhanced local feature of the cigarette and the preset label, and combined with the Top-K strategy, the top 10% of the enhanced local feature features of the cigarette are selected.
[0034] Prior to this, the early stopping mechanism optimization strategy is specifically as follows:
[0035] When the error of the cigarette filter ventilation rate prediction model does not decrease for 20 consecutive rounds, the training of the cigarette filter ventilation rate prediction model will be automatically stopped, and the optimal number of iterations will be used to determine the trained cigarette filter ventilation rate prediction model.
[0036] In its second aspect, this application proposes a cigarette filter rod ventilation rate prediction device based on local feature enhancement and XGBoost, comprising:
[0037] The cigarette perforation multidimensional data acquisition module is used to acquire the hole morphology data of each cigarette filter rod after perforation, and integrate the hole morphology data with the corresponding ventilation rate data of the cigarette filter rod to obtain the cigarette perforation multidimensional data; wherein, the hole morphology data includes the number of holes, hole length, hole width, hole depth, hole spacing, hole upper surface and hole inner surface area; each cigarette is a cigarette after laser perforation;
[0038] The data smoothing / denoising module is used to identify cigarette punch data with large offsets in the multidimensional data of cigarette punching after zero removal, and to perform smoothing / denoising on the cigarette punch data with large offsets. The smoothed / denoising multidimensional data of cigarette punching is divided into cigarette punching test data and cigarette punching training data. The smoothing / denoising process is SG filtering and sliding window moving average method.
[0039] The cigarette filter ventilation rate prediction model module is used to train the cigarette filter ventilation rate prediction model based on cigarette punching training data and early stop mechanism optimization strategy, so as to obtain the trained cigarette filter ventilation rate prediction model; wherein, the cigarette filter ventilation rate prediction model is constructed based on the XGBoost algorithm; the cigarette filter ventilation rate prediction model includes a data local feature enhancement module, a feature selection module and a model prediction module.
[0040] The cigarette filter rod ventilation rate prediction module is used to input the cigarette punching test data into the trained cigarette filter rod ventilation rate prediction model and output the cigarette filter rod ventilation rate corresponding to the cigarette punching.
[0041] This application proposes an electronic device in a third aspect, including...
[0042] One or more processors;
[0043] The memory is used to store one or more programs; when one or more programs are executed by one or more processors, the one or more processors implement a cigarette filter ventilation rate prediction method based on local feature enhancement and XGBoost.
[0044] In its fourth aspect, this application proposes a computer-readable storage medium storing computer instructions for causing a processor to execute a cigarette filter ventilation rate prediction method based on local feature enhancement and XGBoost.
[0045] Compared with the prior art, the present invention has at least one of the following beneficial effects:
[0046] 1. A method for predicting the ventilation rate of cigarette filters based on local feature enhancement and XGBoost in this invention first acquires the hole morphology data of each cigarette filter after perforation, and integrates the hole morphology data with the corresponding ventilation rate data of the cigarette filter to obtain multi-dimensional data of cigarette perforation. Then, it identifies the cigarette perforation data with large offsets in the multi-dimensional data of cigarette perforation after zero-removal processing, and performs smoothing / noise reduction processing on the cigarette perforation data with large offsets. The smoothed / noise-reduced multi-dimensional data of cigarette perforation is divided into cigarette perforation test data and cigarette perforation training data. Then, based on the cigarette perforation training data and the early stop mechanism optimization strategy, the cigarette filter ventilation rate prediction model is trained to obtain the trained cigarette filter ventilation rate prediction model; wherein... The cigarette filter rod ventilation rate prediction model includes a data local feature enhancement module, a feature selection module, and a model prediction module. Finally, the cigarette punching test data is input into the trained cigarette filter rod ventilation rate prediction model, and the output is the cigarette filter rod ventilation rate corresponding to the cigarette punching. In this way, not only can the morphological parameters of the two rows of small holes formed in the cigarette filter rod during the laser punching process be linked to the cigarette filter rod ventilation rate, but the fluctuation of the hole morphology can also reflect the fluctuation of the cigarette filter rod ventilation rate. This supports the prediction and control of the cigarette ventilation rate during the cigarette production process, so as to solve the problem of difficulty in predicting the cigarette filter rod ventilation rate caused by the random fluctuation of the hole morphology and the nonlinear factors between the cigarette filter rod ventilation rate and the hole morphology during the laser punching process.
[0047] 2. In this invention, the local feature enhancement module performs local feature enhancement processing on the cigarette punch training data, and then the feature selection module performs feature selection processing on the enhanced cigarette punch features. In this way, not only is the feature dimension and local correlation increased, but also the foundation is laid for subsequent high-precision prediction, so as to solve the problems of few input features and weak correlation between local features.
[0048] 3. In this invention, the local feature enhancement module is first used to perform local feature enhancement processing on the training data for perforated cigarettes.
[0049] Then, the feature selection module performs feature selection processing on the cigarette perforation features after local feature enhancement. Finally, the feature selection-processed cigarette perforation features are input into the model prediction module, and the model prediction module is trained by combining the early stop mechanism optimization strategy to obtain the trained cigarette filter ventilation rate prediction model. In this way, the feature enhancement and model training are optimized by mutual information selection and early stop mechanism respectively, which further improves the prediction accuracy of the cigarette filter ventilation rate prediction model. Attached Figure Description
[0050] Figure 1 This is a flowchart of a cigarette filter rod ventilation rate prediction method based on local feature enhancement and XGBoost in this invention.
[0051] Figure 2 This is a schematic diagram of the cigarette filter rod ventilation rate prediction model of the present invention.
[0052] Figure 3 This is a schematic diagram of the fitting curve between the predicted and actual values of the filter rod ventilation rate in an embodiment of the present invention.
[0053] Figure 4 This is a scatter plot showing the predicted and actual values of the filter rod ventilation rate in an embodiment of the present invention.
[0054] Figure 5 This is a schematic diagram of a cigarette filter rod ventilation rate prediction device based on local feature enhancement and XGBoost in this invention.
[0055] Figure 6 This is a schematic diagram of the structure of an electronic device according to the present invention.
[0056] Table 1 shows the multidimensional data on cigarette punching.
[0057] Table 2 compares the ventilation rate prediction performance of each model. Detailed Implementation
[0058] like Figure 1-6 As shown, to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.
[0059] Example 1
[0060] Figure 1 This invention provides a flowchart of a method for predicting the ventilation rate of cigarette filters based on local feature enhancement and XGBoost, according to Embodiment 1. This embodiment is applicable to the prediction of the ventilation rate of cigarette filters. The method can be executed by a cigarette filter ventilation rate prediction device based on local feature enhancement and XGBoost. This device can be implemented in hardware and / or software and can be configured in an electronic device with data processing capabilities. Figure 1 As shown, the method includes:
[0061] S110. Obtain the hole shape data after drilling on each cigarette filter rod, and integrate the hole shape data with the corresponding ventilation rate data of the cigarette filter rod to obtain multi-dimensional data of cigarette drilling.
[0062] The hole morphology data includes the number of holes, hole length, hole width, hole depth, hole spacing, and the surface area of the upper and inner surfaces of the holes; each cigarette is a cigarette after laser drilling.
[0063] Optionally, the pore morphology data of each cigarette filter rod after perforation is obtained, and the pore morphology data is integrated with the corresponding ventilation rate data of the cigarette filter rod to obtain multi-dimensional data of cigarette perforation, including steps A1 to A3:
[0064] A1. Determine the number of holes in the multidimensional data of cigarette perforation obtained from each cigarette.
[0065] The number of holes obtained from the multi-dimensional data of the perforation on each cigarette is 30; each cigarette is a cigarette of the same brand after being perforated by a laser perforation device.
[0066] A2. Based on the number of holes obtained from the multi-dimensional data of perforation on each cigarette, obtain the hole morphology data and ventilation rate corresponding to each hole.
[0067] The hole morphology data includes hole length, hole width, hole depth, hole spacing, hole upper surface area, and hole inner surface area; the hole morphology data and ventilation rate for each hole are obtained through measurement.
[0068] A3. Integrate the hole shape data and ventilation rate corresponding to each hole to obtain multi-dimensional data on cigarette punching.
[0069] Among them, the multidimensional data of cigarette punching is the multidimensional data of punching per cigarette. Specifically, refer to Table 1, which contains the multidimensional data of cigarette punching.
[0070] Table 1: Multidimensional Data on Cigarette Perforation
[0071] 1 0.362 0.171 0.103 0.0093 0.010477 1.133 22.22 2 0.359 0.17 0.1 0.0094 0.01054 1.133 26.28 3 0.375 0.174 0.115 0.0099 0.011705 1.133 32.82 4 0.365 0.174 0.113 0.0107 0.012625 1.133 30.43 5 0.344 0.171 0.094 0.0085 0.009385 1.133 24.87 6 0.376 0.171 0.111 0.0104 0.01215 1.133 27.6 7 0.351 0.171 0.092 0.0085 0.009524 1.034 27.06 8 0.357 0.175 0.116 0.0096 0.011419 1.101 28.95 9 0.37 0.176 0.109 0.0104 0.012072 1.133 30.87
[0072] S120. Identify cigarette punching data with large offsets in the multidimensional data of cigarette punching after zero removal. Perform smoothing / noise reduction processing on the cigarette punching data with large offsets. Divide the smoothed / noise reduction multidimensional data of cigarette punching into cigarette punching test data and cigarette punching training data.
[0073] The smoothing / noise reduction process consists of SG filtering and sliding window moving average.
[0074] Optionally, identify cigarette punching data with large offsets in the multidimensional data of cigarette punching after zero-removal processing, perform smoothing / denoising processing on the cigarette punching data with large offsets, and divide the smoothed / denoised multidimensional data of cigarette punching into cigarette punching test data and cigarette punching training data, including steps B1 to B4:
[0075] B1. Remove zero-value data from the multidimensional data of cigarette punching to obtain multidimensional data of cigarette punching after removing zero values.
[0076] Among them, the zero-value data in the multidimensional data of cigarette punching are the zero-value data after removing the hole morphology data and the zero-value data of the ventilation rate corresponding to the hole.
[0077] B2. The excessively deviated cigarette perforation data in the multidimensional data of cigarette perforation after removing zero values is determined by the 3σ criterion. The excessively deviated cigarette perforation data is smoothed by SG filtering based on local polynomial least squares fitting to obtain the smoothed multidimensional data of cigarette perforation.
[0078] The calculation process for the 3σ criterion is as follows:
[0079]
[0080]
[0081]
[0082] in, For signal, For the number of signals, The mean of the signal, when If the condition is met, it indicates that the point is an outlier.
[0083] The SG filter calculation process for outlier data is as follows:
[0084]
[0085] in, Let be the window length. These are the fitting coefficients. The value after signal smoothing
[0086] B3. The smoothed cigarette punching multidimensional data is denoised using the sliding window moving average method to obtain the denoised cigarette punching multidimensional data.
[0087] The calculation process for the sliding window moving average method is as follows:
[0088]
[0089] in, This is the value after signal smoothing. This is the window length.
[0090] The eigenvalue standardization calculation process is as follows:
[0091]
[0092] in, These are the standardized eigenvalues. These are the original eigenvalues. The mean of this feature for the training set, The standard deviation of this feature in the training set.
[0093] B4. The noise-reduced multidimensional data of cigarette punching were divided into cigarette punching test data and cigarette punching training data in an 8:2 ratio.
[0094] S130. Based on the training data of cigarette punching and the optimization strategy of early stop mechanism, the cigarette filter rod ventilation rate prediction model is trained to obtain the trained cigarette filter rod ventilation rate prediction model.
[0095] The cigarette filter ventilation rate prediction model is based on the XGBoost algorithm; the cigarette filter ventilation rate prediction model includes a data local feature enhancement module, a feature selection module, and a model prediction module.
[0096] Optionally, based on the cigarette perforation training data and the early stop mechanism optimization strategy, the cigarette filter rod ventilation rate prediction model is trained to obtain the trained cigarette filter rod ventilation rate prediction model, including steps C1 to C3:
[0097] C1. In the local feature enhancement module of the cigarette punching training data input data, the local feature enhancement module performs local feature enhancement processing on the cigarette punching training data to obtain the cigarette punching features after local feature enhancement.
[0098] The specific processing steps of the local feature enhancement module are as follows: steps ab:
[0099] a. Based on the training data of cigarette punching, the sliding mean and sliding standard deviation are calculated with window lengths of 5, 10 and 20, and feature enhancement is performed on the mean of the 5 holes of the cigarette filter rod.
[0100] b. After calculating the moving mean and moving standard deviation, edge padding is used to ensure that the feature length is consistent with the original sample; among them, 6 features are added for each morphology, for a total of 35 features.
[0101] In industrial testing, data is often instantaneous, and individual data points are highly susceptible to equipment errors, resulting in noise. Therefore, noise reduction and statistical methods are frequently used in data processing to mitigate the impact of noise. Lightweight sliding window features are a data processing method that extracts local features from data using a fixed window length. This effectively extracts local fluctuations and trends, preserves local statistical features, and avoids introducing redundant information. By progressively sliding a fixed-length window, statistical, morphological, and distributional features can be calculated within each window to capture local patterns, trends, and fluctuations.
[0102] The moving average can reflect the central location and overall level of data, enabling analysis of central tendency. Assume there are N signals, and a certain characteristic is... Then the feature has N feature values. Assuming the sliding window length is If the sliding step size is 1, then the formula for calculating the sliding mean is:
[0103]
[0104] in, It is the moving average. This is the moving average index, with values ranging from 1 to 10. N eigenvalues can be used to calculate N moving averages, and the remaining ones... Each moving average is padded with edge fill.
[0105] The moving standard deviation reflects the dispersion and volatility of data, enabling dispersion or variability analysis. The calculation formula is:
[0106]
[0107] in, It is the moving average. This is the moving average index, with values ranging from 1 to 10. N eigenvalues can be calculated One moving standard deviation is used, and the remaining moving standard deviations are filled with edge padding.
[0108] After the sliding window is processed, there will be a final... If there are no sliding feature values after a given feature, and to ensure that the sliding feature is consistent with the original feature and to facilitate model training, edge padding can be used to complete the data. Assuming the sliding window processing... The features are ,but The following Each feature value is filled with the i-th feature value, calculated as follows:
[0109] Moving average:
[0110]
[0111] Sliding standard deviation:
[0112]
[0113] C2. Input the enhanced local feature cigarette punching feature into the feature filtering module. The feature filtering module performs feature selection processing on the enhanced local feature cigarette punching feature to obtain the feature-selected cigarette punching feature.
[0114] The specific processing steps of the feature selection module are as follows: Step de:
[0115] d. For all the enhanced local features of the cigarette punching feature, calculate the mutual information between each enhanced local feature of the cigarette punching feature and the preset label.
[0116] Among them, the enhanced local features of the cigarette punch feature consist of 35 features.
[0117] It should be noted that when performing local feature enhancement through a sliding window, the moving mean and moving standard deviation of all features are calculated. Introducing a large number of features can lead to increased data dimensionality and information redundancy. Furthermore, not all features need to be enhanced locally, so feature selection is still required after feature enhancement.
[0118] e. Based on the mutual information between the enhanced cigarette punch features and the preset labels, and combined with the Top-K strategy, select the top 10% of the enhanced cigarette punch features.
[0119] Mutual information can effectively measure the nonlinear relationship between features and labels, is not limited by data distribution, and has strong applicability. In high-dimensional sliding window features, it can eliminate redundancy and retain key information, reducing data dimensionality and computational cost while improving model generalization ability, providing a stable and efficient feature foundation for subsequent prediction modeling. Its discrete form calculation method is as follows:
[0120]
[0121] Where p(x,y) is the joint probability distribution of feature X and target Y, and p(x) and p(y) are their marginal probability distributions, respectively.
[0122] Meanwhile, to avoid the influence of different dimensions between different features, the mutual information is normalized. The specific calculation method is as follows:
[0123]
[0124] Where H(X) and H(Y) are the information entropy of the feature and the target, respectively.
[0125] C3. The perforation features of the cigarette after feature selection are input into the model prediction module, and combined with the early stop mechanism optimization strategy, the model prediction module is trained to obtain the trained cigarette filter rod ventilation rate prediction model.
[0126] The early stop mechanism optimization strategy is as follows: when the error of the cigarette filter rod ventilation rate prediction model does not decrease for 20 consecutive rounds, the training of the cigarette filter rod ventilation rate prediction model is automatically stopped, and the optimal number of iterations is used to determine the trained cigarette filter rod ventilation rate prediction model.
[0127] The specific training process of the model training and prediction module is as follows:
[0128] For the number of training samples The XGBoost algorithm can be simply represented as an additive model, and the specific calculation method is as follows:
[0129]
[0130] in For the i-th input sample, Let i be the model prediction value for the i-th sample. This represents the prediction result of the k-th sub-model for the i-th sample.
[0131] To improve model generalization ability and avoid overfitting, the XGBoost algorithm introduces a regularization term into the objective function. The new objective function consists of two parts: the loss function and the regularization term. The specific calculation method is as follows:
[0132]
[0133] in, For the sample size, For the loss function, For the sample true value, is the regularization term, and is the k-th weak learner.
[0134] The specific calculation method for the loss function is as follows:
[0135]
[0136] in, and For the samples respectively The first and second gradients are calculated as follows:
[0137] The specific calculation method for the regularization term is as follows:
[0138]
[0139] Where T represents the number of leaf nodes, and The parameter for the regularization term. Let be the weight of the j-th leaf node.
[0140] Introducing the second-order Taylor expansion, the specific calculation method for minimizing the objective function is as follows:
[0141]
[0142] The training and solving of the XGBoost algorithm is to... The training and solution are then performed. Furthermore, to improve model performance, this paper optimizes the number of weak learners using an early stopping mechanism.
[0143] S140. Input the cigarette perforation test data into the trained cigarette filter rod ventilation rate prediction model, and output the cigarette filter rod ventilation rate corresponding to the cigarette perforation.
[0144] The technical solution of this application embodiment first obtains the hole shape data of each cigarette filter rod after drilling, and integrates the hole shape data with the corresponding ventilation rate data of the cigarette filter rod to obtain multi-dimensional data of cigarette drilling. Then, it identifies the cigarette drilling data with large offset in the multi-dimensional data of cigarette drilling after zero-removal processing, and performs smoothing / noise reduction processing on the cigarette drilling data with large offset. The smoothed / noise-reduced multi-dimensional data of cigarette drilling is divided into cigarette drilling test data and cigarette drilling training data. Then, based on the cigarette drilling training data and the early stop mechanism optimization strategy, the ventilation rate prediction model of the cigarette filter rod is trained to obtain the trained cigarette filter rod ventilation rate prediction model; wherein, the cigarette filter rod ventilation rate prediction model includes The model comprises a data local feature enhancement module, a feature selection module, and a model prediction module. Finally, the cigarette perforation test data is input into the trained cigarette filter rod ventilation rate prediction model, which outputs the ventilation rate of the cigarette filter rod corresponding to the cigarette perforation. This not only links the morphological parameters of the two rows of small holes formed during laser perforation of the cigarette filter rod with the cigarette filter rod ventilation rate, but also reflects the fluctuation of the cigarette filter rod ventilation rate through the fluctuation of the hole morphology. This supports the prediction and control of the cigarette ventilation rate during cigarette production, solving the problem of difficulty in predicting the ventilation rate of cigarette filters due to random fluctuations in hole morphology and the nonlinear factors between the cigarette filter rod ventilation rate and hole morphology during laser perforation.
[0145] To verify the predictive performance of the model, the method of this invention was compared with commonly used small-sample regression prediction methods such as Random Forest, CatBoost, LightGBM, XGBoost, and LightGBM+XGBoost. The specific results are shown in Table 2.
[0146] Table 2 compares the ventilation rate prediction performance of each model.
[0147] Random Forest 0.8918 4.0712 1.4196 2.0891 CatBoost 0.9013 4.4603 1.6273 2.1119 LightGBM 0.9147 3.8527 1.5198 1.9628 XGBoost 0.916 3.7971 1.4950 1.9486 LightGBM+XGBoost 0.9128 3.9424 1.5392 1.9856 This article's method 0.9681 1.4393 0.9249 1.1997
[0148] As shown in Table 2, the method of the present invention has the largest R² and the smallest MSE, MAE and RMSE, proving that the method of the present invention has a good predictive effect on the ventilation rate of cigarette filter rods.
[0149] The fitting effect between the predicted value and the true value obtained by the method of the present invention is as follows: Figure 3 As shown; the relative distribution of predicted values and actual values is as follows. Figure 4 As shown.
[0150] Example 2
[0151] Figure 5 This invention provides a structural block diagram of a cigarette filter ventilation rate prediction device based on local feature enhancement and XGBoost. This embodiment is applicable to the prediction of cigarette filter ventilation rate. This cigarette filter ventilation rate prediction device based on local feature enhancement and XGBoost can be implemented in hardware and / or software, and can be configured in an electronic device with data processing capabilities. Figure 5 As shown, the cigarette filter rod ventilation rate prediction device based on local feature enhancement and XGBoost in this embodiment may include: S210, a cigarette perforation multi-dimensional data acquisition module, S220, a data smoothing / noise reduction processing module, S230, a cigarette filter rod ventilation rate prediction model module, and S240, a cigarette filter rod ventilation rate prediction module.
[0152] Among them: S210, the multi-dimensional data acquisition module for cigarette perforation, is used to acquire the hole morphology data after perforation on each cigarette filter rod, and integrate the hole morphology data with the corresponding ventilation rate data of the cigarette filter rod to obtain multi-dimensional data of cigarette perforation; wherein, the hole morphology data includes the number of holes, hole length, hole width, hole depth, hole spacing, hole upper surface and hole inner surface area; each cigarette is a cigarette after laser perforation;
[0153] S220, the data smoothing / denoising module, is used to identify cigarette punch data with large offsets in the multidimensional data of cigarette punching after zero removal, and to perform smoothing / denoising processing on the cigarette punch data with large offsets. The smoothed / denoising multidimensional data of cigarette punching is divided into cigarette punching test data and cigarette punching training data. The smoothing / denoising processing is SG filtering and sliding window moving average method.
[0154] S230, the cigarette filter ventilation rate prediction model module, is used to train the cigarette filter ventilation rate prediction model based on cigarette perforation training data and early stop mechanism optimization strategy, to obtain the trained cigarette filter ventilation rate prediction model; wherein, the cigarette filter ventilation rate prediction model is constructed based on the XGBoost algorithm; the cigarette filter ventilation rate prediction model includes a data local feature enhancement module, a feature selection module, and a model prediction module;
[0155] S240, the cigarette filter rod ventilation rate prediction module, is used to input the cigarette punching test data into the trained cigarette filter rod ventilation rate prediction model and output the cigarette filter rod ventilation rate corresponding to the cigarette punching.
[0156] The technical solution of this application embodiment first obtains the hole shape data of each cigarette filter rod after drilling through a cigarette perforation multidimensional data acquisition module, and integrates the hole shape data with the corresponding ventilation rate data of the cigarette filter rod to obtain cigarette perforation multidimensional data; then, a data smoothing / noise reduction processing module identifies the cigarette perforation data with large offsets in the cigarette perforation multidimensional data after zero-removal processing, performs smoothing / noise reduction processing on the cigarette perforation data with large offsets, and divides the smoothed / noise-reduced cigarette perforation multidimensional data into cigarette perforation test data and cigarette perforation training data; finally, a cigarette filter rod ventilation rate prediction module predicts the cigarette perforation rate... The perforation test data is input into the trained cigarette filter rod ventilation rate prediction model, which outputs the cigarette filter rod ventilation rate corresponding to the perforation of the cigarette. In this way, not only can the morphological parameters of the two rows of small holes formed in the cigarette filter rod during the laser perforation process be linked to the cigarette filter rod ventilation rate, but the fluctuation of the hole morphology can also reflect the fluctuation of the cigarette filter rod ventilation rate. This supports the prediction and control of the cigarette ventilation rate during the cigarette production process, so as to solve the problem of difficulty in predicting the cigarette filter rod ventilation rate caused by the random fluctuation of the hole morphology and the nonlinear factors between the cigarette filter rod ventilation rate and the hole morphology during the laser perforation process.
[0157] Based on the above embodiments, optionally, the S210 and the cigarette punching multi-dimensional data acquisition module are specifically used for:
[0158] The number of holes obtained from the multidimensional data of cigarette perforation on each cigarette is determined; wherein, the number of holes obtained from the multidimensional data of cigarette perforation on each cigarette is 30; each cigarette is a cigarette of the same brand after being perforated by a laser perforation device;
[0159] Based on the number of holes obtained from the multidimensional data of perforation on each cigarette, the hole morphology data and ventilation rate of each hole are obtained; the hole morphology data includes hole length, hole width, hole depth, hole spacing, hole upper surface area, and hole inner surface area; the hole morphology data and ventilation rate of each hole are obtained by measurement.
[0160] By integrating the hole shape data and ventilation rate corresponding to each hole, multi-dimensional data of cigarette perforation is obtained; among which, multi-dimensional data of cigarette perforation is multi-dimensional data of perforation for each cigarette.
[0161] Based on the above embodiments, optionally, the S220 data smoothing / noise reduction processing module is specifically used for:
[0162] Zero-value data in the multidimensional data of cigarette perforation is removed to obtain the multidimensional data of cigarette perforation after removing zero values; wherein, the zero-value data in the multidimensional data of cigarette perforation is the zero-value data of hole morphology data and the zero-value data of ventilation rate corresponding to the hole are removed.
[0163] The excessively deviated cigarette perforation data in the multidimensional data of cigarette perforation after removing zero values is identified by the 3σ criterion. The excessively deviated cigarette perforation data is smoothed by SG filtering based on local polynomial least squares fitting to obtain the smoothed multidimensional data of cigarette perforation.
[0164] The noise reduction process of the smoothed cigarette punch multidimensional data was performed by the sliding window moving average method to obtain the noise-reduced cigarette punch multidimensional data.
[0165] The noise-reduced multidimensional data of cigarette punching were divided into cigarette punching test data and cigarette punching training data in an 8:2 ratio.
[0166] Based on the above embodiments, optionally, the cigarette filter rod ventilation rate prediction model includes a data local feature enhancement module, a feature filtering module, and a model prediction module;
[0167] Based on training data of cigarette perforation and optimization strategies of early stop mechanism, a cigarette filter rod ventilation rate prediction model is trained to obtain the trained cigarette filter rod ventilation rate prediction model, including the following steps:
[0168] In the local feature enhancement module of the cigarette punching training data input data, the local feature enhancement module performs local feature enhancement processing on the cigarette punching training data to obtain the cigarette punching features after local feature enhancement.
[0169] The enhanced local feature features of the cigarette punch are input into the feature filtering module. The feature filtering module performs feature selection processing on the enhanced local feature features of the cigarette punch to obtain the feature-selected cigarette punch features.
[0170] The perforation features of the cigarette after feature selection are input into the model prediction module, and combined with the early stop mechanism optimization strategy, the model prediction module is trained to obtain the trained cigarette filter rod ventilation rate prediction model.
[0171] Based on the above embodiments, optionally, the local feature enhancement module is specifically used for:
[0172] Based on training data of cigarette punching, the sliding mean and sliding standard deviation were calculated with window lengths of 5, 10 and 20, and feature enhancement was performed on the mean of the morphology of the 5 holes in the cigarette filter rod.
[0173] After calculating the moving mean and moving standard deviation, edge padding is used to ensure that the feature length is consistent with the original sample; in this case, 6 features are added to each morphology, for a total of 35 features.
[0174] Based on the above embodiments, optionally, the feature filtering module is specifically used to: calculate the mutual information between each enhanced local feature cigarette punch feature and a preset label for all enhanced local feature cigarette punch features; wherein, all enhanced local feature cigarette punch features are 35 features;
[0175] Based on the mutual information between the enhanced local feature of the cigarette and the preset label, and combined with the Top-K strategy, the top 10% of the enhanced local feature features of the cigarette are selected.
[0176] Based on the above embodiments, optionally, the early stopping mechanism optimization strategy is as follows:
[0177] When the error of the cigarette filter ventilation rate prediction model does not decrease for 20 consecutive rounds, the training of the cigarette filter ventilation rate prediction model will be automatically stopped, and the optimal number of iterations will be used to determine the trained cigarette filter ventilation rate prediction model.
[0178] The cigarette filter rod ventilation rate prediction device based on local feature enhancement and XGBoost provided in the embodiments of the present invention can execute the cigarette filter rod ventilation rate prediction method based on local feature enhancement and XGBoost provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.
[0179] Example 3
[0180] Figure 6 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.
[0181] like Figure 6 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.
[0182] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0183] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the cigarette filter ventilation rate prediction method based on local feature enhancement and XGBoost.
[0184] In some embodiments, the cigarette filter ventilation rate prediction method based on local feature enhancement and XGBoost can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the cigarette filter ventilation rate prediction method based on local feature enhancement and XGBoost described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the cigarette filter ventilation rate prediction method based on local feature enhancement and XGBoost by any other suitable means (e.g., by means of firmware).
[0185] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.
[0186] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0187] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.
[0188] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).
[0189] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.
[0190] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.
[0191] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.
[0192] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.
Claims
1. A method for predicting the ventilation rate of cigarette filter rods based on local feature enhancement and XGBoost, characterized in that: Includes the following steps: The hole morphology data of each cigarette filter rod after drilling is obtained, and the hole morphology data is integrated with the corresponding ventilation rate data of the cigarette filter rod to obtain multi-dimensional data of cigarette perforation; among which, the hole morphology data includes the number of holes, hole length, hole width, hole depth, hole spacing, and the surface area of the upper and inner surfaces of the holes; each cigarette is a cigarette after laser perforation; After removing zeros, identify cigarette punch data with large offsets in the multidimensional data of cigarette punching. Perform smoothing / denoising processing on the cigarette punch data with large offsets. Divide the smoothed / denoised multidimensional data of cigarette punching into cigarette punching test data and cigarette punching training data. The smoothing / denoising processing is SG filtering and sliding window moving average method. Based on the training data of cigarette punching and the optimization strategy of early stop mechanism, the cigarette filter rod ventilation rate prediction model is trained to obtain the trained cigarette filter rod ventilation rate prediction model; the cigarette filter rod ventilation rate prediction model is constructed based on the XGBoost algorithm; the cigarette filter rod ventilation rate prediction model includes a data local feature enhancement module, a feature selection module and a model prediction module. Input the cigarette perforation test data into the trained cigarette filter rod ventilation rate prediction model, and output the cigarette filter rod ventilation rate corresponding to the cigarette perforation.
2. The method for predicting the ventilation rate of cigarette filter rods based on local feature enhancement and XGBoost according to claim 1, characterized in that: The process of acquiring the hole morphology data of each cigarette filter rod after perforation, and integrating the hole morphology data with the corresponding ventilation rate data of the cigarette filter rod to obtain multi-dimensional data on cigarette perforation, includes the following steps: The number of holes obtained from the multidimensional data of cigarette perforation on each cigarette is determined; wherein, the number of holes obtained from the multidimensional data of cigarette perforation on each cigarette is 30; each cigarette is a cigarette of the same brand after being perforated by a laser perforation device; Based on the number of holes obtained from the multidimensional data of perforation on each cigarette, the hole morphology data and ventilation rate of each hole are obtained; the hole morphology data includes hole length, hole width, hole depth, hole spacing, hole upper surface area, and hole inner surface area; the hole morphology data and ventilation rate of each hole are obtained by measurement. By integrating the hole shape data and ventilation rate corresponding to each hole, multi-dimensional data of cigarette perforation is obtained; among which, multi-dimensional data of cigarette perforation is multi-dimensional data of perforation for each cigarette.
3. The method for predicting the ventilation rate of cigarette filter rods based on local feature enhancement and XGBoost according to claim 1, characterized in that: The process involves identifying cigarette punching data with significant offsets in the multidimensional data after zero-removal processing, smoothing / denoising these data, and then dividing the smoothed / denoised multidimensional data into cigarette punching test data and cigarette punching training data. This includes the following steps: Zero-value data in the multidimensional data of cigarette punching is removed to obtain the multidimensional data of cigarette punching after removing zero values; where the zero-value data in the multidimensional data of cigarette punching is the zero-value data of hole morphology data and the zero-value data of ventilation rate corresponding to the hole are removed. The excessively deviated cigarette perforation data in the multidimensional data of cigarette perforation after removing zero values is identified by the 3σ criterion. The excessively deviated cigarette perforation data is smoothed by SG filtering based on local polynomial least squares fitting to obtain the smoothed multidimensional data of cigarette perforation. The noise reduction process of the smoothed cigarette punch multidimensional data was performed by the sliding window moving average method to obtain the noise-reduced cigarette punch multidimensional data. The noise-reduced multidimensional data of cigarette punching were divided into cigarette punching test data and cigarette punching training data in an 8:2 ratio.
4. The method for predicting the ventilation rate of cigarette filter rods based on local feature enhancement and XGBoost according to claim 1, characterized in that: The cigarette filter rod ventilation rate prediction model includes a data local feature enhancement module, a feature filtering module, and a model prediction module. Based on training data of cigarette perforation and optimization strategies of early stop mechanism, a cigarette filter rod ventilation rate prediction model is trained to obtain the trained cigarette filter rod ventilation rate prediction model, including the following steps: In the local feature enhancement module for the input data of cigarette punching training data, the local feature enhancement module performs local feature enhancement processing on the cigarette punching training data to obtain the cigarette punching features after local feature enhancement. The enhanced local feature features of the cigarette punch are input into the feature filtering module. The feature filtering module performs feature selection processing on the enhanced local feature features of the cigarette punch to obtain the feature-selected cigarette punch features. The perforation features of the cigarette after feature selection are input into the model prediction module, and combined with the early stop mechanism optimization strategy, the model prediction module is trained to obtain the trained cigarette filter rod ventilation rate prediction model.
5. The method for predicting the ventilation rate of cigarette filter rods based on local feature enhancement and XGBoost according to claim 4, characterized in that: The local feature enhancement module is specifically used for: Based on training data of cigarette punching, the sliding mean and sliding standard deviation were calculated with window lengths of 5, 10 and 20, and feature enhancement was performed on the mean of the morphology of the 5 holes in the cigarette filter rod. After calculating the moving mean and moving standard deviation, edge padding is used to ensure that the feature length is consistent with the original sample; in this case, 6 features are added to each morphology, for a total of 35 features.
6. The method for predicting the ventilation rate of cigarette filter rods based on local feature enhancement and XGBoost according to claim 4, characterized in that: The feature filtering module is specifically used for: For all the enhanced local features of the cigarette punch, calculate the mutual information between each enhanced local feature of the cigarette punch and the preset label; there are 35 enhanced local feature features of the cigarette punch. Based on the mutual information between the enhanced local feature of the cigarette and the preset label, and combined with the Top-K strategy, the top 10% of the enhanced local feature features of the cigarette are selected.
7. The method for predicting the ventilation rate of cigarette filter rods based on local feature enhancement and XGBoost according to claim 4, characterized in that: The optimization strategy for the early stopping mechanism is as follows: When the error of the cigarette filter ventilation rate prediction model does not decrease for 20 consecutive rounds, the training of the cigarette filter ventilation rate prediction model will be automatically stopped, and the optimal number of iterations will be used to determine the trained cigarette filter ventilation rate prediction model.
8. A cigarette filter rod ventilation rate prediction device based on local feature enhancement and XGBoost, characterized in that: include: The cigarette perforation multidimensional data acquisition module is used to acquire the hole morphology data of each cigarette filter rod after perforation, and integrate the hole morphology data with the corresponding ventilation rate data of the cigarette filter rod to obtain the cigarette perforation multidimensional data; wherein, the hole morphology data includes the number of holes, hole length, hole width, hole depth, hole spacing, hole upper surface and hole inner surface area; each cigarette is a cigarette after laser perforation; The data smoothing / denoising module is used to identify cigarette punch data with large offsets in the multidimensional data of cigarette punching after zero removal, and to perform smoothing / denoising on the cigarette punch data with large offsets. The smoothed / denoising multidimensional data of cigarette punching is divided into cigarette punching test data and cigarette punching training data. The smoothing / denoising process is SG filtering and sliding window moving average method. The cigarette filter ventilation rate prediction model module is used to train the cigarette filter ventilation rate prediction model based on cigarette punching training data and early stop mechanism optimization strategy, so as to obtain the trained cigarette filter ventilation rate prediction model; wherein, the cigarette filter ventilation rate prediction model is constructed based on the XGBoost algorithm; the cigarette filter ventilation rate prediction model includes a data local feature enhancement module, a feature selection module and a model prediction module. The cigarette filter rod ventilation rate prediction module is used to input the cigarette punching test data into the trained cigarette filter rod ventilation rate prediction model and output the cigarette filter rod ventilation rate corresponding to the cigarette punching.
9. An electronic device, characterized in that, The electronic device includes: One or more processors; A memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement a cigarette filter ventilation rate prediction method based on local feature enhancement and XGBoost as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute and implement the cigarette filter ventilation rate prediction method based on local feature enhancement and XGBoost as described in any one of claims 1-7.