Principal Component Analysis Dimensionality Reduction Research Articles

Prediction of Carbon Emission Right Price Based on XGBoost Algorithm

Reasonable carbon market price prediction can facilitate the carbon market participants, such as physical producers, to achieve the goal of efficient emission reduction through the market mechanism. In this paper, we use XGBoost, an integrated algorithm in machine learning, to forecast the domestic carbon price from 2013 to 2021. Pearson coefficient is utilized to calculate the correlation of data features, and perform PCA dimensionality reduction on the features with high correlation coefficients. Before using PCA to reduce the dimensions, in order to make the feature data more suitable for later model training, the feature data is standardized. After the standardization, PCA selects super parameters with maximum likelihood estimation and outputs the dimension reduction results of features. Finally, the integrated algorithm XGBoost is used to form a prediction model for the carbon price. RMSE output from cross validation are used to evaluate the accuracy and error of the prediction results. The results show that this paper confirms that the integrated algorithm XGBoost model has a good prediction ability for carbon price, and provides a novel idea for the field of carbon emission price prediction. It is expected to provide some rational basis for the carbon market participants to make investment decisions, so as to avoid the carbon market risk caused by the change of carbon price.

Generator Fault Classification Method Based on Multi-Source Information Fusion Naive Bayes Classification Algorithm

The existing motor fault classification methods mostly use sensors to detect a single fault feature, which makes it difficult to ensure high diagnostic accuracy. In this paper, a motor fault classification method based on multi-source information fusion Naive Bayes classification algorithm is proposed. Firstly, this paper introduces the concept and advantages of multi-source information fusion, as well as its problems of miscellaneous information and inconsistent data magnitude. For example, as this paper classifies the fault of generators, there are many physical quantities, such as voltage, current and temperature, which are not in the same dimension, therefore it is difficult to fuse. Then, aiming at the corresponding problems, this paper uses a PCA dimension reduction method to remove redundant information and reduce the dimension of multi-dimensional complex information. Aiming at the problem of unequal data magnitude, the interval mapping method is adopted to effectively solve the misjudgment caused by unequal data magnitude. After the initial multi-source information processing, the classical Naive Bayes classification algorithm is used for fault classification, and the algorithm diagnosis and verification are carried out according to the statistical fault data. Use of the algorithm increases accuracy to more than 97%.

A vegetation classification method based on improved dual-way branch feature fusion U-net.

Aiming at the problems of complex structure parameters and low feature extraction ability of U-Net used in vegetation classification, a deep network with improved U-Net and dual-way branch input is proposed. Firstly, The principal component analysis (PCA) is used to reduce the dimension of hyperspectral remote sensing images, and the effective bands are obtained. Secondly, the depthwise separable convolution and residual connections are combined to replace the common convolution layers of U-Net for depth feature extraction to ensure classification accuracy and reduce the complexity of network parameters. Finally, normalized difference vegetation index (NDVI), gray level co-occurrence matrix (GLCM) and edge features of hyperspectral remote sensing images are extracted respectively. The above three artificial features are fused as one input, and PCA dimension reduction features are used as another input. Based on the improved U-net, a dual-way vegetation classification model is generated. Taking the hyperspectral remote sensing image of Matiwan Village, Xiong'an, Beijing as the experimental object, the experimental results show that the precision and recall of the improved U-Net are significantly improved with the residual structure and depthwise separable convolution, reaching 97.13% and 92.36% respectively. In addition, in order to verify the effectiveness of artificial features and dual-way branch design, the accuracy of single channel and the dual-way branch are compared. The experimental results show that artificial features in single channel network interfere with the original hyperspectral data, resulting in reduction of the recognition accuracy. However, the accuracy of the dual-way branch network has been improved, reaching 98.67%. It shows that artificial features are effective complements of network features.

Technical methods of national security supervision: Grain storage security as an example

Grain security guarantees national security. China has many widely distributed grain depots to supervise grain storage security. However, this has led to a lack of regulatory capacity and manpower. Amid the development of reserve-level information technology, big data supervision of grain storage security should be improved. This study proposes big data research architecture and an analysis model for grain storage security; as an example, it illustrates the supervision of the grain loss problem in storage security. The statistical analysis model and the prediction and clustering-based model for grain loss supervision were used to mine abnormal data. A combination of feature extraction and feature selection reduction methods were chosen for dimensionality. A comparative analysis showed that the nonlinear prediction model performed better on the grain loss data set, with R2 of 87.21%, 87.83%, 91.97%, and 89.40% for Gradient Boosting Regressor (GBR), Random Forest, Decision Tree, XGBoost regression on test sets, respectively. Nineteen abnormal data were filtered out by GBR combined with residuals as an example. The deep learning model had the best performance on the mean absolute error, with an R2 of 85.14% on the test set and only one abnormal data identified. This is contrary to the original intention of finding as many anomalies as possible for supervisory purposes. Five classes were generated using principal component analysis dimensionality reduction combined with Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering, with 11 anomalous data points screened by adding the amount of normalized grain loss. Based on the existing grain information system, this paper provides a supervision model for grain storage that can help mine abnormal data. Unlike the current post-event supervision model, this study proposes a pre-event supervision model. This study provides a framework of ideas for subsequent scholarly research; the addition of big data technology will help improve efficient supervisory capacity in the field of grain supervision.

Detection and recognition of toluene and butanone in mixture based on SnO2 sensor via dynamic transient and steady-state response analysis in jump heating voltage mode

The temperature modulation method is used to improve the selectivity of sensor, but there are still some deficiencies in the gas mixture recognition. The dynamic response law is insufficient understood because the dynamic response process itself is a complex change and there are some factors such as competition or suppression among multi-gases. This paper solves this problem by decomposing dynamic response and mapping static response. Based on the grain boundary barrier model, dynamic response is divided into dynamic temperature response and dynamic gas-sensing response. The dynamic temperature response is unaffected by the target gas and can be detected directly. Dynamic gas-sensing response cannot be detected directly but can be mapped through static responses. Therefore, this paper uses static response to grasp the dynamic response law to directionally optimize the experimental scheme. Dynamic transient response and dynamic steady-state response are analyzed one by one to discuss the role of dynamic response law in gas mixture recognition. High recognition rate is achieved in KNN recognition due to the selection of PCA dimensionality reduction results of the second dynamic transient response and dynamic steady-state response value. The study of the dynamic response law provides a new avenue for the recognition of gas mixture.

Comparison of Different Transfer Learning Methods for Classification of Mangrove Communities Using MCCUNet and UAV Multispectral Images

Mangrove-forest classification by using deep learning algorithms has attracted increasing attention but remains challenging. The current studies on the transfer classification of mangrove communities between different regions and different sensors are especially still unclear. To fill the research gap, this study developed a new deep-learning algorithm (encoder–decoder with mixed depth-wise convolution and cascade upsampling, MCCUNet) by modifying the encoder and decoder sections of the DeepLabV3+ algorithm and presented three transfer-learning strategies, namely frozen transfer learning (F-TL), fine-tuned transfer learning (Ft-TL), and sensor-and-phase transfer learning (SaP-TL), to classify mangrove communities by using the MCCUNet algorithm and high-resolution UAV multispectral images. This study combined the deep-learning algorithms with recursive feature elimination and principal component analysis (RFE–PCA), using a high-dimensional dataset to map and classify mangrove communities, and evaluated their classification performance. The results of this study showed the following: (1) The MCCUNet algorithm outperformed the original DeepLabV3+ algorithm for classifying mangrove communities, achieving the highest overall classification accuracy (OA), i.e., 97.24%, in all scenarios. (2) The RFE–PCA dimension reduction improved the classification performance of deep-learning algorithms. The OA of mangrove species from using the MCCUNet algorithm was improved by 7.27% after adding dimension-reduced texture features and vegetation indices. (3) The Ft-TL strategy enabled the algorithm to achieve better classification accuracy and stability than the F-TL strategy. The highest improvement in the F1–score of Spartina alterniflora was 19.56%, using the MCCUNet algorithm with the Ft-TL strategy. (4) The SaP-TL strategy produced better transfer-learning classifications of mangrove communities between images of different phases and sensors. The highest improvement in the F1–score of Aegiceras corniculatum was 19.85%, using the MCCUNet algorithm with the SaP-TL strategy. (5) All three transfer-learning strategies achieved high accuracy in classifying mangrove communities, with the mean F1–score of 84.37~95.25%.

Classification of Different Shoulder Girdle Motions for Prosthesis Control Using a Time-Domain Feature Extraction Technique

Abstract—The upper limb amputation exerts a significant burden on the amputee, limiting their ability to perform everyday activities, and degrading their quality of life. Amputee patients’ quality of life can be improved if they have natural control over their prosthetic hands. Among the biological signals, most commonly used to predict upper limb motor intentions, surface electromyography (sEMG), and axial acceleration sensor signals are essential components of shoulder-level upper limb prosthetic hand control systems. In this work, a pattern recognition system is proposed to create a plan for categorizing high-level upper limb prostheses in seven various types of shoulder girdle motions. Thus, combining seven feature groups, which are root mean square, four-order autoregressive, wavelength, slope sign change, zero crossing (ZC), mean absolute value, and cardinality. In this article, the time-domain features were first extracted from the EMG and acceleration signals. Then, the spectral regression (SR) and principal component analysis dimensionality reduction methods are employed to identify the most salient features, which are then passed to the linear discriminant analysis (LDA) classifier. EMG and axial acceleration signal datasets from six intact-limbed and four amputee participants exhibited an average classification error of 15.68 % based on SR dimensionality reduction using the LDA classifier.

Construction of a diagnostic classifier for cervical intraepithelial neoplasia and cervical cancer based on XGBoost feature selection and random forest model.

The pathological phenotype of early-stage cervical cancer (CC) is similar to that of cervical intraepithelial neoplasia (CIN), which provides a challenge for the diagnosis of cervical precancerous lesions. Meanwhile, the existing diagnostic methods have certain subjectivity and limitations, resulting in the possibility of misdiagnosis or missed diagnosis. Hence, some methods are needed to assist diagnosis of CC and CIN. Based on the data of CIN and CC in gene expression omnibus (GEO) dataset, the eXtreme Gradient Boosting (XGBoost) algorithm was used to screen the feature genes between CIN and CC for constructing the classifier. Incremental feature selection (IFS) curve was also used for screening. The classifier was validated for reliability using principal component analysis (PCA) dimensionality reduction analysis and heat map analysis of gene expression. Then, differentially expressed genes of CIN and CC were intersected with the classifier genes. Genes in the intersection were used as seeds for protein-protein interaction network construction and restart random walk analysis. And the genes with the top 50 affinity coefficients were selected for gene ontology (GO) and kyoto encyclopedia of genes and genome (KEGG) enrichment analyses to observe the biological functions with differences between CIN and CC. The peripheral blood genes of CIN and CC were analyzed, and seven genes were screened. Using this gene for classifier construction, IFS curve screening revealed that the three-feature gene classifier constructed according to the random forest model had the best effect. The results of PCA dimensionality reduction analysis and gene expression heat map analysis showed that the three-gene classifier could effectively distinguish CIN from CC. A three-gene diagnostic classifier can effectively distinguish CIN patients from CC patients and provide a reference for the clinical diagnosis of early CC.

Research on intelligent assembly method of aero-engine multi-stage rotors based on SVM and variable-step AFSA-BP neural network

The quality of the aero-engine rotors assembly determines the overall performance of the engine. Aiming at the problems of rotors assembly with different plane types, we proposes a rotor plane classification method based on SVM by using the profile data of PCA dimension reduction. Meanwhile, for the unilateral-tilt plane rotors, the three-objective rotors assembly method of coaxiality, unbalance amount and perpendicularity based on the rigid rotor model is established. For the hyperbolic paraboloid rotors, an intelligent assembly method based on AFSA-BP neural network for coaxiality, unbalance amount and perpendicularity is established. The experiment is based on the double-column ultra-precision measuring instrument and V4L vertical balancing machine and HL5UB horizontal balancing machine to measure rotors geometry and unbalance data. The experimental results show that the plane type classification accuracy can reach 99 %. The prediction error of the coaxiality of the unilateral-tilt plane rotors assembly is 5.1 μm, the prediction error of the unbalance amount is 196 g·mm, and the prediction error of the perpendicularity is 0.6 μm. The average prediction error of the coaxiality of the hyperbolic paraboloid rotors assembly is 0.9 μm, and the average prediction error of the unbalance amount is 73 g·mm, and the average prediction error of the perpendicularity is 0.2 μm. Our method provides a reliable assembly solution for aero-engine rotors assembly and meets actual assembly requirements.

Study on Health Indicator Construction and Health Status Evaluation of Hydraulic Pumps Based on LSTM–VAE

This paper addresses the difficulty of evaluating operating status in widely used gear pumps. A method for constructing hydraulic pump health indicators and evaluating health status is proposed based on LSTM–VAE. In this study, the vibration signal data source of gear pumps was assessed in the accelerated life test. Firstly, the normalized feature vectors of the whole-life operation data of gear pumps were extracted by wavelet packet decomposition and amplitude feature extraction. Combining an LSTM algorithm with a VAE algorithm, a method for constructing hydraulic pump health indicators based on LSTM–VAE is proposed. By learning the feature vectors of gear pumps in varying health conditions, a one-dimensional HI curve of the gear pumps was obtained. Then, LSTM was used to predict the HI curve of gear pumps. According to the volume efficiency of the gear pumps, the health status of gear pumps is divided into four states: health, sub-health, deterioration, and failure. The health status of the hydraulic pump is accurately evaluated by the health indicator. Finally, the proposed method is compared with the traditional method based on feature selection and PCA dimensionality reduction. The health indicator constructed by the method proposed in this paper is superior to the traditional method in terms of tendency, robustness, and monotonicity, which proves the effectiveness of the method proposed in this paper.

Hyperspectral Estimates of Soil Moisture Content Incorporating Harmonic Indicators and Machine Learning

Soil is one of the most significant natural resources in the world, and its health is closely related to food security, ecological security, and water security. It is the basic task of soil environmental quality assessment to monitor the temporal and spatial variation of soil properties scientifically and reasonably. Soil moisture content (SMC) is an important soil property, which plays an important role in agricultural practice, hydrological process, and ecological balance. In this paper, a hyperspectral SMC estimation method for mixed soil types was proposed combining some spectral processing technologies and principal component analysis (PCA). The original spectra were processed by wavelet packet transform (WPT), first-order differential (FOD), and harmonic decomposition (HD) successively, and then PCA dimensionality reduction was used to obtain two groups of characteristic variables: WPT-FOD-PCA (WFP) and WPT-FOD-HD-PCA (WFHP). On this basis, three regression models of principal component regression (PCR), partial least squares regression (PLSR), and back propagation (BP) neural network were applied to compare the SMC predictive ability of different parameters. Meanwhile, we also compared the results with the estimates of conventional spectral indices. The results indicate that the estimation results based on spectral indices have significant errors. Moreover, the BP models (WFP-BP and WFHP-BP) show more accurate results when the same variables are selected. For the same regression model, the choice of variables is more important. The three models based on WFHP (WFHP-PCR, WFHP-PLSR, and WFHP-BP) all show high accuracy and maintain good consistency in the prediction of high and low SMC values. The optimal model was determined to be WFHP-BP with an R2 of 0.932 and a prediction error below 2%. This study can provide information on farm entropy before planting crops on arable land as well as a technical reference for estimating SMC from hyperspectral images (satellite and UAV, etc.).

Modern enterprise financial accounting abnormal statistical data in the biopharma industry

At present, abnormal data monitoring in biopharma industry is mainly realized through the difference between single characteristic attributes of data, which leads to low effective monitoring rate and high time cost. In order to improve the monitoring effect of modern enterprise financial accounting abnormal statistical data in biopharma industry, this paper studies a modern enterprise financial accounting abnormal statistical monitoring method based on data mining algorithm. After the preliminary processing of statistical data, the combination of information entropy and PCA dimension reduction is introduced to reduce the dimension of data. After the decision tree is established by C5.0 decision tree algorithm, the random forest is constructed to realize the abnormal monitoring of statistical data. The experimental results show that the monitoring rate of this method is more than 97%, the false alarm rate and time cost are low, and has good performance.

Concrete Strength Prediction Using Different Machine Learning Processes: Effect of Slag, Fly Ash and Superplasticizer

Blast furnace slag (BFS) and fly ash (FA), as mining-associated solid wastes with good pozzolanic effects, can be combined with superplasticizer to prepare concrete with less cement utilization. Considering the important influence of strength on concrete design, random forest (RF) and particle swarm optimization (PSO) methods were combined to construct a prediction model and carry out hyper-parameter tuning in this study. Principal component analysis (PCA) was used to reduce the dimension of input features. The correlation coefficient (R), the explanatory variance score (EVS), the mean absolute error (MAE) and the mean square error (MSE) were used to evaluate the performance of the model. R = 0.954, EVS = 0.901, MAE = 3.746, and MSE = 27.535 of the optimal RF-PSO model on the testing set indicated the high generalization ability. After PCA dimensionality reduction, the R value decreased from 0.954 to 0.88, which was not necessary for the current dataset. Sensitivity analysis showed that cement was the most important feature, followed by water, superplasticizer, fine aggregate, BFS, coarse aggregate and FA, which was beneficial to the design of concrete schemes in practical projects. The method proposed in this study for estimation of the compressive strength of BFS-FA-superplasticizer concrete fills the research gap and has potential engineering application value.

Measurement and Analysis of Regional Inclusive Finance Composite Index Based on Contribution Analysis

In recent years, with the development of economic globalization, regional inclusive finance has received increasing attention from the international community. The development concept of inclusive finance has been deeply rooted in the people's hearts, and its practice has been continuously improved. Based on the data of financial services in Dalian, Liaoning Province, this paper uses principal component analysis ( PCA ) and artificial neural network-radial basis function ( ANN-RBF ) algorithm to measure and analyze the contribution of the inclusive financial index ( FII ). Firstly, the index system of regional inclusive finance is confirmed, and based on PCA dimension reduction, the index of regional inclusive finance is constructed. By constructing the evaluation model based on principal component analysis and artificial neural network-radial basis function contribution analysis, the measurement and analysis conclusions of regional inclusive finance comprehensive index is obtained. The results show that: ( 1 ) In recent years, FII in Dalian has gradually fluctuated and increased. (2)Among the influencing factors, the number of insurance institutions, number of employees in monetary and financial service, the amount of RMB loans of financial institutions in Dalian, the premium income of property insurance in Dalian, the proportion of total loans of financial institutions in Dalian to GDP, Dalian property insurance depth have the greatest impact.

Intelligent diagnosis algorithm for thyroid nodules based on deep learning and statistical features

Thyroid nodule is one of the most common endocrine diseases in the adult population. Early screening and diagnosis of thyroid nodules is of great significance to patients’ subsequent treatment. An objective and accurate diagnosis algorithm for thyroid nodules is vital to improve the efficiency of clinical diagnosis and reduce the work pressure of doctors. In this paper, we propose a fused deep learning model for the diagnosis of benign and malignant thyroid nodules. Based on the ACR TI-RADS proposed by the American Academy of Radiology, the model first extracts 33 clinically significant statistical features of composition, echogenicity, shape, margin, and echogenic foci. After PCA dimensionality reduction, the top 4 features are fused with the feature map of EfficientNet B3. The thyroid nodules are classified into benign and malignant based on the fused feature map. This model is trained and tested on 3828 clinically collected ultrasound images of thyroid nodules. It achieves an accuracy of 96.4% and an AUC of 0.965. Compared with traditional machine learning algorithms and state-of-the-art deep learning networks, the fused model we proposed has improved classification performance and amount of parameters, and is also clinically interpretable.

Implementation of Principal Component Analysis and Learning Vector Quantization for Classification of Food Nutrition Status

Balanced nutrition is very good in the process of child development. During the COVID-19 pandemic, consuming a balanced nutritious diet can keep a child's immune system from transmitting the virus. In determining the nutritional content of children's food during the pandemic, a classification of the nutritional content of children's food is carried out by applying the principal component analysis (PCA) dimension reduction method and the learning vector quantization (LVQ) classification method. The data used in this study amounted to 1168 data with 25 indicators of food nutrients. From the tests that have been carried out, the combination of the PCA-LVQ method produces an average accuracy of 58% with the highest accuracy of 60%. In addition, this study also compares the performance of the PCA dimension reduction method, independent component analysis (ICA) and factor analysis (FA) on the LVQ classification process. The final result of testing the three methods is that the FA method takes the fastest time, which is 4.10434 seconds and the PCA method produces the highest accuracy, which is 58.2%

Terahertz Non-Destructive Testing and Imaging of High-Voltage Cables

The evaluation of the internal defects in high-voltage cross-linked polyethylene cables is highly crucial for the security and reliability of power transmission. Here, a terahertz (THz) frequency-modulated-continuous-wave non-destructive testing (NDT) imaging system is used to demonstrate the non-contact detection of a high-voltage cable (35 KV). Combined with linear scanning and axial rotation, the three-dimensional (3D) data of the columnar target is acquired, and a 3D perspective image is reconstructed. The correspondence between the target in the real space and image space is determined to accurately show the internal spiral structure and strand number of metal wires. The results indicate that prefabricated air and water holes in the polyethylene insulation layer can be displayed. Furthermore, feature defect signals are automatically classified and recognised by combining the principal component analysis dimension reduction method and the support vector machine classification method, which can predict abnormal defects more efficiently in mass data. These results provide technical guidance for the non-contact NDT and visual evaluation of the internal state of detected cable targets.

Construction of a prognostic risk assessment model for lung adenocarcinoma based on Integrin β family-related genes.

BackgroundIntegrin β (ITGB) superfamily plays an essential role in the intercellular connection and signal transmission. It was exhibited that overexpressing of ITGB family members promotes the malignant progression of lung adenocarcinoma (LUAD), but the relationship between ITGB superfamily and the LUAD prognosis remains unclear.MethodsIn this study, the samples were assigned to different subgroups utilizing non‐negative matrix factorization clustering according to the expression of ITGB family members in LUAD. Kaplan–Meier (K‐M) survival analysis revealed the significant differences in the prognosis between different ITGB subgroups. Subsequently, we screened differentially expressed genes among different subgroups and conducted univariate Cox analysis, random forest feature selection, and multivariate Cox analysis. 9‐feature genes (FAM83A, AKAP12, PKP2, CYP17A1, GJB3, TMPRSS11F, KRT81, MARCH4, and STC1) in the ITGB superfamily were selected to establish a prognostic assessment model for LAUD.ResultsIn accordance with the median risk score, LUAD samples were divided into high‐ and low‐risk groups. The receiver operating characteristic (ROC) curve of LUAD patients’ survival was predicted via K‐M survival curve and principal component analysis dimensionality reduction. This model was found to have a favorable performance in LUAD prognostic assessment. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses of differentially expressed genes between groups and Gene Set Enrichment Analysis (GSEA) of intergroup samples confirmed that the high‐ and low‐risk groups had evident differences mainly in the function of extracellular matrix (ECM) interaction. Risk score and univariate and multivariate Cox regression analyses of clinical factors showed that the prognostic model could be applied as an independent prognostic factor for LUAD. Then, we draw the nomogram of 1‐, 3‐, and 5‐year survival of LUAD patients predicted with the risk score and clinical factors. Calibration curve and clinical decision curve proved the favorable predictive ability of nomogram.ConclusionWe constructed a LUAD prognostic risk model based on the ITGB superfamily, which can provide guidance for clinicians on their prognostic judgment.

Visible-NIR spectral characteristics and grade inversion model of skarn-type iron ore

Iron ore grade determination is a key step in mining production, and its efficiency has a significant impact on mining efficiency and profitability. At present, the most commonly used method for ore grade determination is chemical analysis, which not only is expensive but also has a long assay cycle and lag effect relative to the ore allocation process and cannot effectively reduce the loss depletion rate of ore mining. Therefore, in situ iron ore grade data determination technology based on visible and near-infrared hyperspectral data analysis is an effective way to solve the above problems, but the poor quality of raw hyperspectral data and low accuracy of the ore grade inversion model are still one of the bottlenecks in the application of this technology. In this paper, 225 ore samples were collected from Hongling skarn-type iron ore in Chifeng, Inner Mongolia Autonomous Region, China. The visible and near-infrared hyperspectral data of the sample set and the grade data obtained from chemical analysis were used as data sources. The raw spectral data were first smoothed, and then the smoothed spectral data were processed using two algorithms, continuum removal (CR) and standard normal variate transform (SNV). Then, two dimensionality reduction algorithms, principal component analysis (PCA) and genetic algorithm (GA), were used to reduce the dimensionality of the spectral data before and after preprocessing, and the data sources were obtained after the four different preprocessing combination algorithms were processed. Finally, several quantitative inversion models were established based on the random forest algorithm and the backpropagation neural network (BPNN) algorithm. The stability, accuracy and credibility of the models were evaluated by four metrics: the coefficient of determination (R2), root mean square error (RMSE), residual prediction deviation(RPD) and mean relative error (MRE). The results showed that the quantitative inversion model (SNV-PCA-BPNN) established based on the BP neural network algorithm using the data processed by SNV and PCA for dimensionality reduction was the best, with an R2 of 0.99, an RMSE of 0.0048, an MRE of 1.10% and a RPD of 37.70, providing an effective method for the high-precision inversion of iron ore grades.

Petrophysical rock typing based on deep learning network and hierarchical clustering for volcanic reservoirs

With the rebound of international oil prices, complex non-clastic volcanic rock reservoirs have been gaining much attention from the scientific community in recent years. However, due to the complexity of volcanic diagenesis and the intensity of tectonic activity at its development location, volcanic reservoirs generally exhibit strong heterogeneity. In particular, the volcanic rock reservoir in Block Jinlong 2 is a low-medium porosity, ultra-low to low permeability reservoir with complex lithology and diverse storage spaces. Therefore, the author proposes a petrophysical rock typing method based on the VGG16 deep learning network and hierarchical clustering, which takes MCP curve images and physical data as input. First, the author uses the learning ability of the convolutional layer and max-pooling layer in the VGG16 network to extract the characteristics of each MICP curve. Second, through the PCA dimensionality reduction method, the feature of each picture is transformed into a low-dimensional feature vector. Then, the distance of the image feature vector and the distance of the physical point are combined into a new distance matrix, which is used as the input of hierarchical clustering. Finally, the scores of the Davies Boulding Index and Sillhouette Coefficient are used to determine the final clustering result. Comparing with the FZI and FZI* methods, this method has a better application effect in Jinlong-2 Block volcanic reservoirs. The volcanic rocks of Jinlong 2 are divided into 4 types. Type 1 has the best reservoir characteristics, type 4 is relatively compact, and type 2 and 3 are in the transition zone, but type 2 will develop some micro-fractures to increase reservoir seepage capacity. Such petrophysical rock typing has laid a good foundation for the subsequent reservoir description and reservoir evaluation research of Jinlong-2 Block volcanic reservoirs. • Volcanic rock reservoirs are highly heterogeneous, and petrophysical rock types are diverse. •VGG16 and Hierarchical Clustering for data-driven petrophysical rock typing. •Petrophysical rock types in the Jinlong-2 Block of Xinjiang Oilfield are divided into four types. •Results analysis by DBI and S score.