Shallow Machine Learning Algorithms Research Articles

MGCN-PolyA: An Integrated Computational Framework for Predicting Poly(A) Signals with Multiscale-gated Convolutional Networks

Background: The accurate recognition of the polyadenylation signal (PAS) from DNA sequences is essential for understanding gene transcriptional regulation. A variety of machine learning-based computational methods have been developed to predict PAS in recent years; however, their performance and their generalization ability are unsatisfactory. It is highly desirable to design more preferable computational approaches for PAS prediction. Methods: In this work, we developed an integrated framework MGCN-PolyA for PAS prediction across four species, including Homo sapiens, Bos taurus, Mus musculus, and Drosophila melanogaster. MGCN-Poly(A) benefits from the diversity of feature engineering and the effectiveness of the model architecture. We combined features from different perspectives, such as word embedding, One-hot encoding, K-mer frequency, and Enhanced Nucleic Acid Composition (ENAC), which complement each other and provide rich and comprehensive information for model learning. In model architecture, MGCN-Poly(A) leverages a two-channel multi-scale gated convolutional network to effectively learn high-level feature representations at different scales, and then combines the statistical features to predict PAS using random forest algorithm. These designs not only speed up network training, but also improves the generalization ability Results: The benchmarking experiments on the independent test datasets demonstrate that MGCNPolyA outperforms other state-of-the-art algorithms in identifying PAS. MGCN-PolyA has the highest accuracy on all test datasets, and its excellent performance on cross-species validation also demonstrates the robustness of our model. Conclusion: Extracting features from different perspectives is important for PAS recognition, and the integration of DNNs and shallow machine learning algorithms can improve the model performance.

Classification of reflective writing: A comparative analysis with shallow machine learning and pre-trained language models

AbstractReflective practice holds critical importance, for example, in higher education and teacher education, yet promoting students’ reflective skills has been a persistent challenge. The emergence of revolutionary artificial intelligence technologies, notably in machine learning and large language models, heralds potential breakthroughs in this domain. The current research on analyzing reflective writing hinges on sentence-level classification. Such an approach, however, may fall short of providing a holistic grasp of written reflection. Therefore, this study employs shallow machine learning algorithms and pre-trained language models, namely BERT, RoBERTa, BigBird, and Longformer, with the intention of enhancing the document-level classification accuracy of reflective writings. A dataset of 1,043 reflective writings was collected in a teacher education program at a German university (M = 251.38 words, SD = 143.08 words). Our findings indicated that BigBird and Longformer models significantly outperformed BERT and RoBERTa, achieving classification accuracies of 76.26% and 77.22%, respectively, with less than 60% accuracy observed in shallow machine learning models. The outcomes of this study contribute to refining document-level classification of reflective writings and have implications for augmenting automated feedback mechanisms in teacher education.

A Systematic Review of Electroencephalography-Based Emotion Recognition of Confusion Using Artificial Intelligence

Confusion emotion in a learning environment can motivate the learner, but prolonged confusion hinders the learning process. Recognizing confused learners is possible; nevertheless, finding them requires a lot of time and effort. Due to certain restrictions imposed by the settings of an online learning environment, the recognition of confused students is a big challenge for educators. Therefore, novel technologies are necessary to handle such crucial difficulties. Lately, Electroencephalography (EEG)-based emotion recognition systems have been rising in popularity in the domain of Education Technology. Such systems have been utilized to recognize the confusion emotion of learners. Numerous studies have been conducted to recognize confusion emotion through this system since 2013, and because of this, a systematic review of the methodologies, feature sets, and utilized classifiers is a timely necessity. This article presents the findings of the review conducted to achieve this requirement. We summarized the published literature in terms of the utilized datasets, feature preprocessing, feature types for model training, and deployed classifiers in terms of shallow machine learning and deep learning-based algorithms. Moreover, the article presents a comparison of the prediction accuracies of the classifiers and illustrates the existing research gaps in confusion emotion recognition systems. Future study directions for potential research are also suggested to overcome existing gaps.

Development of a neural network structure for recognition of attacks in computer systems

The integration of industrial control communication networks and the Internet into the Industrial Control System (ICS) increases their vulnerability to cyber-attacks, leading to devastating consequences. Traditional intrusion detection systems (IDS) mostly rely on predefined models and are trained mostly on specific cyber attacks, which means that traditional IDS cannot deal with unknown attacks In addition, most IDSs do not take into account the imbalance of ICS datasets, and therefore suffer from low accuracy and high false positives when used. In the article, we propose an NCO-double-layer DIFF_RF-OPFYTHON intrusion detection method for ICS, which consists of NCO modules, two-layer DIFF_RF modules, and OPFYTHON modules. Detected traffic will be divided into three categories by the two-layer DIFF_RF module: known attacks, unknown attacks, and normal traffic. Next, the known attacks will be classified by the OPFYTHON module into specific attacks according to the characteristics of the attack traffic. We use the NCO module to improve model inputs and improve model accuracy. The results show that the proposed method outperforms traditional intrusion detection methods such as XGboost and SVM. Detecting unknown attacks is also significant. The accuracy of the data set used in this article reaches 98.13%. The detection rate of unknown and known attacks reaches 98.21% and 95.1%, respectively. In this article, the basic modules of the classifier are shallow machine learning algorithms. This can be improved by using a more powerful neural network architecture. In the article, all detected unknown attacks are marked as one category. In further studies, the unknown attacks can be clustered and analyzed to further classify the detected unknown attacks. It is worth noting that the number of unknown attacks is only a small part, so it is difficult to classify unknown attacks using cluster analysis. Since training a more powerful neural network requires a lot of data, it is important to investigate how to train a new model when the number of samples belonging to that class is limited.

Unsupervised domain adaptation methods for photovoltaic power forecasting

The accurate forecasting of photovoltaic (PV) power generation is of great significance in renewable energy systems, as it enables optimal energy management and grid stability. Despite the importance of this issue, substantial limitations still exist in the majority of existing research initiatives, which employ shallow machine learning algorithms. Recently, some studies have proposed employing convolutional and long short-term memory neural networks (LSTMs) in conjunction with transfer learning techniques; however, these approaches require that the production of PV systems is known during training. To overcome these limitations, we present the first study in the task of PV power forecasting utilizing unsupervised domain adaptation methods. Specifically, we employ two unsupervised methods, namely Domain Adversarial Neural Network and Margin Disparity Discrepancy. Both approaches use a source and a target domain during training, where the target labels of the target domain are unknown during training. We use production and weather data from seven PV systems with nominal capacities ranging from 23.52 kW to 271.53 kW, located in different areas. The findings demonstrate that our proposed architectures improve root mean squared error (RMSE), normalized RMSE, and R2 scores over the smart persistence model across all the PV systems used for testing. Furthermore, our approaches improve the performance of the smart persistence model, with a forecast skill index reaching up to 45.35%. Our extensive experiments demonstrate that our introduced approaches offer valuable advantages over state-of-the-art ones, as the target variable of the target domain is unknown during training. We also demonstrate the robustness of our approaches by conducting a series of ablation experiments.

Data-driven rapid lifetime prediction method for lithium-ion batteries under diverse fast charging protocols

Accurate lithium-ion battery lifetime prediction is essential for equipment maintenance and safety assurance in practical applications. The influence of different charge protocols on the lifetime varies greatly, which raises tremendous challenges for rapid lifetime prediction. In this paper, a comprehensive data-driven rapid lifetime prediction method for batteries under diverse fast charging protocols is proposed. It straightforwardly establishes the relationship between the charging conditions and the impacts on battery lifetime. A deep neural network model with the Bayesian optimization algorithm for hyperparameter search (BOA-DNN) and three traditional shallow machine learning algorithms consisting of support vector machine (SVM), Gaussian process regression (GPR), and extreme gradient boosting (XGBoost) are employed. The verification is performed on an authoritative and popular dataset with 69 types of distinct fast charging conditions. The eventual results indicate that the proposed approach has attained reliable performance of battery lifetime degradation trajectory and battery end-of-life (EOL) prediction, in which the BOA-DNN model outperforms SVM, GPR, and XGBoost. Especially in battery EOL prediction, the prediction mean absolute percentage error of the BOA-DNN model is below 3.58%.

Machine Learning to Predict Apical Lesions: A Cross-Sectional and Model Development Study.

(1) Background: We aimed to identify factors associated with the presence of apical lesions (AL) in panoramic radiographs and to evaluate the predictive value of the identified factors. (2) Methodology: Panoramic radiographs from 1071 patients (age: 11-93 a, mean: 50.6 a ± 19.7 a) with 27,532 teeth were included. Each radiograph was independently assessed by five experienced dentists for AL. A range of shallow machine learning algorithms (logistic regression, k-nearest neighbor, decision tree, random forest, support vector machine, adaptive and gradient boosting) were employed to identify factors at both the patient and tooth level associated with AL and to predict AL. (3) Results: AL were detected in 522 patients (48.7%) and 1133 teeth (4.1%), whereas males showed a significantly higher prevalence than females (52.5%/44.8%; p < 0.05). Logistic regression found that an existing root canal treatment was the most important risk factor (adjusted Odds Ratio 16.89; 95% CI: 13.98-20.41), followed by the tooth type 'molar' (2.54; 2.1-3.08) and the restoration with a crown (2.1; 1.67-2.63). Associations between factors and AL were stronger and accuracy higher when using fewer complex models like decision tree (F1 score: 0.9 (0.89-0.9)). (4) Conclusions: The presence of AL was higher in root-canal treated teeth, those with crowns and molars. More complex machine learning models did not outperform less-complex ones.

Construction of applied talents training system based on machine learning under the background of new liberal arts.

The development of the new liberal arts field places emphasis on the integration of disciplines such as humanities, engineering, medicine, and agriculture. It specifically highlights the incorporation of new technologies into the education and training of liberal arts majors like economics, law, literature, history, and philosophy. However, when dealing with complex talent data, shallow machine learning algorithms may not provide sufficiently accurate evaluations of the relationship between input and output. To address this challenge, this article introduces a comprehensive evaluation model for applied talents based on an improved Deep Belief Network (DBN). In this model, the GAAHS algorithm iteratively generates optimal values that are utilized as connection weights and biases for the restricted Boltzmann machines (RBM) in the pre-training stage of the DBN. This approach ensures that the weights and biases have favorable initial values. Moreover, the paper constructs a quality evaluation index system for creative talents, which consists of four components: knowledge level, innovation practice ability, adaptability to the environment, and psychological quality. The training results demonstrate that the optimized DBN exhibits improved convergence speed and precision, thereby achieving higher accuracy in the classification of applied talent evaluations.

Novel pneumonia score based on a machine learning model for predicting mortality in pneumonia patients on admission to the intensive care unit

BackgroundScores for predicting the long-term mortality of severe pneumonia are lacking. The purpose of this study is to use machine learning methods to develop new pneumonia scores to predict the 1-year mortality and hospital mortality of pneumonia patients on admission to the intensive care unit (ICU). MethodsThe study population was screened from the MIMIC-IV and eICU databases. The main outcomes evaluated were 1-year mortality and hospital mortality in the MIMIC-IV database and hospital mortality in the eICU database. From the full data set, we separated patients diagnosed with community-acquired pneumonia (CAP) and ventilator-associated pneumonia (VAP) for subgroup analysis. We used common shallow machine learning algorithms, including logistic regression, decision tree, random forest, multilayer perceptron and XGBoost. ResultsThe full data set of the MIMIC-IV database contained 4697 patients, while that of the eICU database contained 13760 patients. We defined a new pneumonia score, the "Integrated CCI-APS", using a multivariate logistic regression model including six variables: metastatic solid tumor, Charlson Comorbidity Index, readmission, congestive heart failure, age, and Acute Physiology Score III. The area under the curve (AUC) and accuracy of the integrated CCI-APS were assessed in three data sets (full, CAP, and VAP) using both the test set derived from the MIMIC-IV database and the external validation set derived from the eICU database. The AUC value ranges in predicting 1-year and hospital mortality were 0.784–0.797 and 0.691–0.780, respectively, and the corresponding accuracy ranges were 0.723–0.725 and 0.641–0.718, respectively. ConclusionsThe main contribution of this study was a benchmark for using machine learning models to build pneumonia scores. Based on the idea of integrated learning, we propose a new integrated CCI-APS score for severe pneumonia. In the prediction of 1-year mortality and hospital mortality, our new pneumonia score outperformed the existing score.

Algebraic Structures Induced by the Insertion and Detection of Malware

Since its introduction, researching malware has had two main goals. On the one hand, malware writers have been focused on developing software that can cause more damage to a targeted host for as long as possible. On the other hand, malware analysts have as one of their main purposes the development of tools such as malware detection systems (MDS) or network intrusion detection systems (NIDS) to prevent and detect possible threats to the informatic systems. Obfuscation techniques, such as the encryption of the virus’s code lines, have been developed to avoid their detection. In contrast, shallow machine learning and deep learning algorithms have recently been introduced to detect them. This paper is devoted to some theoretical implications derived from these investigations. We prove that hidden algebraic structures as equipped posets and their categories of representations are behind the research of some infections. Properties of these categories are given to provide a better understanding of different infection techniques.

Hospitalization status and gender recognition over the arboviral medical records using shallow and RNN-based deep models

In global health systems, clinicians have a challenging decision of a triage patient exposed to arbovirus infections to determine they should be hospitalized. Diagnosing symptoms and molecular testing can be uncertain and costly, especially in resource-limited settings. However, machine learning approaches have a high potential to determine through medical record examination whom to hospitalize. The purpose of this study is to determine hospitalized or outpatient individuals correctly by implementing shallow machine learning algorithms on SISA (Severity Index for Suspected Arbovirus) and SISAL (Severity Index for Suspected Arbovirus with Laboratory) datasets. Feed Forward Neural Network (FFNN), Probabilistic Neural Network (PNN), and Decision Tree (DecT) algorithm with three splitting criterions were used to process the SISA and SISAL datasets. The results of classification performances demonstrated that improved area under the curve scores (0.973) and accuracy (reaching up to 98.73% with FFNN) were obtained when compared with previous research study related to machine learning and arbovirus. Moreover, this study also aims to investigate gender recognition over arboviral infection medical records using recurrent neural network-based deep models. Hence vector control policy in the health system reflects the gender roles to control the spread of arboviral infection. Overall the outcomes are potentially very promising.

Fog-cloud based intrusion detection system using Recurrent Neural Networks and feature selection for IoT networks

Deep learning (DL) techniques are being widely researched for their effectiveness in detecting cyber intrusions against the Internet of Things (IoT). Time sensitive Critical Infrastructures (CIs) that rely on IoT require rapid detection of cyber intrusions close to the constrained devices in order to prevent service delays. Deep learning techniques perform better in detecting attacks compared to shallow machine learning algorithms and can be used for intrusion detection. However, communication overheads due to large volume of IoT data and computation requirements for deep learning models prevents effective application of deep learning models closer to the constrained devices. Existing IDS techniques are either based on shallow learning algorithms or not trained on relevant IoT datasets and furthermore not designed for distributed fog-cloud deployment. To counter these issues, we propose a novel fog-cloud based IoT intrusion detection framework which incorporates a distributed processing by splitting the dataset according to attack class and a feature selection step on time-series IoT data. This is followed by a deep learning Recurrent Neural Network (SimpleRNN and Bi-directional Long Short-Term Memory (LSTM)) for attack detection. The effectiveness of the proposed approach was evaluated using the high-dimensional BoT-IoT dataset which contains large volumes of realistic IoT attack traffic. Results show that feature selection methods significantly reduced the dataset size by 90% under the computation requirements without compromising on the attack detection ability. The models built on reduced dataset achieved higher recall rate compared to models trained on full feature set without loosing class differentiation ability. The SimpleRNN and Bi-LSTM models also did not suffer any underfitting or overfitting with the reduced feature space. The proposed deep learning based IoT intrusion detection framework is suitable for fog-cloud based deployment and can scale well even with large volumes of IoT data.

Geological Mapping via Convolutional Neural Network Based on Remote Sensing and Geochemical Survey Data in Vegetation Coverage Areas

Geological mapping in vegetation coverage areas is a challenging task. In this study, convolutional neural networks (CNNs) were employed for geological mapping in a vegetation coverage area based on remote sensing images and geochemical survey data. The Gram-Schmidt fusion technology was first applied to fuse Sentinel-2A and ASTER remote sensing images to enhance the spatial resolution and enrich spectral information of remote sensing data. The fused remote sensing images were then organically integrated with geochemical survey data according to the correlations between the geochemical element contents and spectral reflectance of the objects. A case study of mapping six lithologic units in Jilinbaolige, Inner Mongolia, China was implemented using a CNN model based on the fused data. The classification map obtained an overall accuracy of 83.0%, which exhibited a better performance in contrast to random forest (RF) model. The results showed that CNNs can take full advantage of the spatial features of fused data and solve the problems of the ‘salt and pepper phenomenon’ against the shallow machine learning algorithms, and the fusion of remote sensing and geochemical data can provide rich diagnostic information for geological mapping.

Early-Season Crop Identification in the Shiyang River Basin Using a Deep Learning Algorithm and Time-Series Sentinel-2 Data

Timely and accurate crop identification and mapping are of great significance for crop yield estimation, disaster warning, and food security. Early-season crop identification places higher demands on the quality and mining of time-series information than post-season mapping. In recent years, great strides have been made in the development of deep-learning algorithms, and the emergence of Sentinel-2 data with a higher temporal resolution has provided new opportunities for early-season crop identification. In this study, we aimed to fully exploit the potential of deep-learning algorithms and time-series Sentinel-2 data for early-season crop identification and early-season crop mapping. In this study, four classifiers, i.e., two deep-learning algorithms (one-dimensional convolutional networks and long and short-term memory networks) and two shallow machine-learning algorithms (a random forest algorithm and a support vector machine), were trained using early-season Sentinel-2 images and field samples collected in 2019. Then, these algorithms were applied to images and field samples for 2020 in the Shiyang River Basin. Twelve scenarios with different classifiers and time intervals were compared to determine the optimal combination for the earliest crop identification. The results show that: (1) the two deep-learning algorithms outperformed the two shallow machine-learning algorithms in early-season crop identification; (2) the combination of a one-dimensional convolutional network and 5-day interval time-series Sentinel-2 data outperformed the other schemes in obtaining the early-season crop identification time and achieving early mapping; and (3) the early-season crop identification mapping time in the Shiyang River Basin was identified as the end of July, and the overall classification accuracy reached 0.83. In addition, the early identification time for each crop was as follows: the wheat was in the flowering stage (mid-late June); the alfalfa was in the first harvest (mid-late June); the corn was in the early tassel stage (mid-July); the fennel and sunflower were in the flowering stage (late July); and the melons were in the fruiting stage (around late July). This study demonstrates the potential of using Sentinel-2 time-series data and deep-learning algorithms to achieve early-season crop identification, and this method is expected to provide new solutions and ideas for addressing early-season crop identification monitoring.

Detection of sub-healthy apples with moldy core using deep-shallow learning for vibro-acoustic multi-domain features

The apples with early moldy-core symptoms are in the sub-healthy status and still keep commodity value. The accurate recognition of sub-healthy fruit with moldy core is a valuable and challenging task for the apple industry. In this study, the well-trained IResNet50 network was used as a feature extractor to automatically mine the sensitive features from pixel grid images for time and frequency domains, and the time-frequency images. The extracted multi-domain deep features were fused by adaptive weighting and then fed into two shallow machine learning algorithms including support vector machine (SVM) and extreme learning machine (ELM) for classification. The parameters of SVM and ELM classifiers were optimized by a particle swarm optimization algorithm (PSO). The constructed IResNet50-PSO-ELM hybrid model achieved the overall classification accuracy of 96.7 %, and the recall value of 100 % for healthy fruit, 94.1 % for sub-healthy fruit, 96.2 % for diseased fruit. Also, the F1 values of the IResNet50-PSO-ELM hybrid model were over 95 % for the three-class classification. A comparison with existing studies demonstrated the potential of the proposed method for detecting the sub-healthy apples with moldy core.

Big Data-Based Performance Analysis of Tunnel Boring Machine Tunneling Using Deep Learning

In tunnel boring machine (TBM) construction, the advance rate is a crucial parameter that affects the TBM driving efficiency, project schedule, and construction cost. During the operation process, various types of indicators that are monitored in real-time can help to control the advance rate of TBM. Although some studies have already been carried out in advance rate prediction, the research is almost all based on statistical methods and shallow machine learning algorithms, thereby having difficulties in dealing with a very large amount of monitored data and in modeling the time-dependent characteristics of the parameters. To solve this problem, a deep learning model is proposed based on the CNN architecture, bidirectional Long Short-Term Memory module, and the attention mechanism, which is called the CNN-Bi-LSTM-Attention model. In the first step, the monitored data is processed, and the CNN architecture is adopted to extract features from the data sequence. Then the Bi-LSTM module is adopted to obtain the time-dependent indicators. The significant features can be addressed by the added attention mechanism. In the model training process, the rotation speed of the cutter head (N), thrust (F), torque (T), penetration rate (P), and chamber earth pressure (Soil_P) are adopted to predict the advance rate. The influence of the training periods on the model performance is also discussed. The result shows that not only the data amount, but also the data periods have an influence on the prediction. The long-term data may lead to a failure of the advance rate of TBM. The model evaluation result on the test data shows that the proposed model cannot predict the monitored data in the starting stage, which denotes that the working state of TBM in the starting stage is not stable. Especially when the TBM starts to work, the prediction error is big. The proposed model is also compared with several traditional machine methods, and the result shows the excellent performance of the proposed model.

Comparing Skin Cancer Diagnosis between manual, 4 classical, and 1 deep machine learning algorithms

Skin cancer incidence has increased significantly with approximately 1.2 million melanoma cases diagnosed each year globally. An experienced dermatologist’s visual inspection has only been able to achieve a maximum accuracy of 78%. Recently, convolution neural networks algorithms have successfully outperformed human detection. The study aim was to compare the accuracy and computational time of one deep and four shallow machine learning classification algorithms and demonstrate their statistical superiority over human detection. One deep and four shallow learning algorithms were used to compare the detection accuracy and computational time. Using verified skin lesion images from the ISIC Database, the deep and shallow machine learning algorithms were trained, validated, classified and tested to detect a cancerous skin mole. The results showed that the accuracy of the deep algorithm (VGG-16- 98.86%) was superior to both the shallow algorithms (SVMs – 88.29%, decision trees-88.62%, logistic regression -88.26%, neural network -88.59%) as well as human detection (78.3%). The shallow algorithms were also superior to human detection. Deep Learning algorithms, specifically convolution neural networks, can reduce the false negatives and false positives during melanoma detection and could also be used for early detection of skin cancer at home, saving lives and significant costs.&#x0D;

A Novel Condenser Vacuum Degree Prediction Model Based on LSTM and MemN2N

Condenser vacuum degree prediction of power plants is a challenge task in power system security field. Most existing studies are based on shallow machine learning algorithms, which fail to leverage historical data comprehensively, resulting inaccuracy and unreliable predictions. Therefore, using a serialization model like Recurrent Neural Network to capture time-series information from historical data is necessary. However, these serialization model alone has inherent defects in dealing with long-distance dependence, which may cause historical information forgetting problem. This paper proposes a new prediction model combining LSTM and End-To-End Memory Network (MemN2N). We use LSTM to mine the long-distance dependency information in historical data, and introduce the encoding historical information into the memory pool of MemN2N. MemN2N allows better preservation of historical information for serialization model LSTM, and can make accurate and reliable predictions through soft attention mechanism. Through the experiments on real data from the power plant show that, compared with other prediction models, the model proposed in this paper achieves higher prediction accuracy and has great engineering value.

Assessing Human Motion During Exercise Using Machine Learning: A Literature Review

The World Health Organization promotes healthy living through regular physical activities, such as exercise and sports, as well as access to healthcare and rehabilitation services for people with motor dysfunctions. However, there is a lack of specialized personnel and increased costs associated with such activities. These have led to the increased use of machine learning for the analysis and evaluation of human motion during exercise. To study the latest advancements in this area, a systematic literature review focusing on publications from 2017 to 2021 was performed. As a result, 88 relevant publications were identified, which developed both shallow machine learning and deep learning algorithms. The results indicated that algorithms for human motion assessment should provide personalized and informative assessments, with explainable and interpretable outcomes, that can be computed in real-time or concurrently with the execution of an exercise. Furthermore, they should be easy to adapt based on the needs of applications and should be able to perform with different motion capture systems. This has been challenging because of the usually small amount of collected data, the lack of large open datasets, and the unique characteristics of exercise motions. Based on the above findings, guidelines for the development of such algorithms are proposed and discussed. They relate to the selection of the type of assessment, handling data imbalances, selecting of motion capture technologies, balancing between accuracy and speed, selecting the right algorithm, performing concurrent assessment during an exercise, personalization and scalability, and evaluation.

Transformer Fault Diagnosis Based on Multi-Class AdaBoost Algorithm

Traditional shallow machine learning algorithms cannot effectively explore the relationship between the fault data of oil-immersed transformers, resulting in low fault diagnosis accuracy. This paper proposes a transformer fault diagnosis method based on Multi-class AdaBoost Algorithms in response to this problem. First, the AdaBoost algorithm is combined with Support Vector Machines (SVM), The SVM is enhanced through the AdaBoost algorithm, and the transformer fault data is deeply explored. Then the dynamic weight is introduced into the Particle Swarm Optimization (PSO); through the real-time update of the particle inertia weight, the search accuracy and optimization speed of the particle swarm optimization algorithm is improved, and the improved particle swarm optimization algorithm (IPSO) is used to optimize the parameters of the SVM. Finally, by analyzing the relationship between the dissolved gas in the transformer oil and the fault type, the uncoded ratio method forms a new gas group cooperation. The improved ratio method is constructed as the input feature vector. Simulations based on 117 sets of IECTC10 standard data and 419 sets of transformer fault data collected in China show that the diagnosis method proposed in this paper has strong search ability and fast convergence speed and has a significant improvement in diagnostic accuracy compared with traditional methods.