PCCNN: A CNN classification model integrating EEG time-frequency features for stroke classification

Prostate cancer is a common malignancy in men, and accurately distinguishing between benign and malignant nodules at an early stage is crucial for optimizing treatment. Multimodal imaging (such as ADC and T2) plays an important role in the diagnosis of prostate cancer, but effectively combining these imaging features for accurate classification remains a challenge. This retrospective study included MRI data from 199 prostate cancer patients. Radiomic features from both the tumor and peritumoral regions were extracted, and a random forest model was used to select the most contributive features for classification. Three machine learning models-Random Forest, XGBoost, and Extra Trees-were then constructed and trained on four different feature combinations (tumor ADC, tumor T2, tumor ADC+T2, and tumor + peritumoral ADC+T2). The model incorporating multimodal imaging features and peritumoral characteristics showed superior classification performance. The Extra Trees model outperformed the others across all feature combinations, particularly in the tumor + peritumoral ADC+T2 group, where the AUC reached 0.729. The AUC values for the other combinations also exceeded 0.65. While the Random Forest and XGBoost models performed slightly lower, they still demonstrated strong classification abilities, with AUCs ranging from 0.63 to 0.72. SHAP analysis revealed that key features, such as tumor texture and peritumoral gray-level features, significantly contributed to the model's classification decisions. The combination of multimodal imaging data with peritumoral features moderately improved the accuracy of prostate cancer classification. This model provides a non-invasive and effective diagnostic tool for clinical use and supports future personalized treatment decisions.

The increasing interest of people for consumption of safe food has necessitated the research on fraud detection in foods. In this research, fraud samples of extra virgin olive oil were detected by fusion of E-nose and ultrasound systems. The data were analyzed with seven different classification algorithms and the samples were prepared in six levels of fraud. The principal component analysis was used for feature reduction and seven different classification models were also used for classification and fraud detection. Based on the results, the most effective features in classification were “losses in ultrasound wave amplitude percentage,” “ultrasound signal’s time of flight,” and “difference between maximum and minimum output of TGS2620 gas sensor,” respectively. As well, the results indicated that the Gradient Boosting Classifier (GBC) model represented the best classification accuracy (97.75%) which had the maximum calculation time, compared with the other methods. Furthermore, both NaiveBayes and linear SVM methods were found to be effective methods (95.51%) after GBC.

Recently, machine learning methods have been successfully used for the prediction of cardiovascular disease. Early diagnosis and prediction is necessary for giving effective treatment to avoid higher mortality rates. Several classification algorithms have been developed recently which satisfy the need, but show limited accuracy while predicting the heart disease. Hence, the focus of this study is on early prediction of heart disease and to improve the accuracy of prediction using benchmark heart disease datasets such as UCI Cleveland dataset and Heart disease clinical dataset by implementing effective classification and optimization algorithms. Optimization algorithms generally exhibit the benefit of dealing with complex non‐linear issues with better adaptability and flexibility. The Emperor penguin optimization algorithm, which can select the best features for classification has been utilized in this study to improve the efficiency, minimize reconstruction errors, and increase the quality of heart disease classification. Further, the newly developed stacked sparse convolutional neural network based auto encoder (SSC‐AE) classification algorithm has been employed for significant feature classification with higher robustness and efficacy. Accuracy, Area Under Curve (AUC), and F1 score are some of the measures used to compare the outcomes of several machine learning algorithms to those of the proposed model in this study. The results show that the proposed model, SSC‐AE, is superior to other classification models.

A CNN model embedded with local feature knowledge and its application to time-varying signal classification

Neuroscience studies have demonstrated the phase-locked characteristics of some early event-related potential (ERP) components evoked by stimuli. In this study, we propose a phase preservation neural network (PPNN) to learn phase information to improve the Electroencephalography (EEG) classification in a rapid serial visual presentation (RSVP) task. The PPNN consists of three major modules that can produce spatial and temporal representations with the high discriminative ability of the EEG features for classification. We first adopt a stack of dilated temporal convolution layers to extract temporal dynamics while avoiding the loss of phase information. Considering the intrinsic channel dependence of the EEG data, a spatial convolution layer is then applied to obtain the spatial-temporal representation of the input EEG signal. Finally, a fully connected layer is adopted to extract higher-level features for the final classification. The experiments are conducted on two public and one collected EEG datasets from the RSVP task, in which we evaluated the performance and explored the capability of phase preservation of our PPNN model and visualized the extracted features. The experimental results indicate the superiority of the proposed PPNN when compared with previous methods, suggesting the PPNN is a robust model for EEG classification in RSVP task.

The manual process in recruitment is too hard when the data is huge and complex. This increase in data can bring inaccuracy in selecting a candidate based on the job description. This project presents the involvement of Machine learning techniques in the recruitment process such as ranking of candidate’s resumes, recommendation of JD (Job description) to candidates and recommendation about resumes to recruiters and recommendation about candidate’s resumes to recruiters. This system increases accuracy, consistency, and speed in the recruitment process. Most of the common processes in recruitment got automated in this system. Techniques like TF-IDF, Cosine similarity and Classification models are used to achieve the features of classification, recommendation and ranking of resumes. TF-IDF vectorization usually converts textual data like resumes or JD (Job description) into numerical vectors, then Cosine similarity is used for providing ranking and recommendation to recruiter and candidates. Models that have higher performance in accuracy, precision, recall and F1-Score are used for resume classification features. This project carries a Dataset from Kaggle, which has suitable and required data for the process like resumes with candidate skills, communication address and experience. Later, using the taken Dataset, we train models like Logistic regression, Support vector machine, Random forest and Naive Bayes to find out which mode fits best in the classification feature.

The article analyzes project management methodologies and standards (in particular, PMBOK, P2M, Agile) in the context of models and methods that can be used to create a “green” project management approach. On the other hand, the concept of "green" entrepreneurship and its principles were analyzed. On this basis, 7 principles of “green” project management were proposed for the further development of the “green entrepreneurship” approach (projects should create environmentally friendly products, goods and services; ensure the maximum possible use of environmentally friendly materials and adhere to the desire to gradually increase the environmental friendliness of products; the implementation of project activities contributes to the creation and development of a “green economy” in the state; project-oriented companies mainly implement socially responsible projects with a trend towards increasing their share in the company's project portfolio; social responsibility should be both corporate and personal). Energy efficiency improvement projects are defined as a special case of “green” projects and as an object of scientific research. A classification model for energy efficiency projects is proposed for the further development of such classifications in the form of 11 classification features and project types within each feature (by project product, by use of artificial intelligence, by field of application, by use of “green” entrepreneurship principles, by project management methodology, by complexity, by type of project management team, by energy efficiency models, by degree of innovation, by geographical distribution, by scalability). 4 clusters of development of “green” project management are also identified, within which it is proposed to implement further research: “green” projects, “green” project management methodologies, “green” teams, “green” management tools. Each such cluster is characterized. A model of supplementing the classification model of project management systems (in its further development) is proposed, in which additional features of the classification of “green” project management systems are formulated - in the form of 5 additional features and types of projects within each classification feature. A model of the components of “green” project management is proposed, which contains four components: classic project management, flexible (Agile) project management, the concept of Lean production and “Green” community. As well as a multiple model of the “green” project management system in the form of a formal seven. A SWOT analysis of the formulated “green” project management approach is conducted, its strengths and weaknesses are highlighted, as well as opportunities and threats that may arise when using it. A conclusion is made regarding the potential effectiveness of such an approach. Areas of further research in the chosen direction are formulated. Conclusions from the conducted research are formulated.

In this paper, we address on the task of sentence paraphrase detection which is focused on deciding whether the two sentences have the relationship of paraphrase. A supervised learning strategy for paraphrase detection is described whereby the two sentences are classified to decide the paraphrase relationship and using only the lexical features operated at n-gram as the classification features. Gradient Boosting, K-Nearest Neighbor, Decision Tree and Support vector machine are chosen as the classifiers. The performance of the classification method is compared and the features are analyzed to determine which of them are most important for paraphrase detection. Evaluation is performed on the corpus of 2016 Detecting Paraphrase in Indian Languages task proposed by Forum of Information Retrieval Evaluation (DPIL-FIRE2016). The experimental results show that the Gradient Boosting can achieve the highest Overall Score. By using the learned classifier, we got the highest F1 measure for both Task1 and Task2 on Malayalam and Tamil, and the highest F1 measure for Task2 on Punjabi in DPIL-FIRE2016.

Protein subcellular localization prediction is the problem of predicting where a protein functions within a living cell. In this paper, we apply associative classifications (CMAR and CPAR) and multi-class Support Vector Machines to tackle the problem of protein subcellular localization prediction. We use classification feature sources generated from a protein's SwissProt annotation record. We visualize the applied classification rules in an explain graph for domain experts to interpret. We compare the performance of our approaches to those of Proteome Analyst 3.0, using the same set of classification features; we find that all three classification algorithms outperform Proteome Analyst. Multi-class SVM achieves overall F-measures [0.934 ~ 0.991], while CPAR and CMAR achieve overall F-measures [0.922 ~ 0.989] and [0.880 ~ 0.989], respectively. Our result shows that despite multi-class SVM is still the most accurate prediction algorithm with overall F-measures, CPAR and CMAR achieve very similar accuracy. In most cases, CPAR outperforms CMAR, especially when the feature space is large. Our result indicates that associative classification algorithms, especially CPAR, is a good alternative to SVM with similar accuracy but much better transparency in classification models.

This study aimed to create a caries classification scheme based on cone-beam computed tomography (CBCT) and develop two deep learning models to improve caries classification accuracy. A total of 2713 axial slices were obtained from CBCT images of 204 carious teeth. Both classification models were trained and tested using the same pretrained classification networks on the dataset, including ResNet50_vd, MobileNetV3_large_ssld, and ResNet50_vd_ssld. The first model was used directly to classify the original images (direct classification model). The second model incorporated a presegmentation step for interpretation (interpretable classification model). Performance evaluation metrics including accuracy, precision, recall, and F1 score were calculated. The Local Interpretable Model-agnostic Explanations (LIME) method was employed to elucidate the decision-making process of the two models. In addition, a minimum distance between caries and pulp was introduced for determining the treatment strategies for type II carious teeth. The direct model that utilized the ResNet50_vd_ssld network achieved top accuracy, precision, recall, and F1 score of 0.700, 0.786, 0.606, and 0.616, respectively. Conversely, the interpretable model consistently yielded metrics surpassing 0.917, irrespective of the network employed. The LIME algorithm confirmed the interpretability of the classification models by identifying key image features for caries classification. Evaluation of treatment strategies for type II carious teeth revealed a significant negative correlation (p < 0.01) with the minimum distance. These results demonstrated that the CBCT-based caries classification scheme and the two classification models appeared to be acceptable tools for the diagnosis and categorization of dental caries.

Deep learning and transfer learning features for plankton classification

In this letter, the pH time-frequency vocal source feature is proposed for multistyle emotion identification. A binary acoustic mask is also used to improve the emotion classification accuracy. Emotional and stress conditions from the Berlin Database of Emotional Speech (EMO-DB) and Speech under Simulated and Actual Stress (SUSAS) databases are investigated in the experiments. In terms of emotion identification rates, the pH outperforms the mel-frequency cepstral coefficients (MFCC) and a Teager-Energy-Operator (TEO) based feature. Moreover, the acoustic mask achieves accuracy improvement for both the MFCC and the pH feature.

Anemia is a global health issue that has a significant impact on quality of life and productivity. Early and accurate detection is essential to prevent more serious complications. This study aims to develop an anemia classification model based on machine learning technology using the XGBoost algorithm, as well as compare its performance with Logistic Regression and Random Forest methods. The dataset used in this study was obtained from the Kaggle platform, consisting of 1,421 samples and six clinical attributes, namely Gender, Hemoglobin (HGB), Mean Corpuscular Hemoglobin (MCH), Mean Corpuscular Hemoglobin Concentration (MCHC), Mean Corpuscular Volume (MCV), Result. During the feature engineering process, the derived feature of the hemoglobin-to-MCV ratio (Hb/MCV) was added, which is medically relevant in distinguishing types of anemia. Evaluation results showed that XGBoost and Random Forest achieved an accuracy rate and F1-Score of 100%, while Logistic Regression achieved a rate of 98.9%. XGBoost was selected as the primary model due to its efficient computational capabilities and support for interpretation using SHAP (SHapley Additive exPlanations). SHAP visualization revealed that the Hb/MCV ratio and hemoglobin were the most influential features in classification. This model has the potential to be used as a decision support system for automated anemia screening and can be further integrated into clinical systems.

Natural language processing (NLP) pipelines are usually complex, including several ways of extracting features and processing the inputs and results. The difficulty of the task directly affects the complexity of the system: multiple modules work together, extracting complementary information that is needed for a good performance. In this work we focus on text classification and show that the same intuition applies to end-to-end neural NLP architectures and that the best results are obtained reliably by combining the information from different neural modules. Concretely, we combine convolution, recurrent and attention modules with ensembles and show that they are complementary. We empirically prove that the combination is robust to various and complex text classification jobs and show that it attains or surpasses the state-of-the-art on a wide variety of datasets with no changes to the architecture. In addition, we show that ensembling CNN-RNN stacks with attention improve the performance with respect to only using a subset of the aforementioned modules. These observations hold in both low and high data availability, as well as for multi-class problems.

This article proposes a model for Chinese text classification based on a feature-enhanced nonequilibrium bidirectional long short-term memory (Bi-LSTM) network that analyzes Chinese text information in depth and improves the accuracy of text classification. First, the bidirectional encoder representations from transformers model was used to vectorize the original Chinese corpus and extract preliminary semantic features. Then, a nonequilibrium Bi-LSTM network was applied to increase the weight of text information containing important semantics and further improve the effects of the key features in Chinese text classification. Simultaneously, a hierarchical attention mechanism was used to widen the gap between the important and unimportant data. Finally, the softmax function was used for classification. By comparing the classification performance of the proposed scheme with those of various other models, it was observed that the model substantially improved the precision of Chinese text classification and had a strong ability to recognize Chinese text features. The model achieved 97% precision on the experimental dataset.