Kaggle Platform Research Articles

Credit risk assessment mechanism of personal auto loan based on PSO-XGBoost Model

As online P2P loans in automotive financing grows, there is a need to manage and control the credit risk of the personal auto loans. In this paper, the personal auto loans data sets on the Kaggle platform are used on a machine learning based credit risk assessment mechanism for personal auto loans. An integrated Smote-Tomek Link algorithm is proposed to convert the data set into a balanced data set. Then, an improved Filter-Wrapper feature selection method is presented to select credit risk assessment indexes for the loans. Combining Particle Swarm Optimization (PSO) with the eXtreme Gradient Boosting (XGBoost) model, a PSO-XGBoost model is formed to assess the credit risk of the loans. The PSO-XGBoost model is compared against the XGBoost, Random Forest, and Logistic Regression models on the standard performance evaluation indexes of accuracy, precision, ROC curve, and AUC value. The PSO-XGBoost model is found to be superior on classification performance and classification effect.

Analyzing the Effectiveness of COVID-19 Mitigation Policies Using ARIMA Forecasting

The SARS-CoV-2 virus has triggered a worldwide pandemic situation which countries are desperately trying to adapt to. In order to halt the transmission of this virus, these countries have implemented COVID-19 mitigation policies, which are designed to suppress the spread and deadliness of the virus. However, there has not been much research into the effectiveness of these COVID-19 mitigation policies. Using data from the Kaggle Platform as well as the European Centre for Disease Prevention and Control, we hope to use an ARIMA time series forecasting model in order to identify effective COVID-19 mitigation polices. This will be done by analyzing the cases time series before the mitigation policy was implemented in a certain country and generating a predicted forecast curve during the time range of the mitigation policy. By comparing this generated curve with the actual curve, deviations will be able to be identified, indicating the significance of the mitigation policy during its implementation. Although most forecasting was relatively inaccurate due to a shortage in training data, one social distancing mitigation policy in South Korea had a clear deviation between the forecast curve without the influence of the mitigation policy and the actual curve. Overall, the ARIMA model has its merits and may prove to be useful with the collection of more data. By analyzing the effectiveness of these policies, future research into this topic may lead to a greater understanding about the transmission of COVID-19 and ways to suppress it.

Learning from imbalanced COVID-19 chest X-ray (CXR) medical imaging data.

The trendy task of digital medical image analysis has been continually evolving. It has been an area of prominent and growing importance from both research and deployment perspectives. Nonetheless, it is necessary to realize that the use of algorithms, methodology, as well as the source of medical image data, must be strictly scrutinized. As the COVID-19 pandemic has been gripping much of the world recently, there has been much efforts gone into developing affordable testing for the masses, and it has been shown that the established and widely available chest X-rays (CXR) images may be used as a screening criteria for assistive diagnosis purpose. Thanks to the dedicated work by various individuals and organizations, publicly available CXR of COVID-19 subjects are available for analytic usage. We have also provided a publicly available CXR dataset on the Kaggle platform. As a case study, this paper presents a systematic approach to learn from a typically imbalanced set of CXR images, which consists of a limited number of publicly available COVID-19 images. Our results show that we are able to outperform the top finishers in a related Kaggle multi-class CXR challenge. The proposed methodology should be able to help guide medical personnel in obtaining a robust diagnosis model to discern COVID-19 from other conditions confidently.

Eliciting Best Practices for Collaboration with Computational Notebooks

Despite the widespread adoption of computational notebooks, little is known about best practices for their usage in collaborative contexts. In this paper, we fill this gap by eliciting a catalog of best practices for collaborative data science with computational notebooks. With this aim, we first look for best practices through a multivocal literature review. Then, we conduct interviews with professional data scientists to assess their awareness of these best practices. Finally, we assess the adoption of best practices through the analysis of 1,380 Jupyter notebooks retrieved from the Kaggle platform. Findings reveal that experts are mostly aware of the best practices and tend to adopt them in their daily work. Nonetheless, they do not consistently follow all the recommendations as, depending on specific contexts, some are deemed unfeasible or counterproductive due to the lack of proper tool support. As such, we envision the design of notebook solutions that allow data scientists not to have to prioritize exploration and rapid prototyping over writing code of quality.

The influence of preprocessing on the accuracy of the classification of posts on social networks about the corona virus

At the beginning of 2020, there was a pandemic caused by the corona virus. At the global level, restrictive measures were introduced in order to prevent the further spread of the virus, after which almost all aspects of life were reduced to working from home. The restriction of live communication has led to increased user activity on social networks. Analysis of the published content on them can provide insight into the feelings and attitudes that prevail among users. In this paper, a classification of announcements about the corona virus on the Twitter social network written in English was made. A publicly available database from the Kaggle platform was used. Tweets are classified based on their sentimental meaning into one of five classes: extremely positive, positive, neutral, negative, and extremely negative tweets. The goal of the paper is to examine how data preprocessing affects classification accuracy. The Naive Bayes classifier, KNN, and artificial neural networks were used. The results indicate that the method of preprocessing links and tagging does not affect the classification accuracy, but the method of processing hashtags can have an effect on accuracy.

Classification and Detection of Autism Spectrum Disorder Based on Deep Learning Algorithms.

Autism spectrum disorder (ASD) is a type of mental illness that can be detected by using social media data and biomedical images. Autism spectrum disorder (ASD) is a neurological disease correlated with brain growth that later impacts the physical impression of the face. Children with ASD have dissimilar facial landmarks, which set them noticeably apart from typically developed (TD) children. Novelty of the proposed research is to design a system that is based on autism spectrum disorder detection on social media and face recognition. To identify such landmarks, deep learning techniques may be used, but they require a precise technology for extracting and producing the proper patterns of the face features. This study assists communities and psychiatrists in experimentally detecting autism based on facial features, by using an uncomplicated web application based on a deep learning system, that is, a convolutional neural network with transfer learning and the flask framework. Xception, Visual Geometry Group Network (VGG19), and NASNETMobile are the pretrained models that were used for the classification task. The dataset that was used to test these models was collected from the Kaggle platform and consisted of 2,940 face images. Standard evaluation metrics such as accuracy, specificity, and sensitivity were used to evaluate the results of the three deep learning models. The Xception model achieved the highest accuracy result of 91%, followed by VGG19 (80%) and NASNETMobile (78%).

Classification and Prediction of Cardiovascular Patients Based on Optimal Random Forest Algorithm

Cardiovascular disease is a high-risk disease and therefore machine learning is needed to classify and predict it in order to aid research in the medical field. A prediction model for classifying cardiovascular patients based on an optimised random forest algorithm and comparing the prediction performance of each model. Using publicly available data on cardiovascular disease from the Kaggle platform, classification prediction models for cardiovascular disease were developed based on an integrated learning approach using Random Forest, Parsimonious Bayes, SVM and AdaBoost algorithms based on 12 indicators that may have an impact on the mortality of patients with cardiovascular disease. and classification prediction effects. Using the multiple averaging method to ensure the accuracy of the algorithms, the four types of AUROC values were observed and visualisation using matlab's powerful toolbox yielded the best ROC curve fit for random forest with an AUC value of 0.90.

An Improved VGG16 Model for Pneumonia Image Classification

Image recognition has been applied to many fields, but it is relatively rarely applied to medical images. Recent significant deep learning progress for image recognition has raised strong research interest in medical image recognition. First of all, we found the prediction result using the VGG16 model on failed pneumonia X-ray images. Thus, this paper proposes IVGG13 (Improved Visual Geometry Group-13), a modified VGG16 model for classification pneumonia X-rays images. Open-source thoracic X-ray images acquired from the Kaggle platform were employed for pneumonia recognition, but only a few data were obtained, and datasets were unbalanced after classification, either of which can result in extremely poor recognition from trained neural network models. Therefore, we applied augmentation pre-processing to compensate for low data volume and poorly balanced datasets. The original datasets without data augmentation were trained using the proposed and some well-known convolutional neural networks, such as LeNet AlexNet, GoogLeNet and VGG16. In the experimental results, the recognition rates and other evaluation criteria, such as precision, recall and f-measure, were evaluated for each model. This process was repeated for augmented and balanced datasets, with greatly improved metrics such as precision, recall and F1-measure. The proposed IVGG13 model produced superior outcomes with the F1-measure compared with the current best practice convolutional neural networks for medical image recognition, confirming data augmentation effectively improved model accuracy.

Adli Bilişim İncelenme Süreçlerinde Yapay Zeka Kullanımı: VGG16 ile Görüntü Sınıflandırma

Son yıllarda teknolojide meydana gelen gelişmelerle beraber başta internet ve sosyal medya olmak üzere bulut bilişim, akıllı telefon ve navigasyon sistemleri gibi uygulamaların kullanım oranları artmıştır. İnternet ve bilişim cihazlarının yoğun kullanımı, beraberinde depolanan veya aktarılan veri miktarını arttırmış ve bu artış aynı zamanda dijital dünya ile ilişkilendirilen suç oranının da yükselmesine neden olmuştur. İşlenen suçlara ilişkin elde edilen delil boyutu da paralel olarak artmış ve artan veri miktarı, adli bilişim uzmanlarının mevcut imkânlarla veriyi analiz edebilmesini zorlaştırmıştır. Adli bilişim veri inceleme süreçlerinde yaşanan aksamalar nihai olarak adli yargılama süreçlerini de olumsuz etkilemiştir. Söz konusu sorunların giderilmesi kapsamında makalede, elde edilen görüntü verilerinin hızlı ve doğru olarak analiz edilmesini sağlayan bir model önerilmiştir. Önerilen model, VGG16 ağ yapısı ile görüntü sınıflandırma için özel tasarlanan ağ katmanlarından oluşmaktadır. Çalışmada, 2085’i Kaggle platformundan 915’i farklı kaynaklardan oluşturulan 300*300 piksel çözünürlüklü resimlerden oluşan veri seti kullanılmıştır. Model, FloydHub ortamında Keras ve TensorFlow kütüphaneleri ile test edilmiştir. Test sonuçlarına göre modelde %97.8 doğruluk oranı elde edilmiştir. Elde edilen sonuç, benzer çalışmalarla kıyaslanmış ve önerilen modelin diğer çalışmalara oranla ortalama %5 oranında performans artışı sağladığı görülmüştür.

Research on Flower Image Classification Method Based on YOLOv5

The rapid development of deep learning has accelerated the progress of related technologies in the computer vision field and it has broad application prospects. Due to flower inter-class similarity and intra-class differences, flower image classification has essential research value. To achieve flower image classification, this paper proposes a deep learning method using the current powerful object detection algorithm YOLOv5 to achieve fine-grained image classification of flowers. Overlap and occluded objects often appear in the images of the flowers, so the DIoU_NMS algorithm is used to select the target box to enhance the detection of the blocked objects. The experimental dataset comes from the Kaggle platform, and experimental results show that the proposed model in this paper can effectively identify five types of flowers contained in the dataset, Precision reaching 0.942, Recall reaching 0.933, and mAP reaching 0.959. Compared with YOLOv3 and Faster-RCNN, this model has high recognition accuracy, real-time performance, and good robustness. The mAP of this model is 0.051 higher than the mAP of YOLOv3 and 0.102 higher than the mAP of Raster-RCNN.

Alert System of Earthquake Detection

Abstract: Global warning shows unpredictablenature that changes in subsurface geologicfeatures. Due to the complex nature ofseismic events, it is challengeable task to efficiently identify the prominent features that leads to seismic events. Taking the advantage of availability of Seismic dataset, AI using machine learning is a powerful statistical tools to mitigate these practical challenges for earthquake prediction. The paper focuses on the alert and prediction model of an earthquake using machine learning algorithm. The alert time is a function of distance from epicenter and most alert time is for high magnitude earthquake, because high magnitude earthquake rupture over much larger area and take time to propagate, thus delayed for warning. The main aim of the project is to build precise earthquake prediction model using XGboost regression in machine learning technique. The XGboost regression model implemented on the dataset of 14700 data records. The data is gathered from the Kaggle platform that contains data on events of the earthquake in Indian Subcontinent of 18 years and the last updated is of the year 2018. Tableau 2019 is explored for data visualization of the predicted outcome. The proposed technique is able to give precise prediction of earthquake with sufficient time that will reduce the maximum damages and many lives will be saved. Keywords: Seismic signals, XGboost, Tableau, Prediction.

Implementation of Machine Learning Techniques for the Classification of Lung X-Ray Images Used to Detect COVID-19 in Humans

COVID-19 (Coronavirus disease-2019), commonly called Coronavirus or CoV, is a dangerous disease caused by the SARS-CoV-2 virus. It is one of the most widespread zoonotic diseases around the world, which started from one of the wet markets in Wuhan city. Its symptoms are similar to those of the common flu, including cough, fever, muscle pain, shortness of breath, and fatigue. This article suggests implementing machine learning techniques (Random Forest, Logistic Regression, Naïve Bayes, Support Vector Machine) by Python to classify a series of chest X-ray images that include viral pneumonia, COVID-19, and healthy (Not infected) cases in humans. The study includes more than 1400 images that are collected from the Kaggle platform. The experimental outcomes of this study confirmed that the supported vector machine technique has high accuracy and excellent performance in the classification of the disease, as reflected by values of 91.8% accuracy, 91.7% sensitivity, 95.9% specificity, 91.8% F1-score, and 97.6% AUC.

Personal Credit Default Discrimination Model Based on Super Learner Ensemble

Assessing the default of customers is an essential basis for personal credit issuance. This paper considers developing a personal credit default discrimination model based on Super Learner heterogeneous ensemble to improve the accuracy and robustness of default discrimination. First, we select six kinds of single classifiers such as logistic regression, SVM, and three kinds of homogeneous ensemble classifiers such as random forest to build a base classifier candidate library for Super Learner. Then, we use the ten-fold cross-validation method to exercise the base classifier to improve the base classifier’s robustness. We compute the base classifier’s total loss using the difference between the predicted and actual values and establish a base classifier-weighted optimization model to solve for the optimal weight of the base classifier, which minimizes the weighted total loss of all base classifiers. Thus, we obtain the heterogeneous ensembled Super Learner classifier. Finally, we use three real credit datasets in the UCI database regarding Australia, Japanese, and German and the large credit dataset GMSC published by Kaggle platform to test the ensembled Super Learner model’s effectiveness. We also employ four commonly used evaluation indicators, the accuracy rate, type I error rate, type II error rate, and AUC. Compared with the base classifier’s classification results and heterogeneous models such as Stacking and Bstacking, the results show that the ensembled Super Learner model has higher discrimination accuracy and robustness.

APLICAÇÃO DE MACHINE LEARNING NA IDENTIFICAÇÃO DE E-MAILS COMO SPAM

The e-mail service is one of the main tools used today and is an example that technology facilitates the exchange of information. On the other hand, one of the biggest obstacles faced by e-mail services is spam, the name given to the unsolicited message received by a user. The machine learning application has been gaining prominence in recent years as an alternative for efficient identification of spam. In this area, different algorithms can be evaluated to identify which one has the best performance. The aim of the study is to identify the ability of machine learning algorithms to correctly classify e-mails and also to identify which algorithm obtained the greatest accuracy. The database used was taken from the Kaggle platform and the data were processed bythe Orange software with four algorithms: Random Forest (RF), K-Nearest Neighbors (KNN), Support Vector Machine (SVM) and Naive Bayes (NB). The division of data in training and testing considers 80% of the data for training and 20% for testing. The results show that Random Forest was the best performing algorithm with 99% accuracy.

Метод синтезу стійкої багатозв’язної когнітивної карти складної системи

In the article the more effective than analogues method of synthesis of a stable multi-connected cognitive map of a complex system on the basis of a basic stable cognitive map of a smaller order is considered. The peculiarity of this method is that the higher-order cognitive map is designed with a pseudo-triangular adjacency matrix, which is more interconnected than the adjacency matrices used in other similar methods. Such a pseudo-triangular contiguity matrix is designed taking into account the following rules: 1) all elements of the basic cognitive map are elements of the higher order cognitive map, i.e. the contiguity matrix for this higher order cognitive map is built "around" the adjacent cognitive maps were part of the diagonal elements of the higher order cognitive map; 2) the elements of the adjacency matrix of the cognitive map "around" the base cognitive map above or below the main diagonal should be 0, i.e. have the form of a triangular matrix (if you ignore the elements of the base cognitive map, so for such a matrix or rearranging columns and rows. The paper proves that cognitive maps with pseudo-triangular adjacency matrices that satisfy the above rules will be guaranteed to be stable. The proof is made by using the rules of calculating the determinant of the matrix as the sum of algebraic complements of any column or row and taking into account that the weight of cognitive maps does not exceed 1, by definition. To prove the efficiency of the proposed method of synthesis of a stable multi-connected cognitive map, the example of successively located sections of the river on the example of the Southern Bug near Vinnytsia was considered. In this example, the peaks of the cognitive map are the concentration of biochemical oxygen consumption in each section of the river, which is a characteristic of the concentration of organic matter in water, and the target peak is the concentration of dissolved oxygen in the closing section of the river affected by pollution. To automate calculations, the Python program was developed as a public laptop based on the famous Kaggle platform. Studies and computer experiments have shown the effectiveness and efficiency of the proposed method.

METHODS FOR INCREASING THE CLASSIFICATION ACCURACY BASED ON MODIFICATIONS OF THE BASIC ARCHITECTURE OF CONVOLUTIONAL NEURAL NETWORKS

Object of research: basic architectures of deep learning neural networks. Investigated problem: insufficient accuracy of solving the classification problem based on the basic architectures of deep learning neural networks. An increase in accuracy requires a significant complication of the architecture, which, in turn, leads to an increase in the required computing resources, as well as the consumption of video memory and the cost of learning/output time. Therefore, the problem arises of determining such methods for modifying basic architectures that improve the classification accuracy and require insignificant additional computing resources. Main scientific results: based on the analysis of existing methods for improving the classification accuracy on the convolutional networks of basic architectures, it is determined what is most effective: scaling the ScanNet architecture, learning the ensemble of TreeNet models, integrating several CBNet backbone networks. For computational experiments, these modifications of the basic architectures are implemented, as well as their combinations: ScanNet + TreeNet, ScanNet + CBNet. The effectiveness of these methods in comparison with basic architectures has been proven when solving the problem of recognizing malignant tumors with diagnostic images – SIIM-ISIC Melanoma Classification, the train/test set of which is presented on the Kaggle platform. The accuracy value for the area under the ROC curve metric has increased from 0.94489 (basic architecture network) to 0.96317 (network with ScanNet + CBNet modifications). At the same time, the output compared to the basic architecture (EfficientNet-b5) increased from 440 to 490 seconds, and the consumption of video memory increased from 8 to 9.2 gigabytes, which is acceptable. Innovative technological product: methods for achieving high recognition accuracy from a diagnostic signal based on deep learning neural networks of basic architectures. Scope of application of the innovative technological product: automatic diagnostics systems in the following areas: medicine, seismology, astronomy (classification by images) onboard control systems and systems for monitoring transport and vehicle flows or visitors (recognition of scenes with camera frames).

Credit Card Fraud Detection in Payment Using Machine Learning Classifiers

The fraud detection in payment is a classification problem that aims to identify fraudulent transactions based individually on the information it contains and on the basis that a fraudster's behaviour patterns differ significantly from that of the actual customer. In this context, the authors propose to implement machine learning classifiers (Naïve Bayes, C4.5 decision trees, and Bagging Ensemble Learner) to predict the outcome of regular transactions and fraudulent transactions. The performance of these classifiers is judged by the following ways: precision, recall rate, and precision-recall curve (PRC) area rate. The dataset includes more than 297K transactions via credit cards in September 2013 and November 2017 that have been collected from Kaggle platform, of which 3293 are frauds. The performance PRC ratio of machine learning classifiers is between 99.9% and 100%, which confirms that these classifiers are very good at identifying binary classes 0 in the dataset. The results of the tests have proved that the best classifier is C4.5 decision trees. This classifier has the best accuracy of 94.12% in prediction of fraudulent transactions.

The ASHRAE Great Energy Predictor III competition: Overview and results

In late 2019, ASHRAE hosted the Great Energy Predictor III (GEPIII) machine learning competition on the Kaggle platform. This launch marked the third energy prediction competition from ASHRAE and the first since the mid-1990s. In this updated version, the competitors were provided with over 20 million points of training data from 2,380 energy meters collected for 1,448 buildings from 16 sources. This competition’s overall objective was to find the most accurate modeling solutions for the prediction of over 41 million private and public test data points. The competition had 4,370 participants, split across 3,614 teams from 94 countries who submitted 39,403 predictions. In addition to the top five winning workflows, the competitors publicly shared 415 reproducible online machine learning workflow examples (notebooks), including over 40 additional, full solutions. This paper gives a high-level overview of the competition preparation and dataset, competitors and their discussions, machine learning workflows and models generated, winners and their submissions, discussion of lessons learned, and competition outputs and next steps. The most popular and accurate machine learning workflows used large ensembles of mostly gradient boosting tree models, such as LightGBM. Similar to the first predictor competition, preprocessing of the data sets emerged as a key differentiator.

Identification of Skin Disease Using Deep Learning

Globally, skin diseases are among the most common health problems in all humans irrespective of age. Prevention and early detection of these diseases can improve the chance of surviving. This model illustrates the identification of skin diseases providing more objective and reliable solutions using deep learning technology and convolutional neural network approach. In this model, the system design, implementation and identification of common skin diseases such as acne, blister, eczema, warts etc. are explained. The system applies deep learning technology to train itself with various images of skin diseases from the Kaggle platform. The accuracy obtained by using deep learning is 83.23%. The main objective of this system is to achieve maximum accuracy of skin disease prediction. Moreover, if the disease is identified the system provides detailed information about the diseases along with home remedies.

An algorithm competition for automatic species identification from herbarium specimens.

PremisePlant biodiversity is threatened, yet many species remain undescribed. It is estimated that >50% of undescribed species have already been collected and are awaiting discovery in herbaria. Robust automatic species identification algorithms using machine learning could accelerate species discovery.MethodsTo encourage the development of an automatic species identification algorithm, we submitted our Herbarium 2019 data set to the Fine‐Grained Visual Categorization sub‐competition (FGVC6) hosted on the Kaggle platform. We chose to focus on the flowering plant family Melastomataceae because we have a large collection of imaged herbarium specimens (46,469 specimens representing 683 species) and taxonomic expertise in the family. As is common for herbarium collections, some species in this data set are represented by few specimens and others by many.ResultsIn less than three months, the FGVC6 Herbarium 2019 Challenge drew 22 teams who entered 254 models for Melastomataceae species identification. The four best algorithms identified species with >88% accuracy.DiscussionThe FGVC competitions provide a unique opportunity for computer vision and machine learning experts to address difficult species‐recognition problems. The Herbarium 2019 Challenge brought together a novel combination of collections resources, taxonomic expertise, and collaboration between botanists and computer scientists.