Explaining Machine Learning Models for Clinical Gait Analysis

Machine Learning (ML) is increasingly used to support decision-making in the healthcare sector. While ML approaches provide promising results with regard to their classification performance, most share a central limitation, their black-box character. This article investigates the usefulness of Explainable Artificial Intelligence (XAI) methods to increase transparency in automated clinical gait classification based on time series. For this purpose, predictions of state-of-the-art classification methods are explained with a XAI method called Layer-wise Relevance Propagation (LRP). Our main contribution is an approach that explains class-specific characteristics learned by ML models that are trained for gait classification. We investigate several gait classification tasks and employ different classification methods, i.e., Convolutional Neural Network, Support Vector Machine, and Multi-layer Perceptron. We propose to evaluate the obtained explanations with two complementary approaches: a statistical analysis of the underlying data using Statistical Parametric Mapping and a qualitative evaluation by two clinical experts. A gait dataset comprising ground reaction force measurements from 132 patients with different lower-body gait disorders and 62 healthy controls is utilized. Our experiments show that explanations obtained by LRP exhibit promising statistical properties concerning inter-class discriminativity and are also in line with clinically relevant biomechanical gait characteristics.

Columnar Aerosol Optical Property Characterization and Aerosol Typing Based on Ground-Based Observations in a Rural Site in the Central Yangtze River Delta Region

Accurate and updated aerosol optical properties (AOPs) are of vital importance to climatology and environment-related studies for assessing the radiative impact of natural and anthropogenic aerosols. We comprehensively studied the columnar AOP observations between January 2019 and July 2020 from a ground-based remote sensing instrument located at a rural site operated by Central China Comprehensive Experimental Sites in the center of the Yangtze River Delta (YRD) region. In order to further study the aerosol type, two threshold-based aerosol classification methods were used to investigate the potential categories of aerosol particles under different aerosol loadings. Based on AOP observation and classification results, the potential relationships between the above-mentioned results and meteorological factors (i.e., humidity) and long-range transportation processes were analyzed. According to the results, obvious variation in aerosol optical depth (AOD) during the daytime, as well as throughout the year, was revealed. Investigation into AOD, single-scattering albedo (SSA), and absorption aerosol optical depth (AAOD) revealed the dominance of fine-mode aerosols with low absorptivity. According to the results of the two aerosol classification methods, the dominant aerosol types were continental (accounting for 43.9%, method A) and non-absorbing aerosols (62.5%, method B). Longer term columnar AOP observations using remote sensing alongside other techniques in the rural areas in East China are still needed for accurate parameterization in the future.

Predicting students’ performance in e-learning using learning process and behaviour data

E-learning is achieved by the deep integration of modern education and information technology, and plays an important role in promoting educational equity. With the continuous expansion of user groups and application areas, it has become increasingly important to effectively ensure the quality of e-learning. Currently, one of the methods to ensure the quality of e-learning is to use mutually independent e-learning behaviour data to build a learning performance predictor to achieve real-time supervision and feedback during the learning process. However, this method ignores the inherent correlation between e-learning behaviours. Therefore, we propose the behaviour classification-based e-learning performance (BCEP) prediction framework, which selects the features of e-learning behaviours, uses feature fusion with behaviour data according to the behaviour classification model to obtain the category feature values of each type of behaviour, and finally builds a learning performance predictor based on machine learning. In addition, because existing e-learning behaviour classification methods do not fully consider the process of learning, we also propose an online behaviour classification model based on the e-learning process called the process-behaviour classification (PBC) model. Experimental results with the Open University Learning Analytics Dataset (OULAD) show that the learning performance predictor based on the BCEP prediction framework has a good prediction effect, and the performance of the PBC model in learning performance prediction is better than traditional classification methods. We construct an e-learning performance predictor from a new perspective and provide a new solution for the quantitative evaluation of e-learning classification methods.

A Multibranch of Convolutional Neural Network Models for Electroencephalogram-Based Motor Imagery Classification

Automatic high-level feature extraction has become a possibility with the advancement of deep learning, and it has been used to optimize efficiency. Recently, classification methods for convolutional neural network (CNN)-based electroencephalography (EEG) motor imagery have been proposed, and have achieved reasonably high classification accuracy. These approaches, however, use the CNN single convolution scale, whereas the best convolution scale varies from subject to subject. This limits the precision of classification. This paper proposes multibranch CNN models to address this issue by effectively extracting the spatial and temporal features from raw EEG data, where the branches correspond to different filter kernel sizes. The proposed method’s promising performance is demonstrated by experimental results on two public datasets, the BCI Competition IV 2a dataset and the High Gamma Dataset (HGD). The results of the technique show a 9.61% improvement in the classification accuracy of multibranch EEGNet (MBEEGNet) from the fixed one-branch EEGNet model, and 2.95% from the variable EEGNet model. In addition, the multibranch ShallowConvNet (MBShallowConvNet) improved the accuracy of a single-scale network by 6.84%. The proposed models outperformed other state-of-the-art EEG motor imagery classification methods.

Prediction of ADHD diagnosis using brief, low-cost, clinical measures: a competitive model evaluation

Proper diagnosis of ADHD is costly, requiring in-depth evaluation via interview, multi-informant and observational assessment, and scrutiny of possible other conditions. The increasing availability of data may allow the development of machine-learning algorithms capable of accurate diagnostic predictions using low-cost measures. We report on the performance of multiple classification methods used to predict a clinician-consensus ADHD diagnosis. Classification methods ranged from fairly simple (e.g., logistic regression) to more complex (e.g., random forest), and also included a multi-stage Bayesian approach. All methods were evaluated in two large (N>1000), independent cohorts. The multi-stage Bayesian classifier provides an intuitive approach that is consistent with clinical workflows, and is able to predict ADHD diagnosis with high accuracy (>86%)—though not significantly better than other commonly used classifiers, including logistic regression. Results suggest that data from parent and teacher surveys is sufficient for high-confidence classifications in the vast majority of cases using relatively straightforward methods.

Classification on Digital Pathological Images of Breast Cancer Based on Deep Features of Different Levels

Breast cancer is one of the primary causes of cancer death in the world and has a great impact on women’s health. Generally, the majority of classification methods rely on the high-level feature. However, different levels of features may not be positively correlated for the final results of classification. Inspired by the recent widespread use of deep learning, this study proposes a novel method for classifying benign cancer and malignant breast cancer based on deep features. First, we design Sliding + Random and Sliding + Class Balance Random window slicing strategies for data preprocessing. The two strategies enhance the generalization of model and improve classification performance on minority classes. Second, feature extraction is based on the AlexNet model. We also discuss the influence of intermediate- and high-level features on classification results. Third, different levels of features are input into different machine-learning models for classification, and then, the best combination is chosen. The experimental results show that the data preprocessing of the Sliding + Class Balance Random window slicing strategy produces decent effectiveness on the BreaKHis dataset. The classification accuracy ranges from 83.57% to 88.69% at different magnifications. On this basis, combining intermediate- and high-level features with SVM has the best classification effect. The classification accuracy ranges from 85.30% to 88.76% at different magnifications. Compared with the latest results of F. A. Spanhol’s team who provide BreaKHis data, the presented method shows better classification performance on image-level accuracy. We believe that the proposed method has promising good practical value and research significance.

Research on the Evaluation of Cross-Border E-Commerce Overseas Strategic Climate Based on Decision Tree and Adaptive Boosting Classification Models

At present, China’s cross-border e-commerce has ushered in a golden period of development. When developing cross-border e-commerce, enterprises should first assess the market climate of the target country and reasonably select the target country. Based on the PESTEL theory, an evaluation index system is established for China’s cross-border e-commerce overseas strategic climate. Taking “One Belt, One Road” as the opportunity and background, the overseas strategic climate of cross-border e-commerce in 62 countries along the “One Belt, One Road” is selected as the research object, and the Decision Tree and Adaptive Boosting classification methods in machine learning are applied to train and predict the established index system. Finally an overall picture of the overseas strategic climate of the 62 countries is obtained. The results are compared and analysed in depth to identify the most suitable countries for cross-border e-merchants and to provide reference for cross-border e-merchants investors.

Automatic Web Page Classification System with Improved Accuracy

Nowadays, the Internet contain s a wide variety of online documents, making finding useful information about a given subject impossible, as well as retrieving irrelevant pages. Web document and page recognition software is useful in a variety of fields, including news, medicine, and fitness, research, and information technology. To enhance search capability, a large number of web page classification methods have been proposed, especially for news web pages. Furthermore existing classification approaches seek to distinguish news web pages while still reducing the high dimensionality of features derived from these pages. Due to the lack of automated classification methods, this paper focuses on the classification of news web pages based on their scarcity and importance. This work will establish different models for the identification and classification of the web pages. The data sets used in this paper were collected from popular news websites. In the research work we have used BBC dataset that has five predefined categories. Initially the input source can be preprocessed and the errors can be eliminated. Then the features can be extracted depend upon the web page reviews using Term frequency-inverse document frequency vectorization. In the work 2225 documents are represented with the 15286 features, which represents the tf-idf score for different unigrams and bigrams. This type of the representation is not only used for classification task also helpful to analyze the dataset. Feature selection is done by using the chi-squared test which will be in the task of finding the terms that are most correlated with each of the categories. Then the pointed features can be selected using chi-squared test. Finally depend upon the classifier the web page can be classified. The results showed that list has obtained the highest percentage, which reflect its effectiveness on the classification of web pages.