Semi-supervised Machine Learning Algorithm Research Articles

Detection of human movement by combining supervised machine learning and an embroidered textile capacitance sensor

This study contributes to respiratory pattern detection by introducing a fabric sensor utilizing capacitance measurement and a semi-supervised machine learning algorithm known as an AI-based autoencoder. The sensor, consisting of two embroidered electrodes composed of silver-coated conductive nylon filaments, leverages the body as a dielectric material. In the research, a garment-type respiratory sensor was employed to continuously monitor respiratory data during both static (standing) and dynamic (walking, brisk walking, running) actions. The sparse autoencoder algorithm was particularly employed for individual static and dynamic actions, effectively distinguishing respiratory patterns corresponding to various movements. In addition, the sparse autoencoder helps prevent overfitting, fundamentally minimizing errors between the compression and reconstruction of signals. The maximum number of epochs was set to 2000, and the target error was set at 0.005. All data were compared against the static walking as the training baseline. Ultimately, the root mean squared error (RMSE) between static postures averaged 0.1, while the RMSEs between dynamic actions of walking, brisk walking, and running were 0.61, 0.91, and 2.78, respectively. These results suggest that movement detection through error detection is practically feasible and possesses discernible capabilities.

Developing an atmospheric aging evaluation model of acrylic coatings: A semi-supervised machine learning algorithm

Developing an atmospheric aging evaluation model of acrylic coatings: A semi-supervised machine learning algorithm

A deep semi-supervised machine learning algorithm for detecting transportation modes based on GPS tracking data

AbstractTransportation research has benefited from GPS tracking devices since a higher volume of data can be acquired. Trip information such as travel speed, time, and most visited locations can be easily extracted from raw GPS tracking data. However, transportation modes cannot be extracted directly and require more complex analytical processes. Common approaches for detecting travel modes heavily depend on manual labelling of trajectories with accurate trip information, which is inefficient in many aspects. This paper proposes a method of semi-supervised machine learning by using minimal labelled data. The method can accept GPS trajectory with adjustable length and extract latent information with long short-term memory (LSTM) Autoencoder. The method adopts a deep neural network architecture with three hidden layers to map the latent information to detect transportation mode. The proposed method is assessed by applying it to the case study where an accuracy of 93.94% can be achieved, which significantly outperforms similar studies.

Analysing Semi-Supervised ConvNet Model Performance with Computation Processes

The rapid development of semi-supervised machine learning (SSML) algorithms has shown enhanced versatility, but pinpointing the primary influencing factors remains a challenge. Historically, deep neural networks (DNNs) have been used to underpin these algorithms, resulting in improved classification precision. This study aims to delve into the performance determinants of SSML models by employing post-hoc explainable artificial intelligence (XAI) methods. By analyzing the components of well-established SSML algorithms and comparing them to newer counterparts, this work redefines semi-supervised computation processes for both data preprocessing and classification. Integrating different types of DNNs, we evaluated the effects of parameter adjustments during training across varied labeled and unlabeled data proportions. Our analysis of 45 experiments showed a notable 8% drop in training loss and a 6.75% enhancement in learning precision when using the Shake-Shake26 classifier with the RemixMatch SSML algorithm. Additionally, our findings suggest a strong positive relationship between the amount of labeled data and training duration, indicating that more labeled data leads to extended training periods, which further influences parameter adjustments in learning processes.

Two neurosubtypes of ADHD different from the clinical phenotypes

Clinical and etiological variability of attention deficit hyperactivity disorder (ADHD) presents an obstacle to understand the disorder. The aim of this study was to disentangle the heterogeneity of ADHD using neuroimaging and a semi-supervised machine learning algorithm. We collected brain structural and functional magnetic resonance imaging (MRI) data and clinical profiles of 183 children with ADHD and 396 neurotypical controls from 7 independent sites. We also used an external validation set with 750 subjects. We adopted a semi-supervised clustering method to subtype ADHD by regional volumetric measures of gray matter, white matter, and fractional amplitude of low frequency fluctuation (fALFF). In addition, split sample test, leave-one-site-out test and external validation were applied to evaluate the reproducibility and stability of ADHD subtypes. Two stable and reproducible neurosubtypes of ADHD were disclosed, which were proved by the split-sample test and leave-one-site-out validation. The structural and functional patterns of ADHD subtypes were also stable in the external validation set. The current two neurosubtypes differed in clinical manifestations and volumetric gray matter, white matter volume and fALFF patterns. The current neurosubtypes of ADHD which were different from clinical phenotypes could facilitate understanding the underlying neuropathological and neurobiological mechanism of the disorder.

OPTIMIZATION OF ROUTING USING TRAFFIC CLASSIFICATION IN SOFTWARE DEFINED NETWORKING

Efficient routing is an essential task for any network as it directly impacts the network’s performance. In this paper, we have used traffic classification techniques to optimize routing in Software Defined Network (SDN) and provided a cost aware routing framework. Instead of classifying traffic based on a particular parameter, we perform the traffic classification using a hybrid approach that is based on machine learning techniques. Our framework uses supervised machine learning techniques to classify network flow and detect elephant flows which are too heavy in size and require significant bandwidth. As the requirements of each elephant flow vary with the application, we further perform Quality of Service (QoS) based traffic classification, which classifies these elephant flow into QoS classes as per their QoS requirements. For this purpose, we have used semi-supervised machine learning algorithms. Further, we have proposed a routing algorithm that makes use of the Dijkstra algorithm to compute the best possible shortest path based on the QoS requirements of the elephant flow. The proposed method was implemented and tested on Mininet using a POX controller. The simulation results show that our framework successfully classifies elephant flows with an accuracy of 80% and also computes a low-cost path for each elephant flow based on the actual internet data set.

Investigation of the J-TEXT plasma events by k-means clustering algorithm

Various types of plasma events emerge in specific parameter ranges and exhibit similar characteristics in diagnostic signals, which can be applied to identify these events. A semi-supervised machine learning algorithm, the k-means clustering algorithm, is utilized to investigate and identify plasma events in the J-TEXT plasma. This method can cluster diverse plasma events with homogeneous features, and then these events can be identified if given few manually labeled examples based on physical understanding. A survey of clustered events reveals that the k-means algorithm can make plasma events (rotating tearing mode, sawtooth oscillations, and locked mode) gathering in Euclidean space composed of multi-dimensional diagnostic data, like soft x-ray emission intensity, edge toroidal rotation velocity, the Mirnov signal amplitude and so on. Based on the cluster analysis results, an approximate analytical model is proposed to rapidly identify plasma events in the J-TEXT plasma. The cluster analysis method is conducive to data markers of massive diagnostic data.

Correction: Emotional Distress During COVID-19 by Mental Health Conditions and Economic Vulnerability: Retrospective Analysis of Survey-Linked Twitter Data With a Semisupervised Machine Learning Algorithm

[This corrects the article DOI: 10.2196/44965.].

Emotional Distress During COVID-19 by Mental Health Conditions and Economic Vulnerability: Retrospective Analysis of Survey-Linked Twitter Data With a Semisupervised Machine Learning Algorithm.

Monitoring the psychological conditions of social media users during rapidly developing public health crises, such as the COVID-19 pandemic, using their posts on social media has rapidly gained popularity as a relatively easy and cost-effective method. However, the characteristics of individuals who created these posts are largely unknown, making it difficult to identify groups of individuals most affected by such crises. In addition, large annotated data sets for mental health conditions are not easily available, and thus, supervised machine learning algorithms can be infeasible or too costly. This study proposes a machine learning framework for the real-time surveillance of mental health conditions that does not require extensive training data. Using survey-linked tweets, we tracked the level of emotional distress during the COVID-19 pandemic by the attributes and psychological conditions of social media users in Japan. We conducted online surveys of adults residing in Japan in May 2022 and collected their basic demographic information, socioeconomic status, and mental health conditions, along with their Twitter handles (N=2432). We computed emotional distress scores for all the tweets posted by the study participants between January 1, 2019, and May 30, 2022 (N=2,493,682) using a semisupervised algorithm called latent semantic scaling (LSS), with higher values indicating higher levels of emotional distress. After excluding users by age and other criteria, we examined 495,021 (19.85%) tweets generated by 560 (23.03%) individuals (age 18-49 years) in 2019 and 2020. We estimated fixed-effect regression models to examine their emotional distress levels in 2020 relative to the corresponding weeks in 2019 by the mental health conditions and characteristics of social media users. The estimated level of emotional distress of our study participants increased in the week when school closure started (March 2020), and it peaked at the beginning of the state of emergency (estimated coefficient=0.219, 95% CI 0.162-0.276) in early April 2020. Their level of emotional distress was unrelated to the number of COVID-19 cases. We found that the government-induced restrictions disproportionately affected the psychological conditions of vulnerable individuals, including those with low income, precarious employment, depressive symptoms, and suicidal ideation. This study establishes a framework to implement near-real-time monitoring of the emotional distress level of social media users, highlighting a great potential to continuously monitor their well-being using survey-linked social media posts as a complement to administrative and large-scale survey data. Given its flexibility and adaptability, the proposed framework is easily extendable for other purposes, such as detecting suicidality among social media users, and can be used on streaming data for continuous measurement of the conditions and sentiment of any group of interest.

Developing deep learning algorithms for inferring upstream separatrix density at JET

Predictive and real-time inference capability for the upstream separatrix electron density, $n_\text{e, sep}$, is essential for design and control of core-edge integrated plasma scenarios. In this study, both supervised and semi-supervised machine learning algorithms are explored to establish direct mapping as well as indirect compressed representation of the pedestal profiles for predictions and inference of $n_{\text{e, sep}}$. Based on the EUROfusion pedestal database for JET, a tabular dataset was created, consisting of machine parameters, fraction of ELM cycle, high resolution Thomson scattering profiles of electron density and temperature, and $n_{\text{e, sep}}$ for 608 JET shots. Using the tabular dataset, the direct mapping approach provides a mapping of machine parameters and ELM percentage to $n_{\text{e, sep}}$. Through representation learning, a compressed representation of the experimental pedestal electron density and temperature profiles is established. By conditioning the representation with machine control parameters, a probabilistic generative predictive model is established. For prediction, the machine parameters can be used to establish a conditional distribution of the compressed pedestal profiles, and the decoder that is trained as part of the algorithm can be used to decode the compressed representation back to full pedestal profiles. Although, in this work, a proof-of-principle for predicting and inferring $n_{\text{e, sep}}$ is given, such a representation learning can be used also for many other applications as the full pedestal profile is predicted. An implementation of this work can be found at https://github.com/fusionby2030/moxie.

Characterization and Detection of Cross-Router Covert Channels

In covert channel attacks, an adversary seeks various means to influence a tangible characteristic of a system, and then makes the systems leak information by measuring this characteristic. Covert channels are, by nature, very elusive. This makes it very difficult to identify them and defend against attacks that use these channels to leak sensitive information. Thus, they are a serious threat to the security of many systems.In this paper, we present two network timing covert channel attacks, and a defense mechanism against them. The purpose of the proposed attacks is to leak sensitive information between two logically separated (or isolated) networks that are hosted by a single router – one that is connected to the Internet, and another that is isolated and contains sensitive information. The attacks build on the fact that the response time of the router for a specific type of packet sent from a device that is connected to it is usually predictable, given the network topology. By interacting with the single shared router in a specific manner, an attacker can increase the router’s packet response time. The interaction is determined based on the information to be leaked, so the receiver (a computer located on the Internet-connected network) can measure the delayed packet response times and decode the sender’s signals (which operates from the isolated network) to receive classified information. The two classes of attacks presented in this work differ in the way that the delay is caused. In the cross-router covert channel (CRCC) attack, the sender overloads the router with control-plane packets; in the Wi-Fi micro-jamming attack, the sender uses a pre-installed implant to transmit single-tone signals in the 2.4GHz frequency range, disturbing the router’s packet transmissions. We showed that both attacks can influence a wide range of router brands and Wi-Fi capable devices and evaluated the optimal settings for both attacks when using different types of packets, transmission power, and data transmission rates.Our proposed defense mechanism is based on semi-supervised machine learning and deep learning algorithms, which are both used for novelty detection in network traffic. By detecting unusual traffic, we can identify the disturbances needed to leak the information. The system can then respond by blocking the suspicious devices that are involved in the attack. We evaluated the attacks in noisy and noise-free environments, successfully detecting both attacks in both environments. All the data and code, which includes the implementation of the attacks and the defense mechanism, is published as a benefit to the research community.

A semi-supervised adaptive Markov Gaussian embedding process (SAMGEP) for prediction of phenotype event times using the electronic health record

While there exist numerous methods to identify binary phenotypes (i.e. COPD) using electronic health record (EHR) data, few exist to ascertain the timings of phenotype events (i.e. COPD onset or exacerbations). Estimating event times could enable more powerful use of EHR data for longitudinal risk modeling, including survival analysis. Here we introduce Semi-supervised Adaptive Markov Gaussian Embedding Process (SAMGEP), a semi-supervised machine learning algorithm to estimate phenotype event times using EHR data with limited observed labels, which require resource-intensive chart review to obtain. SAMGEP models latent phenotype states as a binary Markov process, and it employs an adaptive weighting strategy to map timestamped EHR features to an embedding function that it models as a state-dependent Gaussian process. SAMGEP’s feature weighting achieves meaningful feature selection, and its predictions significantly improve AUCs and F1 scores over existing approaches in diverse simulations and real-world settings. It is particularly adept at predicting cumulative risk and event counting process functions, and is robust to diverse generative model parameters. Moreover, it achieves high accuracy with few (50–100) labels, efficiently leveraging unlabeled EHR data to maximize information gain from costly-to-obtain event time labels. SAMGEP can be used to estimate accurate phenotype state functions for risk modeling research.

Understanding the Roles of Video and Sensor Data in the Annotation of Human Activities

Human activities can be recognized in sensor data using supervised machine learning algorithms. In this approach, human annotators must annotate events in the sensor data which are used as input to supervised learning algorithms. Annotating events directly in time series graphs of data streams is difficult. Video is often collected and synchronized to the sensor data to aid human annotators in identifying events in the data. Other work in human activity recognition (HAR) minimizes the cost of annotation by using unsupervised or semi-supervised machine learning algorithms or using algorithms that are more tolerant of human annotation errors. Rather than adjusting algorithms, we focus on the performance of the human annotators themselves. Understanding how human annotators perform annotation may lead to annotation interfaces and data collection schemes that better support annotators. We investigate the accuracy and efficiency of human annotators in the context of four HAR tasks when using video, data, or both to annotate events. After a training period, we found that annotators were more efficient when using data alone on three of four tasks and more accurate when marking event types when using video alone on all four tasks. Annotators were more accurate when marking event boundaries using data alone on two tasks and more accurate using video alone on the other two tasks. Our results suggest that data and video collected for annotation of HAR tasks play different roles in the annotation process and these roles may vary with the HAR task.

Patterns of activity correlate with symptom severity in major depressive disorder patients

Objective measures, such as activity monitoring, can potentially complement clinical assessment for psychiatric patients. Alterations in rest–activity patterns are commonly encountered in patients with major depressive disorder. The aim of this study was to investigate whether features of activity patterns correlate with severity of depression symptoms (evaluated by Montgomery–Åsberg Rating Scale (MADRS) for depression). We used actigraphy recordings collected during ongoing major depressive episodes from patients not undergoing any antidepressant treatment. The recordings were acquired from two independent studies using different actigraphy systems. Data was quality-controlled and pre-processed for feature extraction following uniform procedures. We trained multiple regression models to predict MADRS score from features of activity patterns using brute-force and semi-supervised machine learning algorithms. The models were filtered based on the precision and the accuracy of fitting on training dataset before undergoing external validation on an independent dataset. The features enriched in the models surviving external validation point to high depressive symptom severity being associated with less complex activity patterns and stronger coupling to external circadian entrainers. Our results bring proof-of-concept evidence that activity patterns correlate with severity of depressive symptoms and suggest that actigraphy recordings may be a useful tool for individual evaluation of patients with major depressive disorder.

Investigation of Gaussian mixture clustering model for online diagnosis of tip-wear in nanomachining

This paper studied an unsupervised machine learning method to achieve real-time diagnosis of tip wear in tip-based nanomachining with an Atomic Force Microscope (AFM). The unsupervised Gaussian Mixture Model (GMM) was applied for online pattern recognition using the collected process data. In this work, the time varying signal of machining force was collected and managed in the form of moving data windows. The characteristics of the nanomachining force signal, including the maximum force, peak to peak force, and the variance, were calculated as the data of extracted features to indicate the tip wear conditions. Outliers were identified and removed using the Mahalanobis Distance Detection. A GMM model was trained using the historical processing data, in which the tip wear states were identified by clustering the feature parameters of each data frame. The change of the tip conditions was detected and tracked through the changing numbers of the tip failure points in the data windows. Experiment results showed that the highest and the average recognition accuracy of the proposed model was 1 and 0.8367 with the longest and the average calculation time of 0.031 s and 0.025 s. Compared with supervised and semi-supervised machine learning algorithms, the GMM based on unsupervised machine learning model has slightly lower recognition accuracy, but much shorter calculation time, which is suitable for online automatic diagnosis of tool wear in large-scale nanomanufacturing in real time.

Affinity Propagation-Based Hybrid Personalized Recommender System

A personalized recommender system is broadly accepted as a helpful tool to handle the information overload issue while recommending a related piece of information. This work proposes a hybrid personalized recommender system based on affinity propagation (AP), namely, APHPRS. Affinity propagation is a semisupervised machine learning algorithm used to cluster items based on similarities among them. In our approach, we first calculate the cluster quality and density and then combine their outputs to generate a new ranking score among clusters for the personalized recommendation. In the first phase, user preferences are collected and normalized as items rating matrix. This generated matrix is then clustered offline using affinity propagation and kept in a database for future recommendations. In the second phase, online recommendations are generated by applying the offline model. Negative Euclidian similarity and the quality of clusters are used together to select the best clusters for recommendations. The proposed APHPRS system alleviates problems such as sparsity and cold-start problems. The use of affinity propagation and the hybrid recommendation technique used in the proposed approach helps in improving results against sparsity. Experiments reveal that the proposed APHPRS performs better than most of the existing recommender systems.

Perfusion heterogeneity of cerebral small vessel disease revealed via arterial spin labeling MRI and machine learning

Cerebral small vessel disease (CSVD) is associated with altered cerebral perfusion. However, global and regional cerebral blood flow (CBF) are highly heterogeneous across CSVD patients. The aim of this study was to identify subtypes of CSVD with different CBF patterns using an advanced machine learning approach. 121 CSVD patients and 53 healthy controls received arterial spin label MRI, T1 structural MRI and clinical measurements. Regional CBF were used to identify distinct perfusion subtypes of CSVD via a semi-supervised machine learning algorithm. Statistical analyses were used to explore alterations in CBF, clinical measures, gray and white matter volume between healthy controls and different subtypes of CSVD. Correlation analysis was used to assess the association between clinical measures and altered CBF in each CSVD subtype. Three subtypes of CSVD with distinct CBF patterns were found. Subtype 1 showed decreased CBF in the temporal lobe and increased CBF in the parietal and occipital lobe. Subtype 2 exhibited decreased CBF in the right hemisphere of the brain, and increased CBF in the left cerebrum. Subtype 3 demonstrated decreased CBF in the posterior part of the brain, and increased CBF in anterior part of the brain. The three subtypes also differed significantly in gender (p = 0.005), the proportion of subjects with lacune (p = 0.002), with periventricular white matter hyperintensity (p = 0.043), and CSVD burden score (p = 0.048). In subtype 3, it was found that widespread decreased CBF was correlated with total CSVD burden score (r = -0.324, p = 0.029). Compared with healthy controls, the three CSVD subtypes also showed distinct volumetric patterns of white matter. The current results associate different subtypes with different clinical and imaging phenotypes, which can improve the understanding of brain perfusion alterations of CSVD and can facilitate precision diagnosis of CSVD.

A Deep Semi-Supervised Machine Learning Algorithm for Detecting Transportation Modes Based on Gps Tracking Data

A Deep Semi-Supervised Machine Learning Algorithm for Detecting Transportation Modes Based on Gps Tracking Data

A Comparison on Supervised and Semi-Supervised Machine Learning Classifiers for Gestational Diabetes Prediction

Abstract: Diabetes is the deadliest chronic diseases in the world. According to World Health Organization (WHO) around 422 million people are currently suffering from diabetes, particularly in low and middle-income countries. Also, the number of deaths due to diabetes is close to 1.6 million. Recent research has proven that the occurrence of diabetes is likely to be seen in people aged between 18 and this has risen from 4.7 to 8.5% from 1980 to 2014. Early diagnosis is necessary so that the disease does not go into advanced stages which is quite difficult to cure. Significant research has been performed in diabetes predictions. As time passes, challenges keep increasing to build a system to detect diabetes systematically. The hype for Machine Learning is increasing day to day to analyse medical data to diagnose a disease. Previous research has focused on just identifying the diabetes without specifying its type. In this paper, we have we have predicted gestational diabetes (Type-3) by comparing various supervised and semi-supervised machine learning algorithms on two datasets i.e., binned and non-binned datasets and compared the performance based on evaluation metrics. Keywords: Gestational diabetes, Machine Learning, Supervised Learning, Semi-Supervised Learning, Diabetes Prediction

A Comparative Analysis of Semi-Supervised Learning in Detecting Burst Header Packet Flooding Attack in Optical Burst Switching Network

This paper presents a comparative analysis of four semi-supervised machine learning (SSML) algorithms for detecting malicious nodes in an optical burst switching (OBS) network. The SSML approaches include a modified version of K-means clustering, a Gaussian mixture model (GMM), a classical self-training (ST) model, and a modified version of self-training (MST) model. All the four approaches work in semi-supervised fashion, while the MST uses an ensemble of classifiers for the final decision making. SSML approaches are particularly useful when a limited number of labeled data is available for training and validation of the classification model. Manual labeling of a large dataset is complex and time consuming. It is even worse for the OBS network data. SSML can be used to leverage the unlabeled data for making a better prediction than using a smaller set of labelled data. We evaluated the performance of four SSML approaches for two (Behaving, Not-behaving), three (Behaving, Not-behaving, and Potentially Not-behaving), and four (No-Block, Block, NB- wait and NB-No-Block) class classifications using precision, recall, and F1 score. In case of the two-class classification, the K-means and GMM-based approaches performed better than the others. In case of the three-class classification, the K-means and the classical ST approaches performed better than the others. In case of the four-class classification, the MST showed the best performance. Finally, the SSML approaches were compared with two supervised learning (SL) based approaches. The comparison results showed that the SSML based approaches outperform when a smaller sized labeled data is available to train the classification models.