Below you will find all conference submissions sorted by their session. Each submission listing will have a submission name, a list of authors, the submission’s abstract, and a link to the full text.
To download the full collection of proceedings as a ZIP file, click here.
Note: All papers are made available here with permission from the authors. The authors retain copyright of their work. DCSI does not require the authors to transfer copyright, nor do we publish these papers in any location other than this website.
LIT1: Lightning Talks
Internet of Things based Feedback Control System for Pediatric Pain Assessment and Management
Nupur Gaikwad, Hiroyouki Ohno, and Srinivas Sampalli
A child’s laughter radiates happiness in the most natural and innocent way. However, unlike just smiling to express joy, until at least six years of age children find it very challenging to comprehend and narrate the characteristics of their pain (its nature, location, and type). Thus, it is essential for parents, health care providers and nurses to identify this pain via the child’s behavior like facial expressions or cry pattern, and provide these tiny bundles of joy with adequate comfort. Motivated to support these little ones during these moments while providing automated assistance to nurses and even parents, we decided to explore the applications and role of Internet of Things (IoT) in the monitoring, assessment, and management of pediatric pain, and propose an IoT-based feedback control system that can assess a child’s pain level by analysing their facial expressions and cry pattern, and help reduce it by distracting the child with music or videos as well as alerting the parents and doctors.
A Blockchain Based Framework for Reputation Management and Node Misbehavior Detection in Wireless Sensor Networks
Kartik Bhatia and Srinivas Sampalli
With the growth of smart applications such as smart cities and smart farming, the importance of Wireless Sensor Networks (WSNs) is gradually being realized by many industrial enterprises. In particular, WSNs have shown enormous potential for being an interesting research area and in this decade, it is expected to grow manifold both in terms of applications as well as business revenues. WSNs consist of resource constrained devices which are present in an open and unsecured environment, and this makes them vulnerable to both internal as well as external attacks. Internal attacks can affect the network’s performance by increased energy consumption and introducing transmission delays. Consequently, this represents a critical security challenge for the deployment of WSNs. Many researchers have proposed solutions based on trust management systems that proves to be an efficient way for detecting such attacks by enhancing trust relationships and data routing reliability. In this paper, we extend the trust management system to include a distributed consensus mechanism based on blockchain which validates data packets originating from various source nodes. Additionally, a new algorithm is developed to estimate a node’s reputation based on its historical energy consumption data. Reputation and trust are both crucial factors that characterize malicious behaviour in the network. We have evaluated our proposed work with another existing trust model named Beliefbased Trust Evaluation Mechanism (BTEM) and compared our results in terms of performance metrics after performing various simulation runs. The results show that there is a significant improvement in the detection rate and accuracy. Furthermore, we have shown that our framework fulfils important security requirements such as integrity, authenticity and confidentiality by analyzing it for various security attacks.
An Autoencoder Model of Bathymetry and Multibeam Echosound Backscatter
Shakhboz Abdulazizov, Thomas Trappenberg, and Scott Lowe
Benthic habitat mapping is crucial to monitor and understand ongoing changes to the ocean environment caused by humanity and preserve fragile ocean ecosystems. Benthic habitats can be identified precisely from underwater images, but these images can not be collected at sufficient scale to build a habitat map on their own. Meanwhile, large-scale surveys can be conducted with multi-beam echosound that can collect both bathymetry and backscatter data, however, this data is hard for humans to interpret. Consequently, benthic habitat maps are currently using simple linear models to classify habitats using hand-picked features from the echosound data. We aim to improve benthic habitat mapping by training a model to classify the habitat from the underlying backscatter and bathymetry maps. Towards this end, in this paper, we show an autoencoder model of both bathymetry and backscatter that can extract high-level features from the echosound data. In the future, this model will be used for habitat classification.
A Survey of Security in SCADA networks: Current Issues and Future Challenges
Sagarika Ghosh and Srinivas Sampalli
Supervisory Control and Data Acquisition (SCADA) systems are used for monitoring industrial devices. However, their security faces the threat of being compromised due to the increasing use of open access networks. To secure the communication between nodes of SCADA networks, various security standards have been developed by different organizations. Researchers have proposed various security schemes to overcome the weaknesses of SCADA standards. The primary objective of this survey paper is to provide a study of the impact of possible attacks on SCADA systems. The paper addresses the future challenges that SCADA networks may face from quantum attacks. Furthermore, it outlines directions for further research in the field.
LIT2: Lightning Talks
IOMapper: The Integration of Generalized Signal Control Mapping With the Godot Game Engine
Logan Murphy and Joseph Malloch
We present a novel software solution for dynamic control of properties in a simulated environment using custom signals and devices, enabling thorough creative control for contemporary media synthesis. The software tool is IOMapper, a plugin for the popular open-source Godot game engine, serving as an implementation of the libmapper software library. The portable nature of libmapper’s network signals offers a basis upon which to easily integrate experimental input devices with Godot, creating a solid foundation for both formal study as well as the creation of art. IOMapper is a working tool, available publicly at https://github.com/lemurph/IOMapper.
IoT Device Fingerprinting in Commodity Switches
Pulkit Garg, Israat Haque, and Miguel Neves
The number of IoT devices and the concept of smart homes have become really prevalent these days. This paper presents FingerP4, a stateful solution that used P4 programming language to uniquely identify IoT devices inside a smart home in a BMv2 switch. FingerP4 uses packet lengths, the direction of flow, and the state of the packet defined by a Finite State Machine (FSM) to identify the devices. In our initial experiments, FingerP4 was successfully able to identify events from 7 different IoT devices entirely in the dataplane.
Stress and Anxiety Management Among Working-Class Indian Women
Jaisheen Kour Reen and Rita Orji
This extended abstract presents work in progress on the topic of managing stress and anxiety among Indian working women. There are many applications focused on managing stress and anxiety, but very few applications are there that follow the “user-centric design approach” that aims solely at the mental health of women. The research aims to apply the user-centric design (UCD) approach to design and develop personalized stress and anxiety applications for our target population. The study involves interviewing working-class Indian women and providing the most evidence-based interventions in the application.
An Intrusion Detection System for Internet of Medical Things
Deborah Oladimeji, Srinivas Sampalli, and Hiroyuki Ohno
The terms IoMT (Internet of Medical Things), IoHT (Internet of Health Things) and HIoT (Healthcare Internet of Things) are now all used interchangeably. They describe the connection of medical devices and software applications relating to healthcare information to the Internet using networking technologies. While these technologies bring the promise of improved patient care, improved efficiency, and reduced costs, they also bring new risks as many these connected devices are unmanaged and unprotected. The consequent potential impact is not just on patient data, but on patient care itself.
This thesis focuses on providing a highly secure transmission of medical data in IoMT to ensure accuracy and confidentiality of patients’ data. We propose a novel intrusion detection system (IDS) based on machine learning (ML) methods which uses both network and biometric parameters as features and can differentiate the normal traffic from attack traffic. Six ML methods were selected for the intrusion detection, namely, Random Forest, K-Nearest Neighbor, Support Vector Machine, Artificial Neural Networks, J48 and Decision Table, and tested against man-in-the-middle and denial of service attacks using a dataset consisting of a combination of about 20,000 normal and attack healthcare data. The dataset was generated on our IoMT test bed that was implemented using four modules, namely, a multi-sensor board, a gateway module, a network module, and a visualization module. The communication between the modules employs a Client Server publish/subscribe messaging transport protocol, MQTT, which is a light weight, simple, easy to implement for constrained devices with limited resources, such as IoMT.
Experimental results indicate that our secured healthcare system can detect anomalies in both the network flow and patient’s biometric readings. Furthermore, we generated a new healthcare dataset with the combination of biometric data and network traffic available for other researchers for statistical analysis and further research. Finally, we present a comparative summary of the proposed scheme with an existing scheme in terms of accuracy and execution time.
Enhanced IoT Network Communications Using Multi-PHY 6TiSCH
Chloe Bae, Israat Haque, and Michael Baddeley
Recently, single-radio, multi-protocol wireless chips are introduced to the market. These chips are equipped with multiple physical layer protocols and can switch between different physical layers (PHYs) on demand. Traditionally, IoT devices operate on a single physical layer protocol (e.g., IEEE 802.15.4) and have predetermined performance and limitations. Thus, the multi-protocol wireless chips can enable multi-PHY IoT wireless communications, especially in Low-Power and Lossy Networks (LLNs). A device can then switch between the PHYs to use the most suitable one to satisfy the application demand, e.g., offering a high throughput or a long-range. This study will conduct the detailed performance evaluation and examine the potential benefits of multi-PHYs on the 6TiSCH stack in Industrial IoT wireless network communications, so users choose appropriate protocols to meet their application demand.
LIT3: Lightning Talks
Improving antimicrobial resistance surveillance and diagnostics with machine learning predictions from genomic data
Jee Kim and Rob Beiko
Antimicrobial resistance (AMR) refers to the phenomenon where antibiotic drugs can no longer inhibit the growth of bacteria. AMR is rapidly increasing worldwide, with grave consequences for our ability to treat infectious diseases. Multidrug-resistant pathogens (or ‘superbugs’) are rapidly emerging in healthcare and agriculture settings due to the high volume of antibiotic use and misuse that pressures microbes to quickly develop defence mechanisms. With the development of rapid and affordable genome sequencing, we can study the complete set of genes in an organism, including genes conferring AMR. However, it is still challenging to predict an organisms’ resistance behaviour based on the AMR genes alone. This investigation uses machine-learning (ML) tools to predict how a pathogen will respond to antibiotic treatment. The performance of ML models with 300+ Enterococcus faecium genomes as the learning data demonstrated maximum accuracy of 98% when predicting resistance against antibiotics like vancomycin. Other genomic data encoding methods and feature selection demonstrated that ML can accurately predict organisms’ resistance to specific antibiotics and appropriately select highly relevant features that contribute to resistance. The newly established genetic features that have no previous connection to resistance will be investigated in the laboratory to confirm their contribution to bacteria’s resistance behaviour. The potential research findings will hopefully assist in the early prevention and mitigation effort of AMR.
Novel Approaches to Marker Gene Representation Learning Using Trained Tokenizers and Jointly Trained Transformer Models
Alexander Manuele and Robert Beiko
Next generation DNA sequencing technologies have made DNA sequence data far more widely available, opening new avenues of research. Analysis of marker gene data has many short-comings, including sparsity, high cardinality, and intra-study dependencies during feature engineering. We present two novel approaches to feature representation of DNA marker gene data, first showing that trained tokenizers can replace traditional sliding-window based segmentation techniques, then proposing a training scheme to learn dense vector representations of DNA sequences using transformer language models. We demonstrate that our representations match or exceed previously published approaches while providing fixed-length, low cardinality representations
Ananke 2: Memory-efficient, progressive clustering of large microbial data sets
Michael Hall and Robert Beiko
As data sets grow and combine and the capacity to collect data increases, resource efficient algorithms are becoming more necessary for handling microbial ecology data. Clustering, exploring, and visualizing data becomes difficult without the aid of significant computing resources. In this paper, we present Ananke 2, an algorithm for dynamically and progressively clustering large sets of microbial features. Our algorithm uses bloom filters to store the relationships between microbial features (taxonomic unit counts, amplicon counts, functional gene counts, etc.). These succinct, probabilistic data structures allow the network structure of the microbial features to be stored in a memory-efficient way, enabling the clustering and exploration of very large feature sets. Ananke 2 is particularly suited to clustering time-series data, and supports time-series specific distance measures such as short time-series distance and dynamic time warping.
Toward Semi-Supervised Classification of Underwater Benthic Habitat Imagery
Isaac Xu, Thomas Trappenberg, and Scott Lowe
As part of the Benthic Ecosystem Mapping and Engagement (BEcoME) project, we are working toward automating underwater image classification using modern semisupervised learning approaches. To begin this work, we tested the Bootstrap Your Own Latent (BYOL) model on the well known MNIST dataset. Our interest is in datasets where only a minority of the samples are labelled, a problem typical for many real world datasets, which we simulated by redacting the labels from part of the MNIST dataset. We find that the semi-supervised methodology is more resilient against a decrease in the number of labelled training samples than a fully-supervised model trained only on the labelled images. When very few of the samples were labelled (<0.8%), there is an appreciable performance advantage to the semi-supervised models when compared with a purely supervised model.
“From Kilobytes to Kilodaltons”: A Novel Algorithm for Medical Image Encryption based on the Central Dogma of Molecular Biology
Arjav Gupta and Srinivas Sampalli
With the continued integration of technology in medicine, large amounts of patient data are often vulnerable to cyber-attacks. Medical data must be secured, however traditional cryptography algorithms are inapplicable to medical images. To address the need for new medical image encryption algorithms, a novel approach based on the central dogma of molecular biology is proposed. The resulting algorithm has a complexity of O(n) and is resistant to brute force, differential and statistical attacks. The algorithm meets the standards of literature in DNA-based image encryption and surpasses recent approaches in medical image encryption in its defense against cyber-attacks.
FULL1: Full Talks
Authorship Identification of Source Code Segments Written by Multiple Authors Using Stacking Ensemble Method
Parvez Mahbub, Naz Zarreen Oishie, and Rafizul Haque
Source code segment authorship identification is the task of identifying the author of a source code segment through supervised learning. It has vast importance in plagiarism detection, digital forensics, and several other law enforcement issues. However, when a source code segment is written by multiple authors, typical author identification methods no longer work. Here, an author identification technique, capable of predicting the authorship of source code segments, even in case of multiple authors, has been proposed which uses stacking ensemble classifier. This proposed technique is built upon several deep neural networks, random forests and support vector machine classifiers. It has been shown that for identifying the author-group, a single classification technique is no longer sufficient and using a deep neural network based stacking ensemble method can enhance the accuracy significantly. Performance of the proposed technique has been compared with some existing methods which only deal with the source code segments written exactly by a single author. Despite the harder task of authorship identification for source code segments written by multiple authors, our proposed technique has achieved promising results evident by the identification accuracy, compared to the related works which only deal with code segments written by a single author. This work is previously published on International Conference on Computer and Information Technology (ICCIT) 2019.
Analyzing the Impact of Topology on Flooding Attacks in Low-power IoT Networks
Jack Zhao, Xinyu Liu, Michael Baddeley, and Israat Haque
Low-power Internet of things (IoT) networks can support various applications like smart agriculture or smart manufacturing. These devices usually rely on the commonly-used routing protocol for low-power and lossy networks (RPL) protocol to exchange messages. RPL suffers from various attacks, where the DODAG Information Solicitation (DIS) flooding attack is the most common but effective attacking method. To figure out the different factors in this attack, we analyze how the number of attackers and their location from DODAG root can influence the DIS flooding attack as such systematic analysis is missing in existing works. Extensive evaluation over Contiki-NG-based Cooja simulator reveals that attackers’ number significantly damages devices’ energy consumption and packet delivery ratio than the position of attackers.
Implementation and Optimisations for Computing Maximum Agreement Forests for Rooted Multifurcating Trees
Ben Lee and Christopher Whidden
We implement and optimise a fixed parameter tractable (FPT) algorithm proposed by Whidden et al. (Al- gorithmica, 2016) for computing maximum agreement forests (MAFs) for pairs of multifurcating trees. The sizes of MAFs give the subtree-prune-and-regraft (SPR) distance and has uses in determining when and how often lateral gene transfer (LGT) takes place in phylogenetic trees. We show the running time based on synthetic data and confirm the optimisations speed up the algorithm. The main motive for this algorithm is to study LGT in antimicrobial resistant bacteria.
ArtBeat – Deep Convolutional Networks for Emotional Inference to Enhance Art with Music
Liam Hebert, Elizabeth Eddy, Will Harrington, Lauryn Marchand, Jason d’Eon, and Sageev Oore
Paintings and music are two universal forms of art that are present across all cultures and times in human history. In this paper, we present ArtBeat, a machine learning application to connect the two. Not only are these two art forms universal, but they are also deeply emotionally charged. That emotional factor is what we use as a bridge between these mediums. Using a convolutional neural network, we aimed to create a model that can classify the emotions evoked by a painting, and use the predicted values to pair it with a piece of music to complement the viewing experience. Our system uses a pretrained Wide ResNet model as a base, which we then fine-tuned to fit this project’s needs. In this paper, we describe the process of designing and implementing this model as well as report its results and analyze its behaviour. Source code is available at https://git.cs.dal.ca/hebert/artbeat
Due to right restrictions, this paper is only accessible to faculty and staff of Dalhousie University. Access will be given by request only.
Real Valued Actions In Tangled Program Graphs
Ryan Amaral, Caleidgh Bayer, Alexandru Ianta, Robert Smith, and Malcolm Heywood
Tangled Program Graphs (TPGs), a modular genetic programming algorithm, has been shown in the past to perform well in reinforcement learning environments which assume discrete actions (e.g. Atari console games). It is the goal of this research to expand the capability of TPG into the continuous control domain, where real valued actions are required, as well as to showcase the potential of using real valued action generation to aid in discrete action tasks. In addition to giving TPG real value capabilities, we also explore the impact of diversity maintenance through occasionally introducing new genetic material for TPG to work with during evolution, as well as a method of “speeding up” evolution through repeated mutations with a process called rampancy. With this in mind, a 2D bipedal walker control task will be assumed in which multiple real-valued control actions have to be specified per state, as well as 3D tasks through ViZDoom which accepts discrete actions.
Measuring Genetic Heterogeneity in Psychiatric Phenotypes: A Scoping Review
Harvey Wang, Martin Alda, Thomas Trappenberg, and Abraham Nunes
An improved understanding of genetic etiological heterogeneity in a psychiatric condition may help us (A) isolate a neurophysiological “final common pathway” by identifying its upstream genetic origins and (B) facilitate characterization of the condition’s phenotypic variation. This review aims to characterize existing genetic heterogeneity measurements in the psychiatric literature. The Scopus database was searched for studies that quantified genetic heterogeneity or correlation of psychiatric phenotypes with human genetic data. Ninety studies were included. Eighty-seven reports quantified genetic correlation, five applied genomic structural equation modelling, three evaluated departure from the Hardy-Weinberg equilibrium at one or more loci, and two applied a novel approach known as MiXeR. We found no study that rigorously measured genetic etiological heterogeneity across a large number of markers. Developing such approaches may help better characterize the biological diversity of psychopathology.
FULL2: Full Talks
Computing Matching Statistics on Repetitive Texts
Younan Gao
Computing the matching statistics of a string P[1..m] with respect to a text T[1..n] is a fundamental problem which has application to genome sequence comparison. δ, as a relevant compressibility measure for repetitive texts, was recently introduced by Kociumaka et al.. And they also proved that δ is an even smaller measure than γ, the smallest string attractor. In this paper, we study the problem of computing the matching statistics upon highly repetitive texts. We present that within O(δ lg (n/δ)) words of space, matching statistics can be computed in O(m2 lgε γ + m lg n) time, where ε is an arbitrarily small positive constant.
Design and Implementation of Raspberry House: An IoT Security Framework
Wen Fei, Hiroyuki Ohno, and Srinivas Sampalli
The rising popularity of the Internet of Things (IoT) on a global scale has led to an increase in cyber threats, and researchers are paying more attention to its security issues. So far, research on IoT security has focused on large-scale devices, but there is relatively less research on the security of small IoT devices. Therefore, our objective is to mainly study how to make the operation of small IoT devices safer. Raspberry House is a TCP/IP Layer 3 gateway built with Raspberry Pi, which can connect IoT devices to the private network generated by it, thereby preventing IoT devices from being exposed to outside networks. In addition, through a private network, IoT devices can also update their firmware wirelessly. This paper also studies the communication between IoT devices through different secure connections such as Secure Shell (SSH), and evaluates their results in different environments. Experimental evaluation of TCP/IP Layer 3 Gateway indicates that the proposed framework can provide security for small IoT devices. This work is previously published on 2020 IEEE International Conference on Internet of Things and Intelligence System (IoTaIS).
Using Interactive Visual Analytics to Optimize Blood Products Inventory at a Blood Bank
Jaber Rad, Jason G Quinn, Calvino Cheng, Robert Liwski, Samina Abidi, and Syed Sibte Raza Abidi
Blood products and their derivatives are perishable commodities that require an efficient inventory management to ensure both a low wastage rate and a high product availability rate. To optimize blood product inventory, blood transfusion services need to reduce wastage by avoiding outdates and improve availability of different blood products. We used advanced visualization techniques to design and develop a highly interactive web-based dashboard to (1) monitor the blood product inventory and the on-going blood unit transactions in near-real-time, and (2) audit retrospective data to identify and learn from procedural inefficiencies based on analysis of transactional data. We present pertinent scenarios to show how the blood transfusion staff can use the dashboard to locate units with specific characteristics, investigate the lifecycle of the units, efficiently transfer units between facilities to minimize outdates, and probe blood product lifecycle patterns that led to discard to discover inefficiencies in the Blood Transfusion Services (BTS).
Estimating Severity of Depression from Acoustic Features and Embeddings of Natural Speech
Sri Harsha Dumpala, Sheri Rempel, Katerina Dikaios, Mehri Sajjadian, Rudolf Uher, and Sageev Oore
Major depressive disorder, referred to as depression, is a leading cause of disability, absence from work, and premature death. Automatic assessment of depression from speech is a critical step towards improving diagnosis and treatment of depression. Previous works on depression assessment from speech considered various acoustic features extracted from speech to estimate depression severity. But performance of these approaches is not at clinical standards, and thus requires further improvement. In this work, we examine two novel approaches for improving depression severity estimation from short audio recordings of speech. Specifically, in audio recordings of a narrative by individuals diagnosed with major depressive disorder, we analyze spectral-based and excitation source-based features extracted from speech, and significance of sentiment and emotion classification in estimation of depression severity. Initial results indicate synchrony between depression scores and the sentiment and emotion labels. We propose the use of sentiment and emotion based embeddings obtained using machine learning techniques in estimation of depression severity. We also propose use of multi-task training to better estimate depression severity. We show that the proposed approaches provide additive improvements in the estimation of depression severity.
The Spotlight: A General Method for Discovering Systematic Errors in Deep Learning Models
Jason d’Eon, Greg d’Eon, James R. Wright, and Kevin Leyton-Brown
Supervised learning models often make systematic errors on relatively rare subsets of the data. However, such performance problems can be difficult to identify: model performance can be broken down across sensitive groups, but only when these groups are known and explicitly labelled. This paper introduces a method for discovering systematic errors, which we call the spotlight. The key idea is that similar inputs tend to have similar representations in the final hidden layer of a neural network. We leverage this structure by “shining a spotlight” on this representation space to find contiguous regions where the model performs poorly. We show that the spotlight surfaces semantically meaningful areas of weakness in a surprisingly wide variety of model architectures, including image classifiers, language models, and recommender systems.
Controlling BigGAN Image Generation with a Segmentation Network
Aman Jaiswal, Harpreet Singh Sodhi, Mohamed Muzamil, Rajveen Singh Chandhok, Sageev Oore, and Chandramouli Shama Sastry
GANS have been used for a variety of unconditional and conditional generation tasks; while unconditional generation involves learning and sampling from P(X), conditional generation can be described as sampling from P(X|f(X)=1), where f is a binary indicator function. Most commonly studied conditional generation are class-conditional generation wherein f is a binary class-membership function. While class-conditional generation can be directly integrated into the training process, integrating more sophisticated indicator functions within the training is not as straightforward. In this work , we consider the task of sampling from P(X) such that the silhouette of (the subject of) X matches the silhouette of (the subject of) a given image; that is, we not only specify what to generate, but we also control where to put it: more generally, we allow a mask (this is actually another image) to control the silhouette of the object to be generated. The mask is itself the result of a segmentation system applied to a user-provided image. To achieve this, we use pre-trained BigGAN and SOTA segmentation models (e.g. DeepLabV3 and FCN) as follows: we first sample a random latent vector z from the Gaussian Prior of BigGAN and then iteratively modify the latent vector until the silhouettes of X=G(z) and the reference image match. While the BigGAN is a class-conditional generative model trained on the 1000 classes of ImageNet, the segmentation models are trained on the 20 classes of the PASCAL VOC dataset; we choose the “Dog” and the “Cat” classes to demonstrate our controlled generation model.
FULL3: Full Talks
Intrusion Detection in SCADA-based Power Grids: Feature Selection using Gradient Boosting Scoring Model with Decision Tree Classifiers
Darshana Upadhyay, Jaume Manero, Marzia Zaman, and Srinivas Sampalli
Smart grids rely on SCADA (Supervisory Control and Data Acquisition) systems to monitor and control complex electrical networks in order to provide reliable energy to homes and industries. However, the increased inter-connectivity and remote accessibility of SCADA systems expose them to cyber attacks. As a consequence, developing effective security mechanisms is a priority in order to protect the network from internal and external attacks. We propose an integrated framework for an Intrusion Detection System (IDS) for smart grids which combines feature engineering-based preprocessing with machine learning classifiers. Whilst most of the machine learning techniques fine-tune the hyper-parameters to improve the detection rate, our approach focuses on selecting the most promising features of the dataset using Gradient Boosting Feature Selection (GBFS) before applying the classification algorithm, a combination which improves not only the detection rate but also the execution speed. GBFS uses the Weighted Feature Importance (WFI) extraction technique to reduce the complexity of classifiers. We implement and evaluate various decision tree-based machine learning techniques after obtaining the most promising features of the power grid dataset through a GBFS module, and show that this approach optimizes the False Positive Rate (FPR) and the execution time.
E-Prescription Systems A Comparative Survey
Bader Aldughayfiq and Srinivas Sampalli
Medication errors related to prescriptions are among the most significant risks facing the health care sector. Many countries worldwide have implemented ePrescription systems to reduce medication errors. Further, more research is needed to evaluate these systems and their effect on patients’ care services. Thus, we conducted a survey study involving eight countries implementing ePrescription systems. We found several challenges and limitations of the surveyed systems, such as information availability, information privacy and security, and new technologies adaptability. Therefore, these challenges need to be addressed to provide quality service, improve patients’ medication safety.