ABSTRACT In the early phases of a construction project, generating accurate and timely quotations is important for assessing feasibility. Delays or significant revisions in quotations can lead to project cancellations, resulting in lost business opportunities. To address this challenge, we propose a machine learning framework called Jointly Trained Automation of Explainable Construction Material Knowledge ( JACK ), developed in collaboration with a construction company to estimate material requirements. Our methodology begins with pre‐processing estimation data, where construction materials are categorised into high‐level types to facilitate more efficient learning. To support this process, open‐source synthetic data generators were developed to help clarify structural patterns for JACK , which employs a cascaded learning approach during training. The evaluation phase leverages joint training to enhance model efficiency and presents results across 207 construction projects. We also investigate the effects of dropout layers, regression trees and synthetic data augmentation on prediction accuracy. Finally, we compare JACK against traditional regression‐based methods using a separate project set, where it demonstrates competitive performance. Overall, JACK achieves low error rates across a range of material types, with performance gains largely attributed to the benefits of cascaded learning.

In computer vision, most data are captured in 2D formats, limiting spatial understanding in real-world applications. This presents a challenge for fields such as architecture, construction, and robotics, where interpreting spatial relationships from minimal visual input is increasingly essential. This survey reviews recent advancements in extracting 3D features from 2D imagery, a critical task in these domains, where spatial accuracy and object orientation directly impact performance. We focus on three core areas: (1) disposition estimation, determining object pose; (2) joint modeling, constructing skeletal representations; and (3) scene reconstruction, generating spatially accurate environments. Each category is evaluated based on input modalities, performance metrics, and code availability. By providing a unified overview of these techniques, this paper highlights their practical value in enabling 3D reasoning from conventional 2D data.

In the construction industry, a project typically begins with the creation of two-dimensional (2D) building plans, defining the client’s specifications. Using these plans, a digital three-dimensional (3D) model is developed to visualize the anticipated outcome and to verify the model’s alignment with the client’s expectations. The process of converting from 2D to 3D can become time-intensive if there is a need for modifications or if the project’s overall complexity is high. To enhance efficiency and accuracy, this research introduces an end-to-end framework referred to as BIRD which stands for B uilding I mage R econstruction and D imensioning. BIRD is capable of accepting five 2D perspective drawings of a building as inputs and generating a proportionate 3D model of the building envelope as an output. This is accomplished through the integration of multiple techniques that use convolutional neural networks to extract a refined set of line segments, identify measurements, and align each perspective with the floor plan drawing. The key contributions of this study includes: (1) a novel deep learning model designed for the identification of line segments in building plans; (2) novel algorithms that facilitate the generation of information required for 3D modeling; (3) an end-to-end framework for building reconstruction; and (4) novel performance metrics specifically tailored for the 2D to 3D conversion challenge. The practical application of this research was validated through the use of complete building plans provided by an industry partner. In summary, it was observed that BIRD demonstrated high accuracy in the creation of 3D visualizations, highlighting its real-world efficacy. • A framework is proposed to convert 2D building plans into 3D envelope models. • A novel dual-branch approach is used to extract line segments from building plans. • A novel algorithmic approach works to wrap line segments in three-dimensional spaces. • The framework is ran on building plans from an industry partner with high accuracy. • A preliminary user interface is described to highlight the practical use in industry.

The social NLP researchers and mental health practitioners have witnessed exponential growth in the field of mental health detection and analysis on social media. It has become important to identify the reason behind mental illness. In this context, we introduce a new dataset for Causal Analysis of Mental health in Social media posts (CAMS). We first introduce the annotation schema for this task of causal analysis. The causal analysis comprises of two types of annotations, viz, causal interpretation and causal categorization. We show the efficacy of our scheme in two ways: (i) crawling and annotating 3155 Reddit data and (ii) re-annotate the publicly available SDCNL dataset of 1896 instances for interpretable causal analysis. We further combine them as CAMS dataset and make it available along with the other source codes https://anonymous.4open.science/r/CAMS1/. Our experimental results show that the hybrid CNN-LSTM model gives the best performance over CAMS dataset.

Significant research has been done in the medical domain using machine learning and clinical data sets. Although there are many interesting and influential clinical research works in the fields of healthcare and health services using machine learning, there is a need to apply machine learning in the field of health human resource planning. This study uses physician billing data and machine learning to identify and classify family physicians with the goal of improving health human resource planning. This research is essential for policy makers because it is important to know the number of family physicians practicing in certain geographical regions for providing timely care. Additionally, this issue becomes particularly important when it comes to serving communities with fewer resources such as the rural areas of Northwestern Ontario, where family physicians need to work to their full scope of practice, provide more services than physicians working in urban areas, to meet the needs of patients. In this study, recursive feature elimination method is used to reduce the number of predictors for the classification problems. As the result of this process, the most important features include physician’s rurality, full-time equivalent hours, age, and years of experience. Further, several machine learning models are used to solve binary and multi-class classification problems. Gradient boosting machine learning was the most accurate in predicting family physician practice, with a receiver operating characteristic value, ROC value, of 0.73 and 0.72 for binary and multi-class classification, respectively.

In recent days, with increased population and traffic on roadways, vehicle collision is one of the leading causes of death worldwide. The automotive industry is motivated on developing techniques to use sensors and advancements in the field of computer vision to build collision detection and collision prevention systems to assist drivers. In this article, a deep-learning-based model comprising of ResNext architecture with SENet blocks is proposed. The performance of the model is compared to popular deep learning models like VGG16, VGG19, Resnet50, and stand-alone ResNext. The proposed model outperforms the existing baseline models achieving a ROC-AUC of 0.91 using a significantly less proportion of the GTACrash synthetic data for training, thus reducing the computational overhead.

Online businesses are highly interested in finding practical solutions to opinion mining, but it is challenging to extract aspects and sentiments from the text. One way to solve this problem is to fine-tune good quality extractions from reviews using state-of-the-art pre-trained language models. However, such fine-tuned language models can produce good results if trained with a large amount of relevant data. In this paper, a technique that can fine-tune language models for opinion extractions using unlabelled training data. This paper proposes a novel opinion mining system called ONSET. This system is developed through a fine-tuned language model using an unsupervised learning approach to label aspects using topic modeling and then using semi-supervised learning with data augmentation. With extensive experiments performed during this research, the proposed model can achieve similar results as some state-of-the-art models produce with a high quantity of labelled training data. F1-scores of 87.30% and 88.35% are achieved on SemEval Aspect-Based Sentiment Analysis and Twitter datasets, respectively.

BACKGROUND: Twitter is a popular social networking site where short messages or "tweets" of users have been used extensively for research purposes. However, not much research has been done in mining the medical professions, such as detecting the occupations of users from their biographical contents. Mining such professions can be used to build efficient recommender systems for cost-effective targeted advertisements. Moreover, it is highly important to develop effective methods to identify the occupation of users since conventional classification methods rely on features developed by human intelligence. Although, the result may be favorable for the classification problem. However, it is still extremely challenging for traditional classifiers to predict the medical occupations accurately since it involves predicting multiple occupations. Hence this study emphasizes predicting the medical occupational class of users through their public biographical ("Bio") content. We have conducted our analysis by annotating the bio content of Twitter users. In this paper, we propose a method of combining word embedding with state-of-art neural network models that include: Long Short Term Memory (LSTM), Bidirectional LSTM, Gated Recurrent Unit, Bidirectional Encoder Representations from Transformers, and A lite BERT. Moreover, we have also observed that by composing the word embedding with the neural network models there is no need to construct any particular attribute or feature. By using word embedding, the bio contents are formatted as dense vectors which are fed as input into the neural network models as a sequence of vectors. RESULT: Performance metrics that include accuracy, precision, recall, and F1-score have shown a significant difference between our method of combining word embedding with neural network models than with the traditional methods. The scores have proved that our proposed approach has outperformed the traditional machine learning techniques for detecting medical occupations among users. ALBERT has performed the best among the deep learning networks with an F1 score of 0.90. CONCLUSION: In this study, we have presented a novel method of detecting the occupations of Twitter users engaged in the medical domain by merging word embedding with state-of-art neural networks. The outcomes of our approach have demonstrated that our method can further advance the process of analyzing corpora of social media without going through the trouble of developing computationally expensive features.

Twitter has withstood the test of time as a successful social networking platform. In many circles globally, the majority of users choose Twitter when choosing a social media outlet for reliable scientific information and news. However, the Twitter application programming interface (API) limitations do not allow for low-cost data science options for academia. It becomes very expensive for academic researchers to gain the full potential of data analytics available from Twitter using a free API account. In this article, we present our big data analytics platform developed at our DaTALab at Lakehead University, Canada, that allows users to focus on their Twitter search criteria and gain access to large amounts of Twitter data at the touch of a button. The platform supports the collection of social media data and applies many filters for cleaning and further use for machine learning (ML) and artificial intelligence (AI)-based systems. Our focus has been primarily on healthcare-related research, which shows the strength of the presented platform. However, the platform itself is malleable to any topic of interest. Data collected and processed are suitable for further AI/ML analysis. We present our platform using a specific healthcare search topic to emphasize the power of our system for future research endeavors in the healthcare field.

Despite studies of over six decades, research on automated essay scoring continues to grab ample attention in the Natural Language Processing (NLP) community in part because of its commercial and educational value. However, evaluating such writing compositions or essays in terms of reliability and time is a very challenging process. The need for reliable and rapid scores has elevated the need for a computer system that can answer essay questions that fit precise prompts automatically. NLP and machine learning strategies use Automated Essay Scoring (AES) systems to solve the difficulty of scoring writing tasks. In this paper, we suggest an AES approach that involves not only rule-based grammar and consistency tests, but also the semantic similarity of sentences, thus giving priority to question prompts. Similarity vectors are used obtained after applying semantic algorithms and calculated statistical features. Our system uses 22 features with high predicting power, which is less than current systems, while considering every aspect a human grader may focus on.Predicting scores is achieved using the data provided by Kaggle's ASAP competition using Random Forest. The resulting agreement between the score of the human grader and the prediction of the system is compared with promising results through experimental evaluation.

Homelessness is a complex social problem and there have been limited attempts to use machine learning algorithms to understand the various issues that public health agencies would like to solve. For instance, it is important for the policy makers to know where homeless populations live so that they can provide necessary services accordingly. This article presents a satellite image tent-detection solution with three deep learning methods that utilize transfer learning from the ResNetV2, InceptionV3, and MobileNetV2 models, trained on ImageNet, attached to a unique architecture referred to as "TentNet". The performance of these models are first shown in detecting planes and ships within satellite imagery in previously defined datasets as a baseline. Then, a new dataset is created from a compilation of tents from the xView project to use for testing, along with another dataset of synthetic images from the generative adversarial networks StyleGAN2 and DCGAN for training. After training on a dataset containing only synthetic images for the tents class, the ResNetV2 architecture achieved the highest accuracy of 73.68% when testing on the real satellite imagery.

Abstract As the popularity of the portable document format (PDF) file format increases, research that facilitates PDF text analysis or extraction is necessary. Heading detection is a crucial component of PDF‐based text classification processes. This research involves training a supervised learning model to detect headings by systematically testing and selecting classifier features using recursive feature elimination. Results indicate that decision tree is the best classifier with an accuracy of 95.83%, sensitivity of 0.981, and a specificity of 0.946. This research into heading detection contributes to the field of PDF‐based text extraction and can be applied to the automation of large scale PDF text analysis in a variety of professional and policy‐based contexts.

Nowadays, everyone is surrounded by a large volume of data which is generated from multiple sources. Every day 2.5 quintillion bytes of data are created worldwide. It is a very complicated task to manage and extract useful information from social media platforms. As people are more interested in seeking online health guidance, the medical textual data created by patients, healthcare professionals, and medical staff are expanding fast. In order to address this challenge, an ontology-based approach is used to annotate the medical domain data in this article. We have considered Twitter as our social media platform to filter the medical data from the general category data. In this article, we have used the Open Biomedical Annotator (OBA), which is an ontology-based Web service that annotates raw text with biomedical ontology concepts based on their textual metadata. Domain-specific knowledge is gathered by using MedlinePlus Health Topics and National Drug Data File (NDDF) ontologies licensed under the U. S. National Library of Medicine (UMLS), which will filter out all medical related terms. Furthermore, these medical terms are stored in a medical term database, which will then be forwarded to the UMLS REST Application Programming Interface (API) that provides important entities, such as Concept Unique Identifier (CUIs), semantic relations, attributes, definitions, and much more under the subset of SNOMEDCT_US which is the most precise and comprehensive health terminology platform in the world.

Social Media platforms have become common venue for sharing experiences and knowledge about health-related topics. This research focuses on examining social media based communication patterns related to diabetes on the Twitter platform. Specifically, we apply an updated methodology to examine changes in the current use of hash-tags, trending hash-tags, and the frequency of diabetes-related tweets using a previous study as a baseline. Our results show significant growth in the diabetes community on Twitter over time and also evidence that this community is increasing in it's capacity to spread awareness around diabetes related health topics. Our methodological contributions include an improved framework for collecting, cleaning and analyzing Twitter data related to diabetes as well as the application of regular expressions to categorize subsets of Tweets.

Communications on the popular social networking platform, Twitter, can be mapped in terms of a hashtag graph, where vertices correspond to hashtags, and edges correspond to co-occurrences of hashtags within the same distinct tweet. Furthermore, a vertex in hashtag graphs can be weighted with the number of tweets a hashtag has occurred in, and edges can be weighted with the number of tweets both hashtags have co-occurred in. In this paper, we describe additions to some well-known link prediction methods that allow the weights of both vertices and edges in a weighted hashtag graph to be taken into account. We base our novel predictive additions on the assumption that more popular hashtags have a higher probability to appear with other hashtags in the future. We then apply these improved methods to 3 sets of Twitter data with the intent of predicting hashtags co-occurences in the future. Experimental results on real-life data sets consisting of over 3,000,000 combined unique Tweets and over 250, 000 unique hashtags show the effectiveness of the proposed models and algorithms on weighted hashtag graphs.

Pregnancy is a special condition in which women go through various health complications throughout the period of gestation. Conventional health monitoring systems are either too specific or too general, therefore not flexible enough to be suited for pregnant women. In this paper, requirements and challenges in health monitoring of pregnant women are summarized. First, the specific requirements for health monitoring are derived, including adaptive monitoring, need for big data, and real time monitoring. Then a first ever reference architecture is proposed specifically for health monitoring of pregnant women. Mobile devices, body sensors, cloud and thin client of health care professionals are integrated together through the proposed architecture for an end to end self adaptive health monitoring solution design. Apart from self adaptation based on system dynamics, other features of the proposed architecture are mobile device as a gateway for body and ambient sensors, risk factor evaluation and possible action identification, and prolonged monitoring. The proposed architecture facilitates big data analytic and real time monitoring by remote thin client of health professionals through the assistance of cloud infrastructure.

World wide road traffic fatality and accident rates are high, and this is true even in technologically advanced countries like the USA. Despite the advances in Intelligent Transportation Systems, safe transportation routing i.e., finding safest routes is largely an overlooked paradigm. In recent years, large amount of traffic data has been produced by people, Internet of Vehicles and Internet of Things (IoT). Also, thanks to advances in cloud computing and proliferation of mobile communication technologies, it is now possible to perform analysis on vast amount of generated data (crowd sourced) and deliver the result back to users in real time. This paper proposes SafeRNet, a safe route computation framework which takes advantage of these technologies to analyze streaming traffic data and historical data to effectively infer safe routes and deliver them back to users in real time. SafeRNet utilizes Bayesian network to formulate safe route model. Furthermore, a case study is presented to demonstrate the effectiveness of our approach using real traffic data. SafeRNet intends to improve drivers' safety in a modern technology rich transportation system.

This research study proposes a new method for automatic design of Fuzzy Cognitive Maps (FCM) using ordinal data based on the efficient capabilities of mixed graphical models. The approach is able to model all variables on the proper domain of ordinal data by combining a new class of Mixed Graphical Models (MGMs) with a structure estimation approach based on generalized covariance matrices. It can work with a large amount of categorical data. It represents its structure as a sparser graph, while maintaining a high likelihood, by producing an adjacent weight matrix, where relationships are expressed by conditional independences. By maximizing the likelihood indicates that the model fits better to the data under the assumption that the observed data are the most likely data. The whole approach was implemented in a business intelligence problem of evaluating the attractiveness of Belgian companies. Through the analysis of results and conducted scenarios, the usefulness of the proposed MGM method for designing FCM capable to make decisions, is demonstrated. Comparisons with the previous known methodology for automatic construction of FCMs based on distance-based algorithm, showed that the proposed approach provides more understandable/useful relationships among nodes, through a less complex structure for making decisions.

According to Public Health Agency of Canada, Cardiovascular Disease (CVD) is the leading cause of death among adult men and women. Various research works have applied machine learning/data mining algorithms to predict CVD, but these methods suffer from a) lack of transparency of the predictive model building, b) lack of capability to introduce human wisdom, and c) lack of sufficient data. In this paper we provide a novel approach to tackle these issues and design a very robust and reasonably accurate model. Our approach is based on Structural Equation Modeling (SEM) and Fuzzy Cognitive Map (FCM). We used Canadian Community Health Survey, 2012 data set to test our approach. The designed model has 79% area under the ROC curve and 74% accuracy. We have used only the 20 most significant attributes, but we believe that adding more attributes and having an expert heart specialist panel would further improve the accuracy of the system.

Human immunodeficiency virus (HIV) has been a major health problem throughout the world for decades. This paper introduces a new discrete simulation model for the growth of HIV infection within a host body. The model is developed incrementally, and compared at each stage for congruence with real-world observations of disease dynamics. We used the model to get a better understanding of how HIV-infected cells behave, and particularly to assess the importance of medication adherence for effectiveness of treatment. We found that a small lack of adherence could have a proportionally much larger impact on infection as well as trigger negative effects on health sooner. Consequently, improving adherence can be very beneficial (particularly for those whose adherence is already high), and small issues in adherence should be addressed early on. Our work is one step in the development of detailed discrete simulations for HIV dynamics. However, much work remains to be done in order to accurately capture adherence as well as medication schedule, viral resistance, or multiple medication types.

Abstract Crime attractors are locations (e.g. shopping malls) that attract criminally motivated offenders because of the presence of known criminal opportunities. Although there have been many studies that explore the patterns of crime in and around these locations, there are still many questions that linger. In recent years, there has been a growing interest to develop mathematical models in attempts to help answer questions about various criminological phenomena. In this paper, we are interested in applying a formal methodology to model the relative attractiveness of crime attractor locations based on characteristics of offenders and the crime they committed. To accomplish this task, we adopt fuzzy logic techniques to calculate the attractiveness of crime attractors in three suburban cities in the Metro Vancouver region of British Columbia, Canada. The fuzzy logic techniques provide results comparable with our real‐life expectations that offenders do not necessarily commit significant crimes in the immediate neighbourhood of the attractors, but travel towards it, and commit crimes on the way. The results of this study could lead to a variety of crime prevention benefits and urban planning strategies.

Attempts to model insurgency have suffered from several obstacles. Qualitative research may be vague and conflicting, while quantitative research is limited due to the difficulties of collecting sufficient data in war and inferring complex relationships. We propose an innovative combination of Fuzzy Cognitive Maps and Cellular Automata to capture this complexity. Our approach is computational, thus it can be used to develop a simulation platform in which military and political analysts can test scenarios. We take a step-by-step approach to illustrate the potential of our approach in a population-centric war, similar to the on-going campaign in Afghanistan. While the project still requires validation and improvement of the knowledge base by domain experts as well as construction of accurate simulation scenarios, this example fully specifies the general problem definition and the technical structure of the model.