Domain Expert Knowledge Research Articles

Data Integration into OMOP CDM for Heterogeneous Clinical Data Collections via HL7 FHIR Bundles and XSLT.

Data integration is an important task in medical informatics and highly impacts the gain out of existing health information data. These tasks are using implemented as extract transform and load processes. By introducing HL7 FHIR as an intermediate format, our aim was to integrate heterogeneous data from a German pulmonary hypertension registry into an OMOP Common Data Model. First, domain knowledge experts defined a common parameter set, which was subsequently mapped to standardized terminologies like LOINC or SNOMED-CT. Data was extracted as HL7 FHIR Bundle to be transformed to OMOP CDM by using XSLT. We successfully transformed the majority of data elements to the OMOP CDM in a feasible time.

Bearing fault diagnosis using deep learning techniques coupled with handcrafted feature extraction: A comparative study

Deep learning has seen tremendous growth over the past decade. It has set new performance limits for a wide range of applications, including computer vision, speech recognition, and machinery health monitoring. With the abundance of instrumentation data and the availability of high computational power, deep learning continues to prove itself as an efficient tool for the extraction of micropatterns from machinery big data repositories. This study presents a comparative study for feature extraction capabilities using stacked autoencoders considering the use of expert domain knowledge. Case Western Reserve University bearing dataset was used for the study, and a classifier was trained and tested to extract and visualize features from 12 different failure classes. Based on the raw data preprocessing, four different deep neural network structures were studied. Results indicated that integrating domain knowledge with deep learning techniques improved feature extraction capabilities and reduced the deep neural networks size and computational requirements without the need for exhaustive deep neural networks architecture tuning and modification.

PNS146 BEYOND DEFAULTS - ADAPTING PRINCIPLED PRIOR SELECTION IN BAYESIAN NETWORK META-ANALYSIS

Bayesian methods provide a logical and consistent decision making framework to guide evidence synthesis and commonly used for Network Meta-Analysis (NMA). Bayesian analyses allow for estimation of rare events and can protect against issues of multiple comparisons through incorporation of genuine prior information. We adapt recent guidance documents for principled prior selection to NMA. We provide a summary of recent publications emphasizing the importance of prior selection for leveraging the full benefits of a Bayesian decision making framework. We then adapt and provide a worked example of methods based on simulation from the prior predictive distribution to facilitate understanding of the implication of priors on decision making. Standard default priors recommended in NMA methods guidance documents for health technology assessment imply unrealistic bounds for ratio-based measures and impossible values for continuous summaries like the mean difference. Prior predictive distributions for these statistics imply that prior belief places more weight on implausibly large values than are commonly seen across a variety of domains. A worked example in neonatology shows how the use visualizations of the prior predictive distribution can be incorporated in the protocol development stage in order to set priors for analysis that respect genuine expert domain knowledge. We show that conclusions of NMAs where any node is sparsely informed can differ when reasonable priors are selected. Bayesian inference allows for the incorporation of prior information with data in order to support rational decision-making. The use of standard, default “vague” priors in an effort to “let the data speak for themselves” is likely to be inconsistent with genuine domain knowledge and could meaningfully change conclusions of NMAs. Priors chosen at the protocol development stage based on visualizations of the prior predictive distribution may lead to more consistent decisions.

Analyzing employee behavior related questionnaires by combined fuzzy signature model

The evaluation of data obtained from responses given to questionnaires in humanities and social sciences, such as management, linguistics, etc. is a complex task with the necessity of dealing with the inherent subjectivity and vagueness in such data. In this paper, a method based on fuzzy signatures (FSigs), suitable for analyzing questionnaires with hierarchically connected (partially) vague responses is proposed, and its applicability will be demonstrated by a real life problem; the partial analysis of an ongoing research examining employee behavior in various companies. The linkage of the factors hidden in the data bases obtained from the answers to the questionnaires, containing various factors interconnected in a more or less tight way, are represented by a hierarchical FSig system, allowing further evaluation and the discovery of emerging connections and deeper patterns among the responses, thus extending the idea of the original FSig model towards a more general, fuzzy-fuzzy signature approach. The method proposed here is a combination of some statistical elements with the Fuzzy Signature model, and it also uses Kohonen-maps in order to discover deeper structural components in the data pool. As FSigs are suitable to express hierarchically structured connections among vague and imprecise features of the individual data, the statistical analysis helps reveal the degrees of redundancies and the closeness of connectedness of the individual elements within the responses, and thus enable the construction of a relevant FSig tree graph for the data on hand, while further expert domain knowledge helps with determining the proper fuzzy aggregations in the intermediate nodes of the FSigs. The case study presented is based on data obtained from North Lithuanian companies. The results of the case study focusing on the analysis of the connection between OCB and CWB, and other factors, disclose some interesting and, partly unexpected, results. They indicate a strong and unambiguous relationship between career satisfaction and OCB, which is not very surprising. However, it is found that there is no relationship with gender, age, and actual position in the company, which are generally supposed to be determining factors. These results may be further validated by expert knowledge, and thus the new combined method for evaluating structured multicomponent data and internal dependencies is adequate.

The Mobile Health Multiple Lifestyle Behavior Interventions Across the Lifespan (MoBILE) Research Program: Protocol for Development, Evaluation, and Implementation

BackgroundClustering of multiple lifestyle risk behaviors has been associated with a greater risk of noncommunicable diseases and mortality than one lifestyle risk behavior or no lifestyle risk behaviors. The National Board of Health and Welfare in Sweden reported in 2018 that it is important to provide additional support to individuals with multiple lifestyle risk behaviors, as risks from these behaviors are multiplicative rather than additive. However, the same report emphasized that there is a lack of knowledge regarding interventions that support changes to unhealthy lifestyle behaviors.ObjectiveThe MoBILE (Mobile health Multiple lifestyle Behavior Interventions across the LifEspan) research program has brought together two Swedish research groups supported by international collaborators. Through this collaboration, we aim to design and evaluate a number of novel and tailored mobile health (mHealth) multiple lifestyle behavior interventions across the life span of different health care populations. In addition, the MoBILE research program will extend ongoing research to include mHealth interventions for migrant pregnant women and children.MethodsEach project within the MoBILE program will focus on a specific group: pregnant women, preschool children, high school and university students, and adults in primary and clinical care. All the projects will follow the same 4 phases: requirements, development, evaluation, and implementation. During the requirements phase, implementers and end users will aid the design of content and functionality of the interventions. In the development phase, findings from the first phase will be synthesized with expert domain knowledge and theoretical constructs to create interventions tailored to the target groups. The third phase, evaluation, will comprise randomized controlled trials conducted to estimate the effects of the interventions on multiple lifestyle risk behaviors (eg, alcohol, nutrition, physical activity, and smoking). The final phase will investigate how the interventions, if found effective, can be disseminated into different health care contexts.ResultsThe research program commenced in 2019, and the first results will be available in 2020. Projects involving pregnant women, preschool children, and high school and university students will be completed in the first 3 years, with the remaining projects being planned for the program’s final 3 years.ConclusionsThe development of evidence-based digital tools is complex, as they should be guided by theoretical frameworks, and requires large interdisciplinary teams with competence in technology, behavioral science, and lifestyle-specific areas. Individual researchers or smaller research groups developing their own tools is not the way forward, as it means reinventing the wheel over and over again. The MoBILE research program therefore aims to join forces and learn from the past 10 years of mHealth research to maximize scientific outcomes, as well as the use of financial resources to expand the growing body of evidence for mHealth lifestyle behavior interventions.International Registered Report Identifier (IRRID)PRR1-10.2196/14894

Classification Model on Big Data in Medical Diagnosis Based on Semi-Supervised Learning

Abstract Big data in medical diagnosis can provide abundant value for clinical diagnosis, decision support and many other applications, but obtaining a large number of labeled medical data will take a lot of time and manpower. In this paper, a classification model based on semi-supervised learning algorithm using both labeled and unlabeled data is proposed to process big data in medical diagnosis, which includes structured, semi-structured and unstructured data. For the medical laboratory data, this paper proposes a self-training algorithm based on repeated labeling strategy to solve the problem that mislabeled samples weaken the performance of classifiers. Aiming at medical record data, this paper extracts features with high correlation of classification results based on domain expert knowledge base first, and then chooses the unlabeled medical record data with the highest confidence to expand the training set and optimizes the performance of the classifiers of tri-training algorithm, which uses supervised learning algorithm to train three basic classifiers. The experimental results show that the proposed medical diagnosis data classification model based on semi-supervised learning algorithm has good performance.

A generic hierarchical clustering approach for detecting bottlenecks in manufacturing

The advancements in machine learning (ML) techniques open new opportunities for analysing production system dynamics and augmenting the domain expert's decision-making. A common problem for domain experts on the shop floor is detecting throughput bottlenecks, as they constrain the system throughput. Detecting throughput bottlenecks is necessary to prioritise maintenance and improvement actions and obtain greater system throughput. The existing literature provides many ways to detect bottlenecks from machine data, using statistical-based approaches. These statistical-based approaches can be best applied in environments where the statistical descriptors of machine data (such as distribution of machine data, correlations and stationarity) are known beforehand. Computing statistical descriptors involves statistical assumptions. When the machine data doesn't comply with these assumptions, there is a risk of the results being disconnected from actual production system dynamics. An alternative approach to detecting throughput bottlenecks is to use ML- based techniques. These techniques, particularly unsupervised ML techniques, require no prior statistical information on machine data. This paper proposes a generic, unsupervised ML-based hierarchical clustering approach to detect throughput bottlenecks. The proposed approach is the outcome of systematic and careful selection of ML techniques. It begins by generating a time series of the chosen bottleneck detection metric and then clustering the time series using a dynamic time-wrapping measure and a complete-linkage agglomerative hierarchical clustering technique. The results are clusters of machines with similar production dynamic profiles, revealed from the historical data and enabling the detection of bottlenecks. The proposed approach is demonstrated in two real-world production systems. The approach integrates the concept of humans in-loop by using the domain expert's knowledge.

SPREAD: A Fully Automated Toolkit for Single-Particle Cryogenic Electron Microscopy Data 3D Reconstruction with Image-Network-Aided Orientation Assignment.

For the past decade, cryogenic electron microscopy (cryo-EM) has become an important technology to determine three-dimensional (3D) structures of biomacromolecules. Many software tools have been developed for cryo-EM image processing and 3D reconstruction, covering various computational tasks in cryo-EM data analysis. Despite the recent progress, most of these software tools focus on a single task, such as automatic particle picking or image clustering, whereas software packages covering the whole pipeline of cryo-EM data processing are still few. In this study, we developed a fully automatic single-particle reconstruction and analysis toolkit for cryo-EM data, named SPREAD, which integrates 2D image classification, 3D initial model generation, model selection, and 3D refinement. In SPREAD, we adopt our previously proposed network-based clustering algorithm for 2D image classification, NCEM, and the reference-free resolution measurement method SRes to realize the automatic model ranking and selection procedure. Projection orientation assignment is one of the key steps in initial model generation and 3D refinement. In SPREAD, we use the network-based image similarity metric and introduce a new probabilistic-based orientation searching method, named peak finding, to enhance assignment of the projection orientations. For dealing with both the particle images and projection images in the 3D refinement using SPREAD, we build a mixture image network containing both of these types of images on the basis of the peak-finding results, and then similarities for node pairs are recomputed by a superposed random walk on the network. SPREAD achieves a fully automatic workflow in which nearly no expert domain knowledge and interactive manual operation are involved. Our software can accessed for free at http://www.csbio.sjtu.edu.cn/bioinf/SPREAD/ for academic use.

Model-based development of a mental workload-sensitivity index for subject clustering

Considering individual differences are common praxis in several research areas and subjects’ clustering is important to improve knowledge. The concept of mental workload (WL) comprises a number of individual characteristics but is rarely used for subject clustering. In our article, we introduce an approach for the calculation of a WL-sensitivity index that can be used for such purpose. Based on the hypothesis of cognitive resource limitation, we present a two-parametric (nonlinear) logistic model that predicts WL-sensitivity parameters for sample means across participants of a subjective WL measure. It takes into account the specific scale limits of WL metrics, and integrates domain expert knowledge as prior information. Experimental evidence is provided by means of a human-in-the-loop simulation experiment with air traffic controllers. The WL effects were measured using subjective ‘Instantaneous Self-Assessment’ (ISA) under eight different task load scenarios realised by variation of traffic flow n and a dichotomous non-nominal (priority) event e. Analysis of the ISA(n) data shows that the theoretically predicted ISA vs. n characteristic exhibits good agreement with the experimental parameter estimates when based on the ISA scenario averages despite large inter-individual variance. For those scenarios including the event (e = 1), a significant increase of WL sensitivity is observed for n > critical load nx that gives rise to an additional nonlinearity. Moreover, for the given traffic load range, our two-parameter model may be linearised so that in a simple way participant subgroups of different WL sensitivity may be defined based on a linear dimensionless ISA(n) sensitivity index.

Patent Automatic Classification Based on Symmetric Hierarchical Convolution Neural Network

With the rapid growth of patent applications, it has become an urgent problem to automatically classify the accepted patent application documents accurately and quickly. Most previous patent automatic classification studies are based on feature engineering and traditional machine learning methods like SVM, and some even rely on the knowledge of domain experts, hence they suffer from low accuracy problem and have poor generalization ability. In this paper, we propose a patent automatic classification method via the symmetric hierarchical convolution neural network (CNN) named PAC-HCNN. We use the title and abstract of the patent as the input data, and then apply the word embedding technique to segment and vectorize the input data. Then we design a symmetric hierarchical CNN framework to classify the patents based on the word embeddings, which is much more efficient than traditional RNN models dealing with texts, meanwhile keeping the history and future information of the input sequence. We also add gated linear units (GLUs) and residual connection to help realize the deep CNN. Additionally, we equip our model with a self attention mechanism to address the long-term dependency problem. Experiments are performed on large-scale datasets for Chinese short text patent classification. Experimental results prove our proposed model’s effectiveness, and it performs better than other state-of-the-art models significantly and consistently on both fine-grained and coarse-grained classification.

Multi‐Layer Feature Selection Incorporating Weighted Score‐Based Expert Knowledge toward Modeling Materials with Targeted Properties

AbstractSelecting proper descriptors or features is one of the central problems in exploring structure–activity relationships of materials using machine learning models. The current feature selection algorithms usually require tedious hyperparameter tuning and do not actively consider the prior knowledge of domain experts about the features. Here, this work proposes a data‐driven multi‐layer feature selection method incorporating domain expert knowledge named DML‐FSdek, which is automated, with users entering training data without manual tuning of the hyperparameters. The domain expert knowledge is quantified by means of weighted scoring and integrated into the selection process to eliminate the risk of crucial features being removed. The test studies on ten material properties datasets demonstrate the potential of the approach to automatically search for a reduced feature set with lower root mean square errors than those for the initial feature set. Essentially, the most relevant material features, the number of which is much smaller than that in the original feature set, are automatically selected to establish a closer and more accurate structure–activity relationship for the materials of interest. As a result, the method represents the targeted properties of materials with a smaller and more interpretable set of features while ensuring equal or better prediction accuracy.

Functional Brain Network Estimation with Human-Guided Modularity Representation

Abstract Functional brain network (FBN) has been demonstrated with remarkable advancements in understanding the human brain organization architectures and diagnosis disorders. Thus, it is crucial to accurately estimate both biologically meaningful and discriminative FBNs. Although several FBN estimation approaches have been proposed, the accurate estimation of FBN is still an open field due to the high complexity of human brains and the poor quality of the observed data. Moreover, most existing works fail in incorporating domain expert knowledge. In this paper, we stress the importance of both modular topology prior and domain expert knowledge for FBN estimation, and a human-guided modular representation (MR) FBN estimation framework is proposed. Specifically, we depict the intra- and intermodular structures of FBNs under domain expert knowledge guidance and characterize them with an adversarial low-rank constraint. An efficient ConCave-Convex Procedure (CCCP) is applied to estimate FBN, which is then verified on the Chronic Tinnitus Identification task. The proposed methods achieves a 92.11% classification accuracy, significantly outperformed the state-of-the-art methods. Our method also tends to provide more biologically meaningful functional connections, which benefit for both basic and clinical neuroscience studies.

A Scalable Synthesis of Multiple Models of Geo Big Data Interpretation

The paper proposes a scalable fuzzy approach for mapping the status of the environment integrating several distinct models exploiting geo big data. The process is structured into two phases: the first one can exploit products yielded by distinct models of remote sensing image interpretation defined in the scientific literature, and knowledge of domain experts, possibly ill-defined, for computing partial evidence of a phenomenon. The second phase integrates the partial evidence maps through a learning mechanism exploiting ground truth to compute a synthetic Environmental Status Indicator (ESI) map. The proposal resembles an ensemble approach with the difference that the aggregation is not necessarily consensual but can model a distinct decision attitude in between pessimistic and optimistic. It is scalable and can be implemented in a distributed processing framework, so as to make feasible ESI mapping in near real time to support land monitoring. It is exemplified to map the presence of standing water areas, indicator of water resources, agro-practices or natural hazard from remote sensing by considering different models.

Developing a Prototype Knowledge-Based System for Diagnosis and Treatment of Diabetes Using Data Mining Techniques.

Diabetes is a disease that affects the body's ability to produce or use insulin. A total of 425 million people are suffering from diabetes in the world. Of this, more than 16 million people live in the Africa Region, which is estimated to be around 41 million by 2045. The main objective of this study was to design and develop a prototype knowledge-based system using data mining techniques for diagnosis and treatment of diabetes. For this study, experimental research design was employed, and the researchers used domain expert knowledge as a supplement of data mining techniques whereby three classification algorithms in WEKA; namely J48, PART and JRip were used, and finally the researchers decided to use the results of J48 classification algorithm. Ultimate Visual basic studio 2013 (Vb.net) was used to store knowledge and as front side of prototype. Common lisp prolog (Clisp) was used for obtained knowledge back end coding. Using a decision tree algorithm; namely J48, 2512 (95.1515%) of the instances were classified correctly, and 128 (4.8485 %) were classified incorrectly. The second most performing model was generated by JRip Classier. This model scored the 94.7348% accuracy on the general data to classify the status of diabetic patient datasets. It classified the 2501 instances of the records correctly. The J48 model was the best performing model with the best accuracy of results.

A Combined Deep-Learning and Transfer-Learning Approach for Supporting Social Influence Prediction

Social influence is a phenomenon describing the spread of opinions across the population. Nowadays, social influence analysis (SIA) has a great impact. For example, viral marketing and online content recommendation are applications of SIA. Hand-crafted features, as well as domain expert knowledge, are usually required in convention social influence analysis, but they incur high costs and are not scalable. Deep learning based approaches overcome these issues. For instance, a recently used approach learned latent features of users to predict social influence. In this paper, a teleport probability τ from the page rank domain is integrated into the graph convolution network model for further enhance the performance of such an approach. In addition, a combined personalized propagation of neural predictions (CPPNP) algorithm leads to an impressive prediction accuracy when comparing with existing methods. Evaluation results on three well-known datasets reveal that optimizing τ enhances the performance of CPPNP. Such a combined deep-learning and transfer-learning approach well supports the social influence prediction

Fuzzy knowledge graph system for artificial intelligence-based smart education

Fuzzy knowledge graph system is a semantic network that reveals the relationships between entities, and a tool or methodology that can formally describe things in the real world and their relationships. Smart education is an educational concept or model that uses advanced information technology to build a smart environment, integrates theory and practice to build an educational framework for information age, and provides paths to practice it. Artificial intelligence (AI) is a comprehensive discipline developed by the interpenetration of computer science, cybernetics, information theory, linguistics, neurophysiology and other disciplines, which is a direction for the development of information technology in the future. On the basis of summarizing and analyzing of previous research works, this paper expounded the research status and significance of AI technology, elaborated the development background, current status and future challenges of the construction and application of fuzzy knowledge graph system for smart education, introduced the methods and principles of data acquisition methods and digitalized apprenticeship, realized the process design, information extraction, entity recognition and relationship mining of smart education, constructed a systematic framework for fuzzy knowledge graph, and analyzed the high-quality resources sharing and personalized service of AI-assisted smart education, discussed automatic knowledge acquisition and fusion of fuzzy knowledge graph, performed co-occurrence relationship analysis, and finally conducted application case analysis. The results show that the smart education knowledge graph for AI-assisted smart education can integrate teaching experience and domain knowledge of discipline experts, enhance explainable and robust machine intelligence for AI-assisted smart education, and provide data-driven and knowledge-driven information processing methods; it can also discover the analysis hotspots and main content of research objects through clustering of high-frequency topic words, reveal the corresponding research structure in depth, and then systematically explore its research dimensions, subject background and theoretical basis.

Learning Bayesian Network Parameters With Small Data Set: A Parameter Extension under Constraints Method

Recent advances have illustrated substantial benefits from learning Bayesian networks (BNs). However, when the available data size is small, the BN parameter learning becomes a key challenge in many intelligent applications. By integrating both sample data and expert constraints, we propose a BN parameter learning algorithm with extension method-parameter extension under constraints (PEUC) by introducing related domain expert knowledge. Knowledge is transformed into inequality constraints which candidate parameter sets arise from the relative constraints space.The maximum entropy principle helps to estimate the parameter in statistical averaging model while candidate sets of BN parameters satisfy the constrained knowledge by bootstrapping techniques. Then BN parameters are estimated based on the real available sampled data set and extension streams of candidate parameters samples from the constraints space. The sample size is also taken into account according to the contribution to the final parameters. Experimental results of benchmark BN modeling problems demonstrate that PEUC algorithm tends to the classical MLE algorithm when the modeling data size is sufficient. Furthermore, when the available data size is small, the parameters of BN can be estimated by PEUC as well, and the learned accuracy is superior to MLE, MAP or QMAP algorithm. Finally, PEUC is also applied to a real bearing fault diagnosis case. The presented approach provides a new promising BN parameter learning way for more intelligent system modeling problems, particularly when the data sets are small.

Optimisation of a fuzzy logic-based local real-time control system for mitigation of sewer flooding using genetic algorithms

Abstract Urban flooding damages properties, causes economic losses and can seriously threaten public health. An innovative, fuzzy logic (FL)-based, local autonomous real-time control (RTC) approach for mitigating this hazard utilising the existing spare capacity in urban drainage networks has been developed. The default parameters for the control algorithm, which uses water level-based data, were derived based on domain expert knowledge and optimised by linking the control algorithm programmatically to a hydrodynamic sewer network model. This paper describes a novel genetic algorithm (GA) optimisation of the FL membership functions (MFs) for the developed control algorithm. In order to provide the GA with strong training and test scenarios, the compiled rainfall time series based on recorded rainfall and incorporating multiple events were used in the optimisation. Both decimal and integer GA optimisations were carried out. The integer optimisation was shown to perform better on unseen events than the decimal version with considerably reduced computational run time. The optimised FL MFs result in an average 25% decrease in the flood volume compared to those selected by experts for unseen rainfall events. This distributed, autonomous control using GA optimisation offers significant benefits over traditional RTC approaches for flood risk management.

Incorporating monotonic domain knowledge in support vector learning for data mining regression problems

A common problem of data-driven data mining methods is that they might lack considering domain knowledge, despite possibly having high accuracy with respect to the data. As such, prior knowledge plays an important role in many data mining applications. Incorporating prior knowledge into data mining techniques is not trivial and remains a partially open issue drawing much attention. In this paper, we propose a new support vector regression (SVR) model that takes into account the prior knowledge of domain experts in the form of inequalities, which reflect the monotonic relationship between the output and some of the attributes of the input. A dual quadratic programming problem corresponding to the SVR model is derived, along with algorithms for solving it and creating constraints, respectively. The experiment results, which were conducted on two artificial and two practical datasets, show that the proposed model, which considers the monotonicity defined by domain experts, performs better than the original SVR. Moreover, the proposed method is also suitable for prior domain knowledge of piecewise monotonicity.

MRSC

When evaluating the impact of a policy (e.g., gun control) on a metric of interest (e.g., crime-rate), it may not be possible or feasible to conduct a randomized control trial. In such settings where only observational data is available, synthetic control (SC) methods [1-3] provide a popular data-driven approach to estimate a "synthetic" or "virtual" control by combining measurements of "similar" alternatives or units (called "donors"). Recently, robust synthetic control (RSC) [4] was proposed as a generalization of SC to overcome the challenges of missing data and high levels of noise, while removing the reliance on expert domain knowledge for selecting donors. However, both SC and RSC (and its variants) suffer from poor estimation when the pre-intervention period is too short.