scholarly journals Applying Artificial Intelligence Methods for the Estimation of Disease Incidence: The Utility of Language Models

2020 ◽  
Vol 2 ◽  
Author(s):  
Yuanzhao Zhang ◽  
Robert Walecki ◽  
Joanne R. Winter ◽  
Felix J. S. Bragman ◽  
Sara Lourenco ◽  
...  
Keyword(s):  
Machine Learning ◽  
Digital Health ◽  
Contextual Information ◽  
Disease Incidence ◽  
Language Models ◽  
Mathematical Representation ◽  
Learning Models ◽  
Neural Network Models ◽  
Universal Sentence ◽  
Machine Learning Models

Background: AI-driven digital health tools often rely on estimates of disease incidence or prevalence, but obtaining these estimates is costly and time-consuming. We explored the use of machine learning models that leverage contextual information about diseases from unstructured text, to estimate disease incidence.Methods: We used a class of machine learning models, called language models, to extract contextual information relating to disease incidence. We evaluated three different language models: BioBERT, Global Vectors for Word Representation (GloVe), and the Universal Sentence Encoder (USE), as well as an approach which uses all jointly. The output of these models is a mathematical representation of the underlying data, known as “embeddings.” We used these to train neural network models to predict disease incidence. The neural networks were trained and validated using data from the Global Burden of Disease study, and tested using independent data sourced from the epidemiological literature.Findings: A variety of language models can be used to encode contextual information of diseases. We found that, on average, BioBERT embeddings were the best for disease names across multiple tasks. In particular, BioBERT was the best performing model when predicting specific disease-country pairs, whilst a fusion model combining BioBERT, GloVe, and USE performed best on average when predicting disease incidence in unseen countries. We also found that GloVe embeddings performed better than BioBERT embeddings when applied to country names. However, we also noticed that the models were limited in view of predicting previously unseen diseases. Further limitations were also observed with substantial variations across age groups and notably lower performance for diseases that are highly dependent on location and climate.Interpretation: We demonstrate that context-aware machine learning models can be used for estimating disease incidence. This method is quicker to implement than traditional epidemiological approaches. We therefore suggest it complements existing modeling efforts, where data is required more rapidly or at larger scale. This may particularly benefit AI-driven digital health products where the data will undergo further processing and a validated approximation of the disease incidence is adequate.

First-Break Picking Classification Models Using Recurrent Neural Network

2021 ◽  
Author(s):  
Mohammed Ayub ◽  
SanLinn Kaka
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Deep Neural Network ◽  
Contextual Information ◽  
Classification Model ◽  
Superior Performance ◽  
Learning Models ◽  
Neural Network Models ◽  
Minimum Number ◽  
Machine Learning Models

Abstract Manual first-break picking from a large volume of seismic data is extremely tedious and costly. Deployment of machine learning models makes the process fast and cost effective. However, these machine learning models require high representative and effective features for accurate automatic picking. Therefore, First- Break (FB) picking classification model that uses effective minimum number of features and promises performance efficiency is proposed. The variants of Recurrent Neural Networks (RNNs) such as Long ShortTerm Memory (LSTM) and Gated Recurrent Unit (GRU) can retain contextual information from long previous time steps. We deploy this advantage for FB picking as seismic traces are amplitude values of vibration along the time-axis. We use behavioral fluctuation of amplitude as input features for LSTM and GRU. The models are trained on noisy data and tested for generalization on original traces not seen during the training and validation process. In order to analyze the real-time suitability, the performance is benchmarked using accuracy, F1-measure and three other established metrics. We have trained two RNN models and two deep Neural Network models for FB classification using only amplitude values as features. Both LSTM and GRU have the accuracy and F1-measure with a score of 94.20%. With the same features, Convolutional Neural Network (CNN) has an accuracy of 93.58% and F1-score of 93.63%. Again, Deep Neural Network (DNN) model has scores of 92.83% and 92.59% as accuracy and F1-measure, respectively. From the pexperiment results, we see significant superior performance of LSTM and GRU to CNN and DNN when used the same features. For robustness of LSTM and GRU models, the performance is compared with DNN model that is trained using nine features derived from seismic traces and observed that the performance superiority of RNN models. Therefore, it is safe to conclude that RNN models (LSTM and GRU) are capable of classifying the FB events efficiently even by using a minimum number of features that are not computationally expensive. The novelty of our work is the capability of automatic FB classification with the RNN models that incorporate contextual behavioral information without the need for sophisticated feature extraction or engineering techniques that in turn can help in reducing the cost and fostering classification model robust and faster.

scholarly journals Quantifying the Impacts of Pre- and Post-Conception TSH Levels on Birth Outcomes: An Examination of Different Machine Learning Models

2021 ◽  
Vol 12 ◽  
Author(s):  
Yuantong Sun ◽  
Weiwei Zheng ◽  
Ling Zhang ◽  
Huijuan Zhao ◽  
Xun Li ◽  
...  
Keyword(s):  
Machine Learning ◽  
Birth Outcomes ◽  
Loss Function ◽  
Apgar Score ◽  
Thyroid Stimulating Hormone ◽  
Adverse Birth Outcomes ◽  
Learning Models ◽  
Neural Network Models ◽  
Tsh Levels ◽  
Machine Learning Models

BackgroundWhile previous studies identified risk factors for diverse pregnancy outcomes, traditional statistical methods had limited ability to quantify their impacts on birth outcomes precisely. We aimed to use a novel approach that applied different machine learning models to not only predict birth outcomes but systematically quantify the impacts of pre- and post-conception serum thyroid-stimulating hormone (TSH) levels and other predictive characteristics on birth outcomes.MethodsWe used data from women who gave birth in Shanghai First Maternal and Infant Hospital from 2014 to 2015. We included 14,110 women with the measurement of preconception TSH in the first analysis and 3,428 out of 14,110 women with both pre- and post-conception TSH measurement in the second analysis. Synthetic Minority Over-sampling Technique (SMOTE) was applied to adjust the imbalance of outcomes. We randomly split (7:3) the data into a training set and a test set in both analyses. We compared Area Under Curve (AUC) for dichotomous outcomes and macro F1 score for categorical outcomes among four machine learning models, including logistic model, random forest model, XGBoost model, and multilayer neural network models to assess model performance. The model with the highest AUC or macro F1 score was used to quantify the importance of predictive features for adverse birth outcomes with the loss function algorithm.ResultsThe XGBoost model provided prominent advantages in terms of improved performance and prediction of polytomous variables. Predictive models with abnormal preconception TSH or not-well-controlled TSH, a novel indicator with pre- and post-conception TSH levels combined, provided the similar robust prediction for birth outcomes. The highest AUC of 98.7% happened in XGBoost model for predicting low Apgar score with not-well-controlled TSH adjusted. By loss function algorithm, we found that not-well-controlled TSH ranked 4th, 6th, and 7th among 14 features, respectively, in predicting birthweight, induction, and preterm birth, and 3rd among 19 features in predicting low Apgar score.ConclusionsOur four machine learning models offered valid predictions of birth outcomes in women during pre- and post-conception. The predictive features panel suggested the combined TSH indicator (not-well-controlled TSH) could be a potentially competitive biomarker to predict adverse birth outcomes.

scholarly journals Personalized prediction of rehabilitation outcomes in multiple sclerosis: a proof-of-concept using clinical data, digital health metrics, and machine learning

2020 ◽  
Author(s):  
Christoph M. Kanzler ◽  
Ilse Lamers ◽  
Peter Feys ◽  
Roger Gassert ◽  
Olivier Lambercy
Keyword(s):  
Machine Learning ◽  
Multiple Sclerosis ◽  
Clinical Data ◽  
Linear Models ◽  
Digital Health ◽  
Learning Models ◽  
Rehabilitation Outcomes ◽  
Non Linear ◽  
Linear Machine ◽  
Machine Learning Models

AbstractBackgroundA personalized prediction of upper limb neurorehabilitation outcomes in persons with multiple sclerosis (pwMS) promises to optimize the allocation of therapy and to stratify individuals for resource-demanding clinical trials. Previous research identified predictors on a population level through linear models and clinical data, including conventional assessments describing sensorimotor impairments. The objective of this work was to explore the feasibility of providing an individualized and more accurate prediction of rehabilitation outcomes in pwMS by leveraging non-linear machine learning models, clinical data, and digital health metrics characterizing sensorimotor impairments.MethodsClinical data and digital health metrics were recorded from eleven pwMS undergoing neurorehabilitation. Machine learning models were trained on data recorded pre-intervention. The dependent variables indicated whether a considerable improvement on the activity level was observed across the intervention or not (binary classification), as defined by the Action Research Arm Test (ARAT), Box and Block Test (BBT), or Nine Hole Peg Test (NHPT).ResultsIn a cross-validation, considerable improvements in ARAT or BBT could be accurately predicted (94% balanced accuracy) by only relying on patient master data. Considerable improvements in NHPT could be accurately predicted (89% balanced accuracy), but required knowledge about sensorimotor impairments. Assessing these with digital health metrics instead of conventional scales allowed increasing the balanced accuracy by +17% . Non-linear machine-learning models improved the predictive accuracy for the NHPT by +25% compared to linear models.ConclusionsThis work demonstrates the feasibility of a personalized prediction of upper limb neurorehabilitation outcomes in pwMS using multi-modal data collected before neurorehabilitation and machine learning. Information from digital health metrics about sensorimotor impairment was necessary to predict changes in dexterous hand control, thereby underlining their potential to provide a more sensitive and fine-grained assessment than conventional scales. Non-linear models outperformed ones, suggesting that the commonly assumed linearity of neurorehabilitation is oversimplified.clinicaltrials.gov registration number: NCT02688231

scholarly journals MODIS Fractional Snow Cover Mapping Using Machine Learning Technology in a Mountainous Area

2020 ◽  
Vol 12 (6) ◽  
pp. 962 ◽  
Author(s):  
Changyu Liu ◽  
Xiaodong Huang ◽  
Xubing Li ◽  
Tiangang Liang
Keyword(s):  
Machine Learning ◽  
Random Forest ◽  
Snow Cover ◽  
Back Propagation ◽  
Network Models ◽  
Support Vector ◽  
Learning Models ◽  
Neural Network Models ◽  
Fractional Snow Cover ◽  
Machine Learning Models

To improve the poor accuracy of the MODIS (Moderate Resolution Imaging Spectroradiometer) daily fractional snow cover product over the complex terrain of the Tibetan Plateau (RMSE = 0.30), unmanned aerial vehicle and machine learning technologies are employed to map the fractional snow cover based on MODIS over this terrain. Three machine learning models, including random forest, support vector machine, and back-propagation artificial neural network models, are trained and compared in this study. The results indicate that compared with the MODIS daily fractional snow cover product, the introduction of a highly accurate snow map acquired by unmanned aerial vehicles as a reference into machine learning models can significantly improve the MODIS fractional snow cover mapping accuracy. The random forest model shows the best accuracy among the three machine learning models, with an RMSE (root-mean-square error) of 0.23, especially over forestland and shrubland, with RMSEs of 0.13 and 0.18, respectively. Although the accuracy of the support vector machine and back-propagation artificial neural network models are worse over forestland and shrubland, their average errors are still better than that of MOD10A1. Different fractional snow cover gradients also affect the accuracy of the machine learning algorithms. Nevertheless, the random forest model remains stable in different fractional snow cover gradients and is, therefore, the best machine learning algorithm for MODIS fractional snow cover mapping in Tibetan Plateau areas with complex terrain and severely fragmented snow cover.

Vibration characteristic analyses of medium-and small-span girder bridge groups in highway systems based on machine learning models

2021 ◽  
pp. 136943322199772
Author(s):  
Guanya Lu ◽  
Kehai Wang ◽  
Weizuo Guo
Keyword(s):  
Machine Learning ◽  
Random Forest ◽  
Large Scale ◽  
Vibration Characteristics ◽  
Structural Vibration ◽  
Learning Models ◽  
Longitudinal Vibrations ◽  
Neural Network Models ◽  
Artificial Neural ◽  
Machine Learning Models

There are large amounts of small-and medium-span girder bridges which bear structural similarity, while the large-scale bridge structures are generally limited in the timely applications of structural vibration characteristics. Therefore, in this study a framework based on machine learning models was proposed to analyze the vibration characteristics of specific line bridge groups. The probability distributions of structural, geometric, and material properties of bridge groups in specific lines were obtained using statistical tools and a Latin hypercube sampling method was used to generate reasonable sample sets for the bridges group, and parameterized finite element models of the bridges were established. Then, the optimal models were tuned and determined to predict fundamental mode and period by the 10-fold cross-validation method applying the numerical simulation results. This study’s results showed that the random forest models divided the vibration modes of the bridge groups into the longitudinal vibrations of the main girders and the longitudinal vibrations of the adjacent spans and side piers with a classification accuracy of greater than 90%, while the artificial neural network models exhibited the lowest normalized mean square error for the periods. The periods mainly ranged between 0.7 and 1.5 s. Furthermore, the bearing settings, ratios of the pier height to section diameters, and boundary types were determined to be the most significant properties influencing the fundamental modes and periods of the examined bridges, by respectively observing the reduced value of the random forest Gini indices and distribution of the generalized weight value of the input variables in artificial neural networks. This study provides an intelligent and efficient method for obtaining vibration characteristics of bridges group for a specific network.

scholarly journals S&P 500 Stock Price Prediction Using Technical, Fundamental and Text Data

2021 ◽  
Vol 9 (4) ◽  
pp. 769-788
Author(s):  
Shan Zhong ◽  
David Hitchcock
Keyword(s):  
Machine Learning ◽  
Stock Prices ◽  
Stock Price ◽  
Language Models ◽  
News Item ◽  
Learning Models ◽  
Stock Price Prediction ◽  
Price Prediction ◽  
Using Data ◽  
Machine Learning Models

We summarized both common and novel predictive models used for stock price prediction and combined them with technical indices, fundamental characteristics and text-based sentiment data to predict S&P stock prices. A 66.18% accuracy in S&P 500 index directional prediction and 62.09% accuracy in individual stock directional prediction was achieved by combining different machine learning models such as Random Forest and LSTM together into state-of-the-art ensemble models. The data we use contains weekly historical prices, finance reports, and text information from news items associated with 518 different common stocks issued by current and former S&P 500 large-cap companies, from January 1, 2000 to December 31, 2019. Our study's innovation includes utilizing deep language models to categorize and infer financial news item sentiment; fusing different models containing different combinations of variables and stocks to jointly make predictions; and overcoming the insufficient data problem for machine learning models in time series by using data across different stocks.

scholarly journals Causality in digital medicine

2021 ◽  
Vol 12 (1) ◽  
Keyword(s):  
Machine Learning ◽  
Data Science ◽  
Digital Health ◽  
Imperial College ◽  
Learning Models ◽  
Digital Medicine ◽  
Health Expert ◽  
Causality Inference ◽  
University College London ◽  
Machine Learning Models

AbstractBen Glocker (an expert in machine learning for medical imaging, Imperial College London), Mirco Musolesi (a data science and digital health expert, University College London), Jonathan Richens (an expert in diagnostic machine learning models, Babylon Health) and Caroline Uhler (a computational biology expert, MIT) talked to Nature Communications about their research interests in causality inference and how this can provide a robust framework for digital medicine studies and their implementation, across different fields of application.

scholarly journals COMPARATIVE ANALYSIS OF MACHINE LEARNING MODELS AND REGRESSIONS FOR CAR PRICE PREDICTION

2019 ◽  
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Comparative Analysis ◽  
Random Forest ◽  
Network Models ◽  
Gradient Boosting ◽  
Learning Models ◽  
Neural Network Models ◽  
Boosting Algorithms ◽  
Machine Learning Models

The purpose of the research described in this article is a comparative analysis of the predictive qualities of some models of machine learning and regression. The factors for models are the consumer characteristics of a used car: brand, transmission type, drive type, engine type, mileage, body type, year of manufacture, seller's region in Ukraine, condition of the car, information about accident, average price for analogue in Ukraine, engine volume, quantity of doors, availability of extra equipment, quantity of passenger’s seats, the first registration of a car, car was driven from abroad or not. Qualitative variables has been encoded as binary variables or by mean target encoding. The information about more than 200 thousand cars have been used for modeling. All models have been evaluated in the Python Software using Sklearn, Catboost, StatModels and Keras libraries. The following regression models and machine learning models were considered in the course of the study: linear regression; polynomial regression; decision tree; neural network; models based on "k-nearest neighbors", "random forest", "gradient boosting" algorithms; ensemble of models. The article presents the best in terms of quality (according to the criteria R2, MAE, MAD, MAPE) options from each class of models. It has been found that the best way to predict the price of a passenger car is through non-linear models. The results of the modeling show that the dependence between the price of a car and its characteristics is best described by the ensemble of models, which includes a neural network, models using "random forest" and "gradient boosting" algorithms. The ensemble of models showed an average relative approximation error of 11.2% and an average relative forecast error of 14.34%. All nonlinear models for car price have approximately the same predictive qualities (the difference between the MAPE within 2%) in this research.

Sign in / Sign up

Export Citation Format

Share Document