Selective Prediction With Long Short-term Memory Using Unit-Wise Batch Standardization for Time Series Health Data Sets: Algorithm Development and Validation

Abstract

Background In any health care system, both the classification of data and the confidence level of such classifications are important. Therefore, a selective prediction model is required to classify time series health data according to confidence levels of prediction. Objective This study aims to develop a method using long short-term memory (LSTM) models with a reject option for time series health data classification. Methods An existing selective prediction method was adopted to implement an option for rejecting a classification output in LSTM models. However, a conventional selection function approach to LSTM does not achieve acceptable performance during learning stages. To tackle this problem, we proposed a unit-wise batch standardization that attempts to normalize each hidden unit in LSTM to apply the structural characteristics of LSTM models that concern the selection function. Results The ability of our method to approximate the target confidence level was compared by coverage violations for 2 time series of health data sets consisting of human activity and arrhythmia. For both data sets, our approach yielded lower average coverage violations (0.98% and 1.79% for each data set) than those of the conventional approach. In addition, the classification performance when using the reject option was compared with that of other normalization methods. Our method demonstrated superior performance for selective risk (12.63% and 17.82% for each data set), false-positive rates (2.09% and 5.8% for each data set), and false-negative rates (10.58% and 17.24% for each data set). Conclusions Our normalization approach can help make selective predictions for time series health data. We expect this technique to enhance the confidence of users in classification systems and improve collaborative efforts between humans and artificial intelligence in the medical field through the use of classification that considers confidence.

A generalized classification and coding system of Human Disease Animal Model Resource data with a case study to show improving database retrieval efficiency

“…According to the first dimension of classification-the human disease [1], animal models can be classified into the following major categories: (1) cardiovascular system models, (2) digestive system models, (3) respiratory system models, (4) urinary system models, (5) reproductive system models, (6) endocrine diseases models, (7)ophthalmology and otolaryngology models, (8) oral diseases models, (9) bone diseases models, (10) skin diseases models, (11) nervous system models, (12) blood system models, (13) infectious diseases models, (14) tumor models, (15) traditional Chinese medicine (TCM) viscera dialectics models, and (16) animal models for other diseases. These categories can be further subdivided into 118 intermediate categories, which can further be subcategorized into more minor categories.…”

“…Considering the characteristics of data such as multi-source, heterogeneity and imbalance, embedded multi-function data mining technology based on granular computing has been developed [ 14 ]. In order to use time series health data sets for selective prediction, researchers developed an algorithm using long short-term memory and unit-wise batch standardization [ 15 ]. Due to the lack of standard analytical research methods in the field of psychology, there has been a long-standing problem of replication research difficulty.…”

InsightSleepNet: the interpretable and uncertainty-aware deep learning network for sleep staging using continuous Photoplethysmography

“…Moreover, introducing a technique to measure prediction uncertainty in cases where confusion arises among specific sleep stage classes could enhance human decision-making by providing information about the degree of uncertainty in the predictions and allowing for decisions to be postponed [ 25 , 26 ]. This approach is similar to studies that have measured uncertainty in other biomedical domains and can enable synergies between humans and AI [ 27 – 29 ]. Existing methods for measuring uncertainty in the predictions of deep learning models include computing uncertainty over the trained model, such as normalized entropy [ 30 ] and softmax response [ 31 ], and computing uncertainty using a dropout layer embedded in the model, such as Monte-Carlo dropout [ 32 ].…”