When building predictive models for real-world applications, many data are discarded because conventional learning algorithms cannot utilize it, although such data could be very informative. This paper focuses on representation learning using two types of additional data: privileged information (PI) and unlabeled data. PI refers to data available only during training but not at test time. Existing methods transfer the knowledge embedded in PI via supervised mechanisms, making them unable to use unlabeled data. In contrast, self-supervised learning methods can use unlabeled data but cannot learn from PI. While these techniques appear complementary, as we demonstrate, combining them is non-trivial. This paper introduces the Privileged Information Regularized (PIReg) self-supervised learning framework, which utilizes both PI and unlabeled data to learn better representations.

Predicting treatment medicines is a key task in many intelligent healthcare systems. Prediction of treatment medicines can assist doctors in making informed prescription decisions for patients according to their Electronic Health Records (EHRs). However, predicting treatment medicines is a challenging task due to the following reasons: (1) heterogeneous nature of EHR data that typically includes laboratory results, treatment records, disease conditions, and demographic information; (2) complex correlations among EHR sequences, including inter-correlations between sequences and temporal intra-correlations within each sequence; (3) temporal dynamics of these correlations changing with disease progression. In this paper, we predict treatment medicines for patients with dual adaptive sequential networks (DASNet). Specifically, DASNet is designed with three components. First, a decomposed adaptive long short-term memory network (DA-LSTM) is designed to capture the intra- and inter-correlations in multiple heterogeneous temporal sequences. Then, we develop an attentive meta learning network (AT-MetaNet) to learn dynamic weight parameters for DA-LSTM, thus enabling it to model various correlation structures. Finally, we employ an attentive fusion network (AT-FuNet) to incorporate historical information and collectively fuse representation embeddings of heterogeneous data to predict treatment medicines. Our results on the public MIMIC-III dataset covering 11 medical conditions demonstrate that the proposed end-to-end model can achieve the state-of-the-art prediction performance while providing clinically useful insights.

Stock market timing is regarded as a challenging task of financial prediction. An accurate prediction of stock trend can yield great profits for investors. At present, recurrent neural networks (RNNs) have a good performance in stock market forecasting. However, there has been a relative lack of research in the stock market timing using RNNs. In this paper, a novel model named hybrid RNN model is proposed for stock market timing by incorporating multi-layer long short-term memory, multi-layer gated recurrent unit and one-layer ReLU layer. Moreover, based on five popular benchmark datasets from UCI Machine Learning Repository and six daily securities from Shanghai Stock Exchange, comparisons with 12 state-of-the-art models are conducted to verify the superiority of the proposed hybrid RNN model in terms of nine technical indicators. The findings from the experiment demonstrate that: (1) as opposed to 12 models, the average accuracy, MSE and AUC of hybrid RNN model (0.7406, 0.2592, 0.7368) are significantly better than other comparison models, and (2) the proposed hybrid RNN classification procedure can be considered as a feasible and effective tool for stock market timing.

Advanced analytics such as event detection, pattern recognition, clustering, and classification with electroencephalogram (EEG) data often rely on extracted EEG features. Most of the existing EEG feature extraction approaches are hand-designed with expert knowledge or prior assumptions, which may lead to inferior analytical performances. In this paper, we develop a fully data-driven EEG feature extraction method by applying recurrent autoencoders on multivariate EEG signals. We use an Echo State Network (ESN) to encode EEG signals to EEG features, and then decode them to recover the original EEG signals. Therefore, we name our method feature extraction based on echo state network, or simply FE-ESN. We show that the well-known autoregression-based EEG feature extraction can be seen as a simplified variation of our FE-ESN method. We have conducted experiments on real-world EEG data to evaluate the effectiveness of FE-ESN for both classification tasks and clustering tasks. Experimental results demonstrate the superiority of FE-ESN over the state-of-the-art methods. This paper not only provides a novel EEG feature extraction method but also opens up a new way towards unsupervised EEG feature design.

Recent years have witnessed an opportunity for improving healthcare efficiency and quality by mining Electronic Medical Records (EMRs). This paper is aimed at developing a treatment engine, which learns from historical EMR data and provides a patient with next-period prescriptions based on disease conditions, laboratory results, and treatment records of the patient. Importantly, the engine takes consideration of both treatment records and physical examination sequences which are not only heterogeneous and temporal in nature but also often with different record frequencies and lengths. Moreover, the engine also combines static information (e.g., demographics) with the temporal sequences to provide personalized treatment prescriptions to patients. In this regard, a novel Long Short-Term Memory (LSTM) learning framework is proposed to model inter-correlations of different types of medical sequences by connections between hidden neurons. With this framework, we develop three multifaceted LSTM models: Fully Connected Heterogeneous LSTM, Partially Connected Heterogeneous LSTM, and Decomposed Heterogeneous LSTM. The experiments are conducted on two datasets: one is the public MIMIC-III ICU data, and the other comes from several Chinese hospitals. Experimental results reveal the effectiveness of the framework and the three models. The work is deemed important and meaningful for both academia and practitioners in the realm of medical treatment and prediction, as well as in other fields of applications where intelligent decision support becomes pervasive.

Adverse drug-drug interactions (DDIs) remain a leading cause of morbidity and mortality around the world. Identifying potential DDIs during the drug design process is critical in guiding targeted clinical drug safety testing. Although detection of adverse DDIs is conducted during Phase IV clinical trials, there are still a large number of new DDIs founded by accidents after the drugs were put on market. With the arrival of big data era, more and more pharmaceutical research and development data are becoming available, which provides an invaluable resource for digging insights that can potentially be leveraged in early prediction of DDIs. Many computational approaches have been proposed in recent years for DDI prediction. However, most of them focused on binary prediction (with or without DDI), despite the fact that each DDI is associated with a different type. Predicting the actual DDI type will help us better understand the DDI mechanism and identify proper ways to prevent it. In this paper, we formulate the DDI type prediction problem as a multitask dyadic regression problem, where the prediction of each specific DDI type is treated as a task. Compared with conventional matrix completion approaches which can only impute the missing entries in the DDI matrix, our approach can directly regress those dyadic relationships (DDIs) and thus can be extend to new drugs more easily. We developed an effective proximal gradient method to solve the problem. Evaluation on real world datasets is presented to demonstrate the effectiveness of the proposed approach.

Background Key opinion leaders (KOLs) are people who can influence public opinion on a certain subject matter. In the field of medical and health informatics, it is critical to identify KOLs on various disease conditions. However, there have been very few studies on this topic. Objective We aimed to develop a recommender system for identifying KOLs for any specific disease with health care data mining. Methods We exploited an unsupervised aggregation approach for integrating various ranking features to identify doctors who have the potential to be KOLs on a range of diseases. We introduce the design, implementation, and deployment details of the recommender system. This system collects the professional footprints of doctors, such as papers in scientific journals, presentation activities, patient advocacy, and media exposure, and uses them as ranking features to identify KOLs. Results We collected the information of 2,381,750 doctors in China from 3,657,797 medical journal papers they published, together with their profiles, academic publications, and funding. The empirical results demonstrated that our system outperformed several benchmark systems by a significant margin. Moreover, we conducted a case study in a real-world system to verify the applicability of our proposed method. Conclusions Our results show that doctors’ profiles and their academic publications are key data sources for identifying KOLs in the field of medical and health informatics. Moreover, we deployed the recommender system and applied the data service to a recommender system of the China-based Internet technology company NetEase. Patients can obtain authority ranking lists of doctors with this system on any given disease.