Traffic congestion is a thorny issue to many large and medium-sized cities, posing a serious threat to sustainable urban development. Recently, intelligent traffic system (ITS) has emerged as an effective tool to mitigate urban congestion. The key to the ITS lies in the accurate forecast of traffic flow. However, the existing forecast methods of traffic flow cannot adapt to the stochasticity and sheer length of traffic flow time series. To solve the problem, this paper relies on deep learning (DL) to forecast traffic flow through time series analysis. The authors developed a traffic flow forecast model based on the long shortterm memory (LSTM) network. The proposed model was compared with two classic forecast models, namely, the autoregressive integrated moving average (ARIMA) model and the backpropagation neural network (BPNN) model, through long-term traffic flow forecast experiments, using an actual traffic flow time series from OpenITS. The experimental results show that the proposed LSTM network outperformed the classic models in prediction accuracy. Our research discloses the dynamic evolution law of traffic flow, and facilitates the decision-making of traffic management. INDEX TERMS Traffic flow forecast, time series analysis, deep learning (DL), long short-term memory (LSTM).
Exploring the complicated relationships underlying the clinical information is essential for the diagnosis and treatment of the Coronavirus Disease 2019 (COVID-19). Currently, few approaches are mature enough to show operational impact. Based on electronic medical records (EMRs) of 570 COVID-19 inpatients, we proposed an analysis model of diagnosis and treatment for COVID-19 based on the machine learning algorithms and complex networks. Introducing the medical information fusion, we constructed the heterogeneous information network to discover the complex relationships among the syndromes, symptoms, and medicines. We generated the numerical symptom (medicine) embeddings and divided them into seven communities (syndromes) using the combination of Skip-Gram model and Spectral Clustering (SC) algorithm. After analyzing the symptoms and medicine networks, we identified the key factors using six evaluation metrics of node centrality. The experimental results indicate that the proposed analysis model is capable of discovering the critical symptoms and symptom distribution for diagnosis; the key medicines and medicine combinations for treatment. Based on the latest COVID-19 clinical guidelines, this model could result in the higher accuracy results than the other representative clustering algorithms. Furthermore, the proposed model is able to provide tremendously valuable guidance and help the physicians to combat the COVID-19.
Purpose/Objective(s): Accurate lymph node metastasis (LNM) status evaluation for intrahepatic cholangiocarcinoma (ICC) patients is essential for surgical planning. This study aims to develop a prediction model to noninvasively evaluate LNM status in ICC patients preoperatively, based on diagnostic MR T1 contrast-enhanced images and clinical features, with machine learning SVM method and multivariate analysis. Materials/Methods: Intrahepatic cholangiocarcinoma (ICC) is a rare and aggressive disease. Overall survival of ICC patients was highly correlated with LNM status. LNM is usually confirmed by pathological results from surgical resection. The 148 patients diagnosed from 2011 to 2017 with ICC preoperative were included in this study: 106 patients (2011-2016) as a training set; 42 patients (2016-2017) were included as a validation data set. Radiomics imaging features were extracted from T1 contrast-enhanced MRI images. Final imaging features were selected with a maximum relevance minimum redundancy (mRMR) selection algorithm. A support vector machine (SVM) model was built by using the most LN-related ranked features. An SVM score was calculated to reflect the LNM correlation. Prognosis data for multivariate analysis included: gender, age, primary hepatic lobe site, Hepatitis (W/O), Cirrhosis (W/O), Cholelithiasis (W/O), serum carbohydrate antigen 19-9 (Normal Y/N), serum carcinoembryonic antigen (CEA). Physician based MR-reported LNM (Y/N) status was also reported. A nomogram was then constructed by combining the SVM score from the MR images, clinical features and physician reported LMN, to determine the LNM prediction. Results: The 491 Histogram, Geometry, Texture and Wavelet features were extracted from the T1 MR images. Top five features by mRMR were selected to highly correlate with LNM and obtain the final SVM score. Significant differences were found between patients with and without LNM with P<0.0001 in the validation cohort (0.6027 AE 0.2689 vs. 0.2900 AE 0.2628 by Mann-Whitney U-test). The final combined nomogram was based on the top imaging features from SVM, the CA 19-9 from clinical data by AIC, and the physician-reported LNM, which showed better performance than MR imaging alone for correlating LMN. The model training AUC was 0.842 vs 0.788; testing AUC 0.870 vs 0.787 for the combination model and imaging alone for training data (bootstrapped cross-validated). A decision curve analysis was then calculated from the combination nomogram demonstrating better clinical utility compared to SVM alone and 'treat-all' scenarios. Conclusion: A prediction model for LNM preoperatively based on the combination nomogram developed based on SVM score from MR images and clinical prognosis provides a novel approach to determining LMN status noninvasively to assist effective surgical planning.
We demonstrate a new application of fiber-optic-sensing and machine learning techniques for vehicle run-off-road events detection to enhance roadway safety and efficiency. The proposed approach achieves high accuracy in a testbed under various experimental conditions.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.