1 January 2017 in Journal of Physics: Conference Series
A paper has a deal with the problem of quality assessment in sensor data streams accumulated by proactive decision support systems. The new problem is stated where outliers need to be detected and to be classified according to their nature of origin. There are two types of outliers defined; the first type is about misoperations of a system and the second type is caused by changes in the observed system behavior due to inner and external influences. The proposed method is based on the data-driven forecast approach to predict the values in the incoming data stream at the expected time. This method includes the forecasting model and the clustering model. The forecasting model predicts a value in the incoming data stream at the expected time to find the deviation between a real observed value and a predicted one. The clustering method is used for taxonomic classification of outliers. Constructive neural networks models (CoNNS) and evolving connectionists systems (ECS) are used for prediction of sensors data. There are two real world tasks are used as case studies. The maximal values of accuracy are 0.992 and 0.974, and F1 scores are 0.967 and 0.938, respectively, for the first and the second tasks. The conclusion contains findings how to apply the proposed method in proactive decision support systems.
1 May 2021 in Programming and Computer Software
This paper describes an approach based on machine learning technology that is of particular interest for the localization and characterization of both single focal stenoses and multivessel multifocal lesions. Due to the complexity of analyzing large amounts of data for the cardiac surgeon, we pay special attention to the analysis, training, and comparison of popular neural networks that classify and localize foci of stenosis on coronary angiography data. From the complete coronarography dataset collected at the Research Institute for Complex Issues of Cardiovascular Diseases, we retrospectively select data of 100 patients. For the automated analysis of the medical data, the paper considers in detail three models (SSD MobileNet V1, Faster-RCNN ResNet-50 V1, and Faster-RCNN NASNet), which differ in their architecture, complexity, and the number of weights. The models are compared in terms of their basic efficiency characteristics: accuracy, training time, and prediction time. The test results show that the training and prediction times are directly proportional to the complexity of the models. In this regard, Faster-RCNN NASNet exhibits the lowest prediction time (the average processing time for one image is 880 ms), while Faster-RCNN ResNet-50 V1 has the highest prediction accuracy. The latter model reaches the mean average precision (mAP) level of 0.92 on the validation dataset. On the other hand, SSD MobileNet V1 is the fastest model, capable of making predictions with a prediction rate of 23 fps.