Search academic texts

Information × Registration Number 0224U033089, (0124U004362) , R & D reports Title Comprehensive care for the next generation, iCare4Next, С2020/1-8 popup.stage_title Розроблення ієрархічного ансамблю моделей машинного навчання для опрацювання великих даних Head Shakhovska Natalya B., Доктор технічних наук Registration Date 17-12-2024 Organization Lviv Polytechnic National University popup.description1 The goal of the NDR "Complex care for the next generation" is to develop methods and means of analysis, to process data to expand opportunities for improving health and home care for older people, to develop data-driven methods of analysis and forecasting. popup.description2 A new approach to determining the weighting threshold for weak classifiers is proposed that improves classification accuracy and adaptability in real-world applications. Another possible approach is to use an imputed dataset. The authors used the k-nearest neighbor method to impute missing values of a function based on other nonmissing values of that function for that district, with a few exceptions. The core of our innovation is the use of mathematical expectation to select the cutoff coefficient in the ensemble. This dynamic voting mechanism takes into account the individual scores of weak classifiers in the ensemble, allowing for context-sensitive decisions. Instead of relying on a static threshold, our approach calculates the average score for each vote, which is then subjected to mathematical expectation to obtain the optimal cutoff coefficient. This adaptive strategy ensures that classification ensembles are finely tuned to the specific characteristics of the input data, leading to improved performance in a number of classification tasks. The proposed hybrid hierarchical ensemble, combining both supervised and unsupervised learning, allows us to improve the accuracy of the regression task by 11% in terms of MSE, 29% in terms of ROC area, and 43% in terms of MPP metric. The ROC-AUC value increased from 0.609 to 0.790; MSE decreased from 112.6 to 101.3, and MPP from 18.8 to 13.1, respectively. Thus, using the proposed approach, it is possible to predict the number of COVID-19 cases and deaths with fairly high accuracy based on demographic, geographic, climatic, transportation, public health, compliance with social distancing policies, and political characteristics.   Product Description popup.authors Basystiuk Oleh A. Melnykova Nataliia I. Pobereiko Petro P popup.nrat_date 2024-12-17 Close