| Study program / study programs: Advanced data analytics |
| Course: Practical analysis of noisy and uneven time series |
| Teacher(s): Luka Č. Popović, Anđelka Kovačević, Dragana Ilić |
| Course status: Elective |
| ECTS points: 7 |
| Prerequisites: – |
| Course objective: Most of the phenomena in nature, medicine, science, business and engineering are measured at certain time moments that are most often non-homogeneous in time. Extracting information from such series is a great challenge for analysts because standard techniques are mostly developed for evenly distributed time series without a prominent noise. Therefore, specific methods for analyzing such time series are extremely important for all mentioned data types. This course aims to explain the theoretical and practical core of the concept of time series analysis with such disadvantageous characteristics. |
| Learning outcomes: The student is trained for an effective analysis of noisy time series that are unevenly distributed in time, which can be encountered in the sciences, medicine, business, engineering, as well as in the analysis of the time series of social networks and those found in the sociological research. |
| Course structure and content: Missing data. Sample size. Stochastic and deterministic processes. The concept of stationary time series. Extrapolative and decomposition models. Methods of exponential smoothness. The concept of nonstationary time series. Non-stationary tests. Stabilization of variance, structural or regime stability. Overview of homogeneous and non-homogeneous series. Signal and noise information in the time series. Gaussian process for time series modeling. Poisson’s process for time series modeling. Random Walk model. Fourier’s analysis. Wavelet analysis. Detection difficulties: (1 / f) noise in the time series, Signal detection methods in noisy and non-homogeneous series. Maximum Likelihood Estimate (MLE). |
| Literature/Readings: Terence Mills, Applied Time Series Analysis, Academic Press, 2019 Asis Kumar Chattopadhyay, Tanuka Chattopadhyay, Statistical Methods for Astronomical Data Analysis, Springer, 2014 |
| The number of class hours per week: Lectures: 5 Labs: 0 Workshops: 0 Research study: 3 Other classes: 0 |
| Teaching methods: Individual and group work; lectures and labs |
| Evaluation/Grading (maximum 100 points): Pre-exam requirements: 70 (Class activity – 10, Hands-on activity – 60) Final exam (Tests / Final exam in writing): 30 |
