SUMMARY
Context. Monitoring natural systems of diverse nature is an essential condition of rational environmental management. Data Mining technologies, cluster analysis in particular, provide a wide range of capabilities for data sets visualization, which makes it possible for these technologies to be used by individuals with no specialized background in mathematics. The task of monitoring a system that changes its state in time requires extended interpretation of clustering result, which would allow accounting for historical data. Technical capabilities for revealing the nature of changes occurring in the object represented by a data set are of particular importance in water resources monitoring area, as they are strongly related to their environment, and the quantity of the available reserves depend on multiple factors, which are external to the water-bearing system. Upon commissioning, an artesian well requires constant monitoring in order to ensure proper management of groundwater processing, protection against pollution and exhaustion, and preventing negative effects of groundwater mining on the environment. In addition, high redundancy of the parameter space is typical for complex natural systems, as well as existence of both known and not yet discovered correlations between parameters. These factors necessitate the use of cluster analysis methods, which would be capable of operating within the conditions of uncertainty and parameter redundancy.Objective. The goal of the research is extending the capabilities for analyzing changes in a system’s state in time by accounting for uncertainties present in observations data.Method. An application of the interval fuzzy cluster analysis method for investigating changes in data set characteristics in time, and for revealing general trends, is proposed. Formalizing the technological problem faced by the research in terms of Data Mining provides for a possibility of simultaneously processing multiple input vectors. A step-by-step algorithm for interval evaluation of the state of a natural system based on historical observations data and current values is developed.Results. The proposed model is adapted for solving the technological task of an artesian well monitoring, and its capabilities for revealing hidden patterns on early stages are demonstrated experimentally.Conclusions. Interval fuzzy cluster analysis allows taking into account and modeling uncertainties of any given nature, which may occur in artesian well research data on different stages of monitoring. It is shown, that concurrent input of multiple wells data may allow to evaluate not only there position against the standard compact classes according to (potential) water quality, but also their position against each other, and eventually indicate a previously unknown pattern.