Makarova I.L., Ignatenko A.M., Kopyrin A.S. —
Detection and interpretation of erroneous data in statistical analysis of consumption of energy resources
// Software systems and computational methods. – 2021. – ¹ 3.
– P. 40 - 51.
DOI: 10.7256/2454-0714.2021.3.36564
URL: https://en.e-notabene.ru/itmag/article_36564.html
Read the article
Abstract: Monitoring and analysis of consumption of energy resources in various contexts, as well as measuring of parameters (indicators) in time are of utmost importance for the modern economy. This work is dedicated to examination and interpretation of the anomalies of collecting data on consumption of energy resources (on the example of gas consumption) in the municipal formation. Gas consumption is important for the socioeconomic sphere of cities. Unauthorized connections are the key reason for non-technological waste of the resource. The traditional methods of detection of stealing of gas are ineffective and time-consuming. The modern technologies of data analysis would allow detecting and interpreting the anomalies of consumption, as well as forming the lists for checking the objects for unauthorized connections. The author’s special contribution lies in application of the set of statistical methods aimed at processing and identification of anomalies in energy consumption of a municipal formation. It is worth noting that the use of such technologies requires the development of effective algorithms and implementation of automation and machine learning algorithms. The new perspective upon time-series data facilitates identification of anomalies, optimization of decision-making, etc. These processes can be automated. The presented methodology tested on time-series data that describes the consumption of gas can be used for a broader range of tasks. The research can be combined with the methods of knowledge discovery and deep learning algorithms.
Ignatenko A.M., Makarova I.L., Kopyrin A.S. —
Methods for preparing data for the analysis of poorly structured time series
// Software systems and computational methods. – 2019. – ¹ 4.
– P. 87 - 94.
DOI: 10.7256/2454-0714.2019.4.31797
URL: https://en.e-notabene.ru/itmag/article_31797.html
Read the article
Abstract: The aim of the study is to prepare for the analysis of poorly structured source data, their analysis, the study of the influence of data "pollution" on the results of regression analysis. The task of structuring data, preparing them for a qualitative analysis is a unique task for each specific set of source data and cannot be solved using a general algorithm, it will always have its own characteristics. The problems that may cause difficulties when working (analysis, processing, search) with poorly structured data are considered. Examples of poorly structured data and structured data that are used in the preparation of data for analysis are given. These algorithms for preparing weakly structured data for analysis are considered and described. The cleaning and analysis procedures on the data set were carried out. Four regression models were constructed and compared. As a result, the following conclusions were formulated: Exclusion from the analysis of various kinds of suspicious observations can drastically reduce the size of the population and lead to an unreasonable decrease in variation. At the same time, such an approach would be completely unacceptable if, as a result, important objects of observation are excluded from the analysis and the integrity of the population is violated. The quality of the constructed model may deteriorate in the presence of abnormal values, but may also improve due to them.