Improving the Methodological, Information and Software Technology Support for Statistics Using the Analytical Capabilities of Artificial Intelligence and Machine Learning

The purpose of the study. Development of a systematic approach to improving the system of official statistics and its methodological, information and software support using artificial intelligence and machine learning capabilities . Materials and methods. The article was prepared using materials from international statistical organizations, the community of developers of open implementations of HyperLogLog and Neo4j, lisa and lisa meta projects, the works of leading Russian and foreign scientists and specialists in the field of statistics, data analysis and artificial intelligence, as well as a graph model of meta-information and an algorithmic approximation of the contents of composite information units based on the extended probabilistic algorithm HyperLogLog (structure HllSet), a method for restoring tabular data and searching through metadata, as well as proximity metrics (Jaccard index, cosine similarity) . Results. A systematic approach to improving official statistics and its methodological, information and software support has been proposed: from clarifying the foundations of the general theory of statistics (through the measurement language) to a formal model of statistical information systems, including static and dynamic structures. The role of metadata as a single layer of data description and analysis results is shown; a graph model of meta-information and an algorithmic approximation of the contents of composite information units based on the extended HyperLogLog algorithm (HllSet structure) are proposed. Examples of tabular data recovery and metadata search, as well as proximity metrics (Jaccard index, cosine similarity) are demonstrated. The issues of integration of statistics with machine learning and artificial intelligence methods with an emphasis on solving the problems of official statistics, as well as the possibility of local deployment of the DeepSeek artificial intelligence system in national statistical services using the example of Rosstat were discussed . Conclusion. The approach proposed in the article clarifies the subject of the general theory of statistics through the concept of statistical measurement and systematic consideration of statistical information systems, combines static and dynamic structures, formalizes metadata as a graph and introduces effective content approximations (HllSet). This creates a methodological and technological basis for integrating statistics with machine learning and artificial intelligence and constructing geographically distributed management support systems. At the same time, job prospects include : - development of algorithms for identifying cause-effect relationships in the dynamic structure of statistical information systems ; - extension of the ontological metadata layer ; - empirical validation of similarity metrics for different data domains; industrialization of solutions based on container platforms ; - development of standards for the integration of regulatory reporting into the unified information space of official statistics .