In many commercial sectors, the analysis of time series data has become a pressing issue. This encompasses tasks ranging from forecasting currency metrics to identifying failures in industrial equipment. Such demands underline the relevance of developing tools for the creation of automated time series analysis systems. This paper proposes the use of an AutoML approach to automate the development of machine learning models employed in the automatic analysis of time series data. The analysis is suggested to be conducted on the basis of two models: one aimed at time series forecasting and the other at time series classification. This combination of models enables the resolution of a majority of practical problems encountered in the field. The automation of structural-parametric synthesis of the models and optimization of hyperparameters is demonstrated using these models. The presented system is implemented within the AutoGenNet software platform. This platform embodies the No-Code development concept, enabling users to circumvent the complexities inherent in the processes of model creation and training. The adoption of the No-Code development paradigm contributes to lowering the entry barrier for working with the software. Additionally, a mechanism for generating software wrappers for the utilization of trained models has been implemented within the AutoGenNet platform. Collectively, these advancements facilitate the application of the AutoML approach for automating the processes of model generation and training, thereby simplifying and accelerating the resolution of tasks associated with automatic time series analysis. The developed system is scalable and can be utilized for the automated generation of additional time series analysis models to address a wide range of practical challenges.