Digitalisation and the wide availability of machine learning techniques have made the widespread use of artificial intelligence in various fields of science and economics possible. However, applying different methods to datasets to find models that describe hidden dependencies in the data requires knowledge of programming and applied mathematics to test different machine learning methods and evaluate the results to select optimal models. In this sense, it is important to provide tools that can lower the entry barrier for applying artificial intelligence to such problems.

We have developed a multi-user application with a graphical interface designed to evaluate the effectiveness of different algorithms for solving regression problems. At present, regression methods from the scikit-learn package are available in the system, as well as a number of the author's multilevel methods, allowing in some cases to achieve better results when solving problems of predicting properties of compounds in the field of inorganic materials science. A distinctive feature of the system is its extensibility, which is achieved through the opportunity to connect new machine learning methods, their fine-tuning, and the availability of modes for estimating the efficiency of models with different cross-validation strategies.

The development of the presented open-source system contributes to the democratisation of artificial intelligence, lowering the entry barrier for users by using a no-code approach to automate the exploration of the best models for solving practical regression problems.

The developed system is demonstrated for solving the problem of lattice parameters prediction for melilites of A₂B₂CO₇ and AB₂C₂O₇ compositions. Thanks to the use of the system, it was possible to develop models capable of predicting the lattice parameters with an error not greater than 0.21 Å.