This article outlines regulatory legal acts, standards and the application of measurement procedures. The description of the norm of characteristic indicators (requirements) for each product, be it bread, milk, petroleum products, light or heavy industry products, etc. is regulated in technical regulations, and further in regulatory documents for products, where for each indicator there are usually references to test methods of this product characteristic. For testing products, a test method is always necessary. In most cases, standard methods of testing and measurements have already been developed for these purposes, which are described in domestic or interstate standards: ST RK or GOST, as well as in international IEC, ISO, etc. However, there are no standard methods for testing in some cases. The solution to the problem in this case is the development of its own measurement methodology, the so-called non-standardized methodology, where measurement errors, subject to compliance with all the requirements of the document, are guaranteed. The article describes the stages of development, certification, approval and application of both conventional methods and reference measurement procedures. The article contains references from the legislation of the Republic of Kazakhstan in the field of ensuring the unity of measurements to standards, regulatory legal acts and the Law «On ensuring the unity of measurements»

The present paper concerns the notion of weak o-minimality that was initially deeply studied by D. Macpherson, D. Marker and C. Steinhorn. A subset A of a linearly ordered structure M is convex if for all a, b Î A and c Î M whenever a < c < b we have c Î A. A weakly o-minimal structure is a linearly ordered structure M = áM, =, <, …ñ such that  any definable (with parameters) subset of  M is a union of finitely many convex sets in M. A criterion for equality of the binary convexity ranks for non-weakly orthogonal non-algebraic 1-types in almost omega-categorical weakly o-minimal theories in case of existing an element of the set of realizations of one of these types the definable closure of which has a non-empty intersection with the set of realizations of another type is found.

This article implements and analyzes machine-learning algorithms, for predicting carsprices. Predicting prices is one of the most challenging but interesting tasks. There are so many factors involved in the prediction - year of manufacture, condition, mileage, engine size, etc. All these aspects combine to make auto prices volatile and very difficult to predict with a high degree of accuracy. Machine learning techniques can uncover patterns and ideas that we have not seen before, and can be used to predict and classify data accurately and accurately. The choice of the proper data classification algorithm, which would be suitable for a given task, depends on the volume, quality and nature of the data, on the computing resources of the computer, and how you plan to use the result. Each classification algorithm has its own characteristics and is based on certain assumptions. Also requires practical skills. In practice, it is always recommended to compare the quality of at least several different learning algorithms in order to choose the best model for a particular task, since the most experienced data scientists will not be able to tell which algorithm is more efficient. Algorithms can differ in the number of features or samples, the noise level in the dataset, and whether the classes are linearly separable or not. Ultimately, the quality of the classifier, its computational and predictive power, depends on the underlying data intended for training the algorithm. The purpose of this article is to consider the stages of pre-processing training data, and show how machine learning in particular and information technology in general have succeeded in developing tools for modeling, designing, predicting, planning and decision support in the field of auto sales. This study proposes a hybrid approach to forecasting problems, that is, solving forecasting problems using statistical analysis and machine learning methods.

The paper [11] raised the question of describing the cardinality and types of approximations for
natural families of theories. In the present paper, a partial answer to this question is given, and the study
of approximation and topological properties of natural classes of theories is also continued. We consider a
cycle graph consisting of one cycle or, in other words, a certain number of vertices (at least 3 if the graph
is simple) connected into a closed chain. It is shown that an infinite cycle graph is approximated by finite
cycle graphs. Approximations of regular graphs by finite regular graphs are considered. On the other hand,
approximations of acyclic regular graphs by finite regular graphs are considered. It is proved that any infinite
regular graph is pseudofinite. And also, for any k, any k-regular graph is homogeneous and pseudofinite.
Examples of pseudofinite 3-regular and 4-regular graphs are given.

The paper outlines the methods, which improve the controlling process of separating methanol from water in the distillation column to produce crude oil products. Nowadays, many industries use PID controllers to control process variables like temperature, flow, pressure, level, which helps maintain good performances. However, PID controllers can have slightly bad performances in complicated control systems, such as in Multiple Input and Multiple Output (MIMO) systems; due to this, optimization methods of improving PID are considered. А tremendous amount of work has been done rеfіnіng, studyingаndimprоvіng the PID controlling techniquesand methods. However, PID still faces challenges in a variety of common control problems. This article represents NеuralNеtworkAlgоritmbаsеd PID cоntrollеr, whіchіsusеdtоcоntrоlthе separating process of methanol from water in the distillation column, due to Nеuralnеtwork’s good generalization results. The Wood and Berry mathematical Model was chosen as the main control object.