Digitalization of technological and design decisions making processes in modern construction

. The results of the study can expand the implementation field of the “decisions tree” method. This method can be used in the design decision making processes and technological equipment options when choosing construction equipment for the production of earthwork. The study was carried out on the example of the “Voznesensky” Architectural Ensemble″ in Tyumen, hereinafter “Voznesensky AE”. Following goals were achieved: options for a traditional approach to determining the need for construction equipment were considered; models and methods of making managerial decisions were analyzed; the method of “decisions tree” was studied, advantages and disadvantages were identified. The “decisions tree” method ia successfully applied in a digital experiment and recommended for real implementation in the development of an algorithm for making design decisions and when choosing options for mechanizing the production of excavation work using the Deductor Academic Studio program at the construction of a real facility. Currently, there are no cases of direct application of the «decision tree» method in the management of projects in the Russian construction industry. There are no published research results to introduce the «decision tree» method in the processes of design and production decisions.


Introduction
In the modern world, it is difficult to imagine our life without information. It has become the main product and resource in almost all areas of activity. Information technologies have become an integral part of our existence. In the field of software development, a flexible development concept has appeared. This methodology allows one to cope with a high rate of environmental changes and create an actual product at the request of a customer in modern conditions of digital reality. At the moment, this concept has grown out the framework of traditional package software and is a promising alternative to the usual approach to project management.
In the traditional setting, the purpose of project management in construction is to achieve high technical and economic indicators by compliance with the terms for the construction of objects at rational costs of resources. The design approach to production process management has proven its effectiveness in practice and is successfully applied by leading world companies. Unfortunately, the project management in the Russian Federation is not widely used. It is a fairly new product for Russian leaders and managers. Having even a small list of planned work, a modern manager can use abundance of technologies. A variety of its capabilities force to think and build the technological process as efficiently as possible. In such a staging, an important factor in decision making, which will be unambiguously effective, is the formation and the most complete consideration of many possible alternative options.

Methods
The decision making issue attracts scientists and ordinary people from different areas of activity. An important aspects of the study is understanding, development and implementation of process of making effective and rational decisions.
The theory of decision making as a scientific direction was traced in the works of J. von Neumann. The essence of this theory was formed on the basis of formal logical and mathematical methods, but not on the basis of experimental data.
Most often, two directions are distinguished in the study of the decision making process:  normative, i.e. based on formalized methods, models;  descriptive -by studying the psychology of decision making.
Both approaches to decision making are necessary and complement each other. Most often, decision making takes place in conditions of uncertainty, which does not always allow one to build a formal model of searching for the optimal solution.
In the standard process of effective decision developing, it is necessary to have: 1) Information or raw data; 2) Professional knowledge that allows one to assess the situation or problem; 3) Knowledge of formal decision making rules.
Any decision making process implies the presence of following elements:  the one who has to make a decision;  situation variables, i.e. those data, the impact of which can still be influenced;  uncontrollable situation variables, the impact of which cannot be avoided.
Managed and unmanaged variables create an external environment for the problem:  some limiters of change of situation variables;  several options for the result or choice.
The entire decision making process can be thought of as a list of sequential actions that will eventually lead to the optimal decision. The decision making process is subject to a certain list of management operations. They can be combined into certain target complexes. Operations are performed in strict sequence and can be represented in the form of a diagram shown in Fig. 1.  The authors studied the processes of organization and production of construction and installation works at the construction of the object "Voznesensky AE". The study site for the production of works is located within the boundaries of residential development, in the cramped conditions of existing urban development. The organizational and technological scheme of construction assumes classical order of construction of objects. All stages of construction are supposed to be carried out by the in-line method.
The subject of the study includes theoretical and practical aspects of applying the decision tree method as a technology for choosing alternative design and technological solutions in project management on the example of a real construction object using the Deductor Academic Studio program.
To achieve the results presented in the article, following set of research methods was used:  theoretical method (scientific literature study);  private methods: modeling -creation of an algorithm ("decision tree") reflecting the processes of work with the help of information technology;  general scientific methods: system analysis and generalization of research results, comparison of the obtained results with expert opinion.

Discussion
The rapid development of information technologies has led to modernization in the methods of information collecting, processing and storing, which has made it possible to form databases that need to be analyzed. As a result, there has been a demand for methods helping to automatically explore such data. It is important when it is necessary to make several decisions under conditions of uncertainty and dependence of these decisions on certain events or characteristics, when each decision depends on the outcome of the previous one. One of the methods for automated analysis of such data arrays is the decision tree method. The first fundamental to the development of decision trees works were published in 1966. It was a book by E.B. Hunt, J. Marin and P.J. Stone "Experiments in Induction". The first ideas were seen in the works of Hoveland and Hunt of the late 50s of the XX century [1][2][3][4][5].
At the stage of choosing relevant alternatives, the decision tree method is one of the most popular methods. According to the authors of this article, the use of this method and its improved modifications, for example, for the construction industry, will allow participants in the construction industry to successfully implement projects, effectively analyze various methods of construction work, and promptly offer effective solutions in the process of project implementation.
A certain human decision may be unconscious, not every decision may be optimal. Making the right decision involves choosing from a variety of options.
The main feature of decision tree analytics is scenarios, which is a hierarchical tree of nodes. The node is launched, exposing all parent nodes, the node's renderers, and the information contained within the node. As in the classical structure, the nodes interact with each other at the level of the software kernel. Each node collects a set of initial data, and one or more data sets are available during processing and at the output (Fig. 2).
Decision trees allow one to solve a large number of problems using mathematical models and machine learning. As a result, the software package allows one to build a hierarchical structure based on the initial data that were entered into the program for its training. It also allows to visualize the effect and result of the tree, forms the structure of data or knowledge in an intuitive way. The decision tree is a graphical representation of a decision process or problem that can contain alternative solutions and the most likely payoffs for any combination.
The tree is formed from vertices -key states that imply a choice. Vertices (points) of decision-making are the moments of time when the choice of alternatives takes place. In a decision tree, each branch, which is either an existing course of action or a possible consequence of the chosen action, is divided at certain points into a collection of other branches. The vertices are interconnected through the branches of the tree, which represent connection between the decision point and various events (decisions), including random ones. Decision trees also have a probabilistic component. The probabilities of occurrence of events can be located on the branches of the tree, the sum of the probabilities at each decision point is equal to one. At the end of each branch, there is a quantitative expression of each alternative. Most often, the construction of a tree occurs from top to bottom, it is descending. It is also convenient to build a tree from left to right. The classic version is shown in Fig. 3.   Fig. 3. An example of a classical decision tree structure Decision tree allows one to make a choice from the available options, which will be the best outcome of an event or action. The tree operates on the basis of rules, which are depicted as a hierarchical, sequential structure, where each object corresponds to a single node, leading to a specific solution.
A decision tree rule is a logical construction with a main condition in the form "IF ... THEN ...". Moreover, the number of results is not limited by anything, the tree will be built until there are alternative options.
This method is suitable for the analysis of projects, the result of which has several development options. An important aspect of compiling a "tree" is the availability of reliable information, taking into account the probability and moment of occurrence of these events. In the course of the research, the reliability of the work of the decision tree method in the Deductor Academic Studio software package was verified. The results obtained were compared with the actual ones. To check the correctness of the choice of the final measure for certain parameters, the significance of the applied attributes was tracked and the rules for choosing decisions on a real construction site were visualized. This software positions itself as an advanced analytics platform that allows to create complete applied analytical solutions for business. Deductor technologies allow one to go through all stages from data consolidation to building visual models. The problem with most of these algorithms is the inability to process a large amount of input data, for which built-in programming languages are often used. The Deductor platform does not require from the user any special training in programming.

Results
Digital experiment. The output data set for the object under study was the parameter "Special measures for the installation of enclosing structures of the excavation edge". In geomorphological terms, the work site is confined to the left-bank floodplain terrace of the Tura River. Based on the results of pre-project engineering and geological surveys, a significant technogenic impact was found in the study area. The natural relief is disturbed, the soil-vegetative layer is not fully preserved. In connection with human economic activity, the natural runoff of atmospheric precipitation and infiltration has been disturbed. As a result, the following conditions were taken as key input data: groundwater level above the bottom of the excavation, soil temperature > 0° С, seasonal water rise is present, water inflow is weak, excavation depth < 5 m, cohesive soil, external factors -new construction.
The algorithm created in the program has proposed to use following measures: freezing, sloping at a right angle and an angle < 90°, chemical fixing and grouting. Based on the technical and economic indicators, slopes were actually made at an angle of < 90°.
Theoretically, out of all possible measures, the program has proposed to carry out following measures: installation of gabions, open drainage and perimeter drainage.
In fact, dewatering measures were taken at the work site: open drainage and perimeter drainage. The arrangement of stable slopes at an angle of <90°. The arrangement of gabions was theoretically possible, but the event was not carried out based on the preferences of the customer. These activities in the proposed design solutions can be clearly seen in the decision tree. As a result of the digital experiment, out of the three options proposed by the program, two activities were implemented in practice in real conditions. The main results were published in [7][8][9][10][11].

Conclusion
Based on the expert method for evaluating the decisions made, we obtained a traditional set of special measures and construction equipment for earthworks at the site "Voznesensky Architectural Ensemble", located in the quarter of Shcherbakov, Zaozernaya, Krasnoarmeyskaya, Beregovaya streets in Tyumen (hereinafter -"Voznesensky AE") . The used Deductor Academic Studio software package made it possible to obtain machine visualization of decision making based on the given initial data. As a result, an algorithm was developed for making design decisions for the production of earthworks, namely, for the installation of enclosing structures for the edge of the excavation, and a "decision tree" was created.
It was found that the use of databases formed on the basis of previously developed and implemented projects, expert assessments and previous experience, transferred to an effective machine learning model, makes it possible to obtain the most rational design solutions and technological options for choosing equipment. The use of this method and software can be developed for use at all stages of design activities in construction.
The figurative model of the process of erecting the object under study at the stage of earthworks, depicted in the form of a decision tree, turned out to be intuitive and simplified the understanding of the problem being solved. At each decision node, one can clearly see what decision the model makes, where errors or inaccuracies come from, and what data will affect the outcome of the final decision.
The created decision tree algorithm does not require the user to select input data, it is enough to specify all existing ones and the algorithm itself will select the most significant of them. The obtained results of the algorithms are easily interpreted by the user.
The method has high self-learning and forecast accuracy at the level of statistics and neural networks. It was found that the algorithms for constructing decision trees embedded in the methodology have the possibility of special processing of missing values. Decision trees work equally well with both numeric and categorical data types. Decision trees, unlike many methods, build non-parametric models and are capable of handling categorical values. Thus, the created decision trees are able to cope with such Data Mining problems, in which there is no a priori information about the type of dependence between the studied data [2,4].
However, the experience gained with the algorithm of the program used and the generated database allows the author to conclude that, due to the possibility of learning, decision trees are subject to retraining. It is necessary to more carefully monitor the signs and dimensions that, with the maximum weight, affect the construction of decision-making trajectories. Trees are also susceptible to class confusion, which means that a program can sometimes shift the correct decision process in favor of an unlikely or incorrect choice. The solution to this problem seems to be periodic balancing of data classes, their weights and the definition of class functions.
Nevertheless, it can be concluded that the decision tree method can be successfully applied at all stages of the project life cycle. This development will allow all construction participants to increase the speed of making design, operational decisions and rationally use technological resources in real projects.
According to the results of the study, the authors received a visualized result of the work of the algorithm for making design decisions for the installation of enclosing structures of the edge of the excavation during the construction of "Voznesensky AE". The used principles and algorithms for making design decisions can be applied in the production of other types of construction work throughout all stages of construction of facilities and management of construction projects. This digital experiment [9] showed that in order to obtain a fast and reliable solution, it is enough to have an initial (experimental) database (in our case, a table) and a proven software tool for compiling a decision tree, for example, AP "Deductor Studio Academic".