Selecting the optimal forecasting method. Classification of forecasting methods

25.09.2019

Differences in the nature of the forecasted objects, as well as in the timing of forecasting, the degree of completeness and reliability of the initial data, predetermine the use of different forecasting methods. The specificity of the methods is reflected in the sequence and content of the work to compile the forecast.

The entire set of forecasting methods, primarily according to the degree of their homogeneity, can be broadly divided into groups of simple and complex methods (Fig. 7.3).

Group of simple methods combines forecasting methods that are homogeneous in content and tools used (morphological analysis, extrapolation of trends, etc.).

Complex methods reflect aggregates, combinations of methods, most often implemented by special forecasting systems (forecast graph method, Pattern method, etc.).

Depending on the nature of the information on the basis of which the forecast is made, all forecasting methods are divided into three classes: factual, expert and combined.

Factual methods based on factual information material about the past and present development of the forecast object. These methods are most often used in exploratory forecasting for evolutionary processes and are attractive for their relative simplicity and objectivity. However, if unforeseen constraints appear that hinder the development process, the use of these methods may lead to errors in forecasts. When choosing these methods, it should be taken into account that they are applicable subject to the following conditions:

– the probability of maintaining the factors that determined the development process in the past is greater than the probability of their change;
– the probability of the combined influence of all these factors on development in the previous direction is greater than the probability of its change.

The reliability and accuracy of factual methods can be increased by combining them with expert methods.

Among the factual methods, a group of statistical (parametric) and a group of advanced methods are distinguished. The first of them includes methods based on the construction and analysis of dynamic series of characteristics (parameters) of the forecast object. Among them, the most widespread are: extrapolation,

Rice. 7.3.

interpolation, method of analogies (similarity models), parametric method.

The second factual group consists of methods based on the use of the property of scientific and technical information to advance the implementation of scientific and technical achievements in practical activities. One of the methods belonging to this group (publication) is based on the analysis and assessment of the dynamics of publications.

Another method (patent) is based on the assessment of inventions and studies of the dynamics of their patenting. Based on the dynamics of patenting intensity, one can evaluate and predict the development of a particular direction.

In contrast to factual forecasting methods, expert methods are based on expert judgment. They can be divided into groups according to the number of experts involved and the availability of analytical processing of examination data.

Let's consider the main methods of research (search) forecasting.

There are methods of extrapolation and interpolation of development trends. The basis of extrapolation is the analysis of time series, which represent time-ordered sets of measurements of the main characteristics of the object under study. Methods of predictive extrapolation include: trend extrapolation, extrapolation of envelope curves, correlation dependencies, etc. A trend is an analytical or graphical representation of the change in a variable over time, obtained as a result of isolating the regular (systematic) component of a time series. The time sequence of retrospective values of a forecast object variable is called a dynamic series, consisting of deterministic and stochastic components.

The deterministic component (or trend) characterizes the natural dynamics of the development of the object as a whole, and the stochastic component reflects random fluctuations (noise) in the process of functioning of the object over time. The task of forecasting is to determine the type of extrapolating functions X t and e t based on initial actual data about the object.

When using extrapolation to form a forecast, it is necessary to take into account the effect of external factors, since the extrapolation function reflects the development trend in the past and present, but it may not always persist in the future. Extrapolation of trends without taking into account the expected impact of active forecast background factors is called “naive extrapolation”.

Extrapolation is carried out in two stages: at the first, the type of function that describes the empirical series is selected; at the second stage, the parameters of the selected extrapolation function are calculated.

Extrapolation can be used not only to predict parameters individual species products, but also to determine the prospects for the development of a wide class of systems. In this case, they resort to using the envelope curve method. Envelope curve- this is a line that at all points has a common tangent with one of the lines of the family of development of characteristic parameters of the system, the analysis of changes in which makes it possible to effectively predict the development of the entire system.

The most difficult thing is to determine the characteristic parameters of the system. One of the possible ways to overcome this difficulty is to divide all parameters into intensive and extensive, depending on which factors of development of a given system will predominate - internal or external.

Examples intensive parameters may be:

– ratio of input to output (efficiency of conversion of matter, energy, information transfer, growth rate of products, etc.);
– functional characteristics (cleaning efficiency, limits of change in physical and chemical parameters, labor intensity, etc.).

Extensive options may include characteristics such as weight, product dimensions, production volume, profit margin, volume of industrial discharges and emissions. The reliability of the assessment method can be increased in combination with other methods (for example, expert assessment).

Along with various options extrapolations in forecasting are also used interpolation method. It is used in the case when unknown intermediate values are determined from the known initial and final values of the desired characteristic of an object. A classic example of using the interpolation method is D. I. Mendeleev’s prediction of the appearance of new chemical elements based on previously opened and installed functions periodic law distribution of elements.

Forecasting by analogy consists in establishing equivalence (similarity) between compared objects, when, based on a study of the behavior of one system, the behavior (development) of another system is predicted.

The use of analogy for forecasting purposes presents certain difficulties. It is necessary, first of all, to identify elements of continuity and criteria for the similarity of compared objects, as well as to analyze the identity of the action of environmental factors (active background). Based on the analysis of a similar model and the predicted object, the possibility (conditions and restrictions) of using the similarity model when developing a forecast is established.

Expert methods (individual and collective), which are based on the judgments of specialists regarding the prospects for the development of objects. These methods are based on the mobilization of professional experience and intuition. Typically, expert methods are used when objects that cannot be formalized mathematically are analyzed.

Method idea generation based on the activation of psycho-intellectual activity. This method can have several varieties. As a method of individual expert assessment, it involves identifying an expert's opinion using programmed control, including access to human memory or a computer storage device.

Another option is collective idea generation. The promotion of ideas is activated with the direct participation of a team of experts. The purpose of the method is to obtain large number original ideas in a short period of time. To ensure an atmosphere of free creativity, conducting brainstorming sessions requires adherence to a number of rules: inadmissibility of criticism, concise expression of ideas, strict adherence to regulations, continuation and development of ideas previously expressed by other experts, listening to any opinions (even fantastic ones at first glance). The composition of the expert group is 5–7 (up to 10) people. Discussion duration is 30–40 minutes. The main problem when using this method is the formation of a group of experts. Most often they are staffed by specialists different profiles, but prone to generating ideas. Brainstorming is based on the use associative method of thinking.

Morphological analysis– an expert forecasting method, which belongs to the group of individual ones and is aimed at identifying possible options for the development of a forecast object by constructing a matrix of the object’s characteristics and their possible values, followed by enumeration and evaluation of combination options. This method can serve as a tool for predicting inventions that have not yet been made or ordering likely inventions by the time of their appearance.

The main provisions of morphological analysis are the need to:

– accurately formulate the problem to be solved;
– determine the necessary means to carry out specified functions;
– identify the characteristic parameters of the object;
– classify each of the parameters into ranges or modes and draw up a morphological map (matrix);
– generate options in the form of chains of map elements through combinations that include one element from each row of the matrix;
– evaluate the effectiveness and feasibility of implementing options.

If we depict all the characteristics of an object (parameters) and alternatives for each of them in matrix form, then design options can be represented by lines passing through only one element in each row.

Method "goal tree"can be considered the most effective expert method of normative forecasting. Its main task is to link distant goals with actions that need to be taken in the present. The “tree of goals” is a hierarchical structure reflecting cause-and-effect relationships between elements (goals).

Logically related to the “goal tree” method matrix method based on the use of matrices reflecting the influence of factors on the forecast object. When constructing matrices, a number of sequential actions are carried out:

– identify and group factors influencing the achievement of the goal, according to the nature of their contribution;
– identify homogeneous complexes of factors;
– determine the influence of complexes on each other;
– determine the full impact of each factor on achieving the final goals.

Among expert forecasting methods that involve the use of the opinions of groups of experts, the most famous are the Delphi method and the method of collective expert assessments.

Delphi method (Delfi) - one of the common methods of drawing up a picture of the future - consists of generalizing and statistically processing expert opinions regarding the prospects for the development of those areas in which they specialize, as well as related ones.

The essence of the method is that the survey of experts is carried out in several rounds (using special questionnaires). Each round refines previous answers and gradually leads experts to an agreed solution. Expert assessments do not require personal interaction and replace direct debate with a carefully designed program of sequential individual interviews.

The results of processing collective expertise are used when constructing a “goal tree” and when choosing ways to study (or implement) a problem.

Regardless of the level of analytical processing, all expert methods still suffer from subjective assessments. Thus, each of the groups discussed above has certain advantages and disadvantages. The desire to reduce forecasting errors and provide solutions to problems of a wide range (from formalized to non-formalizable) has led to the emergence of complex forecasting methods, most often implemented in forecasting systems.

Combined methods include methods with a mixed information base, in which both expert and factual information are used as primary information.

One such method is Pattern method– justification of planning through scientific and technical assessment of quantitative data. The forecast is carried out on the basis of a scenario - a summary review a certain number simulated trends in the development of real possible events. The work scheme includes several stages:

– construction of a prognostic scenario;
– development of search forecasts of the main characteristics of the forecast object;
– morphological analysis;
– development of a “goal tree”;
– assessment of elements of the levels of the “goal tree” according to the criteria of relative importance, usefulness, condition, development, etc.;
– development of options for optimal distribution of resources among the elements of the “goal tree” levels.

The scenario attempts to analyze, without quantitative assessments, the goals of the enterprise, the direction of efforts and tasks. The findings are used to create a “goal tree” and establish significance coefficients. At the same time, using the methods of extrapolation and envelope curve, a technological forecast is compiled at all levels.

When compiling a “goal tree”, two additional groups of characteristics are determined:

– coefficient of mutual utility (use of the results obtained in other areas);
– state of project readiness (research, search and technical development, design, finished product) and deadlines for work on systems and subsystems.

One of most important tasks forecasting should be considered an increase in the reliability, reliability, and accuracy of forecasts.

This article describes forecasting methods, their significance, classification and brief characteristics. The main criteria for selecting these methods are presented and examples of their effective practical application are given. Also underlined special role forecasting methodology in modern world increased instability.

The essence and significance of forecasting methodology

IN general concept forecasting is the process of predetermining the future based on initial parameters (experience, identified patterns, trends, connections, possible prospects, etc.). On a scientific basis, forecasting is used in a variety of areas of human activity: economics, sociology, demography, political science, meteorology, genetics and many others. Most illustrative example use of forecasting in Everyday life a person is familiar with the daily weather forecast.

In its turn, efficient use forecasts on a scientific basis require the use of certain techniques, including whole line forecasting methods. At the beginning of the last century, when scientific research in this area began, only a few similar methods with a limited range of applications were proposed. On this moment There are many such methods (more than 150), although no more than a few dozen basic forecasting methods are practically used. At the same time, the choice of certain methods depends both on the scope of their application and on the goals of the forecast research being carried out, as well as on the availability of specific forecasting tools for the researcher.

Basic concepts in forecasting methodology

A forecasting method is a specific method aimed at studying the forecast object in order to obtain a target forecast.

Forecasting methodology is a general body of knowledge about methods, techniques and tools for making forecasts.

Forecasting methodology is a combination of methods, techniques and tools chosen to obtain a target forecast.

The object of forecasting is a certain area of processes within which research is carried out on the subject of forecasting.

Subject of forecasting - legal or individual, carrying out research work in order to obtain forecasts.

Differences and relationships between planning and the forecasting process

Forecasting versus planning:

is informative and not prescriptive;
covers not only the activities of a particular enterprise or organization, but the entirety of the external and internal environment;
may be of a more long-term nature;
does not require significant detail.

However, despite all the differences, forecasting and planning have a close relationship, especially in the economic field. The resulting target forecast shows the area of potential risks and opportunities, in the context of which specific problems, tasks and goals are formed that need to be solved and taken into account when drawing up plans of various forms (strategic, operational, etc.). In addition, forecasts provide the opportunity for an analytically sound, multivariate view of potential development, which is necessary for constructing alternative plans. IN in a general sense we can say that the relationship between forecasting and planning lies in the fact that although the forecast does not determine specific planning tasks, it contains the necessary information materials for effective target planning.

Basic classifiers in forecasting methodology

The main classification of forecasting methods is usually carried out according to the following criteria:

By degree of formalization:

intuitive, which are used for difficult to predict tasks using expert assessments (interviews, scenario method, Delphi method, brainstorming, etc.);
formalized methods, which mainly involve more accurate mathematical calculations (extrapolation method, least squares method, etc., as well as various modeling methods).

By the nature of the prognostic process:

qualitative methods based on expert assessments and analytics;
quantitative methods based on mathematical methods;
combined methods, including (synthesizing) elements of both qualitative and quantitative techniques.

According to the method of obtaining and processing information data:

statistical methods, implying the use of quantitative (dynamic) structural patterns for processing information data;
methods of analogies based on logical conclusions about the similarity of development patterns various processes;
advanced methods, characterized by the ability to build forecasts based on the latest trends and patterns of development of the object under study.

Also, the entire set of these methods can be divided into general forecasting methods and specialized methods. General methods include those that cover a wide range of solutions to prognostic problems in various fields life activity. An example of such forecasts can be expert assessments in various fields. On the other hand, there are methods that are focused only on a certain field of activity, such as, for example, the balance sheet method, which has become widespread in the economic sphere and is focused on accounting information.

Brief description of forecasting methods

As already noted, there are currently many methods in forecasting. The main forecasting methods include those that are currently most widespread and used in various fields.

Method Since in solving many forecasting problems there is often insufficient reliable formalized data, including mathematical data, this method is quite popular. It is based on the professional opinion of experienced experts and specialists in various fields, followed by processing and analysis of surveys conducted.
The extrapolation method is used for stable system dynamics of various processes, when development trends persist in the long term and there is the possibility of projecting them onto future results. This method is also used for objects of the same field of activity with similar parameters, assuming that the impact of certain processes on one object, causing certain consequences, will cause similar results in other similar objects. This forecasting is also called the method of analogies.
Modeling methods. The development of models is carried out on the basis of certain objects or systems, their elements and processes, with subsequent experimental testing of the constructed model and making the necessary adjustments to it. At the moment, predictive modeling methods have the widest range of applications in various fields from biology to the socio-economic sphere. In particular, the possibilities of this technique were revealed with the advent of modern computer technologies.
is also one of the main methods. It implies an approach to making forecasts focused on specific goals and objectives formulated by the subject of forecasting with the establishment of certain normative values.
The scenario method has become widespread in the development of management decisions that allow assessing the probabilistic development of events and possible results. That is, this method involves analyzing the situation with the subsequent determination of likely trends in its development under the influence of making certain management decisions.
Foresight methods. The latest methodology, which includes a whole range of different methods and techniques aimed not only at analyzing and forecasting the future, but also at shaping it.

Statistical forecasting methods

One of the main methods for making forecasts are statistical methods. Forecasts developed by such methods can be the most accurate, provided that the initial information data for the analysis of the necessary quantitative and semi-quantitative characteristics of forecast objects is complete and reliable. These methods are a form of mathematical forecasting techniques that make it possible to build promising time series. Statistical forecasting methods include:

research and application of modern mathematical and statistical methods for making forecasts based on objective data;
theoretical and practical research in the field of probabilistic and statistical modeling of expert forecasting methods;
theoretical and practical studies of forecasting in a risk environment, as well as combined methods of symbiosis of economic-mathematical and econometric (including formalized and expert) models.

Supporting tools for forecasting methodology

The auxiliary tools of heuristic forecasting methods include: questionnaires, maps, questionnaires, various graphic material and so on.

The tools of formalized and mixed methods include a wide range of tools and techniques of auxiliary mathematical apparatus. In particular:

linear and nonlinear functions;
differential functions;
statistical and mathematical tools for correlation and regression;
least square method;
matrix techniques, apparatus of neural and analytical networks;
apparatus of the multidimensional central limit theorem of probability theory;
fuzzy set apparatus, etc.

Criteria and factors for choosing certain methods when making forecasts

The choice of forecasting methods is influenced by various factors. Thus, operational tasks require more operational methods. At the same time, long-term (strategic forecasts) require the use of forecasting methods of an integrated, comprehensive nature. The choice of certain methods also depends on the scope of application, the availability of relevant information, the possibility of obtaining formalized (quantitative) assessments, the qualifications and technical equipment of forecasting subjects, etc.

The main criteria for the methodology can be:

systematic nature in the formation of forecasts;
adaptability (variability) to possible parametric changes;
validity of the choice of methodology in terms of reliability and relative accuracy of the forecast;
continuity of the forecasting process (if a one-time task is not set);
economic feasibility - the costs of implementing the forecasting process should not exceed the effect of the practical application of its results, especially in the economic sphere.

Examples of effective use of existing prognostic tools

Effective practical application of forecasting methods, the most common example of which at the moment is their use in the business environment. Thus, the most progressive companies no longer do without making forecasts when carrying out full planning of their activities. In this context, forecasts of market conditions, price dynamics, demand, innovation prospects and other prognostic indicators, including seasonal climatic natural fluctuations and socio-political climate, are important.

In addition, there are many examples of the effective application of forecasting methodology in various:

use of mathematical modeling to predict potential emergency situations at hazardous enterprises;
systemic environmental and economic forecasting across the country and regions;
socio-economic forecasting of development trends of society as a whole and its individual elements;
forecasting in the field of quantum physics, new biotechnologies, information technologies and many other areas.

The role of forecasting methodology in the modern world of increased uncertainty and global risks

In conclusion, it must be said that forecasting methodology has long been fully integrated into human life, but it is becoming most relevant precisely in our days. This trend is associated both with the rapid development technological processes in the world, and with increasing uncertainty in the internal and external environment. Numerous crisis phenomena in the economy, politics, social sphere provoke an increase in risk load in all areas of activity. The deepening of globalization processes has led to the emergence of systemic global risks generating a possible domino effect, when problems in individual corporations or countries have a serious negative impact on the economic and political state of the entire world community. also in Lately The risks associated with natural and climatic instability, major man-made disasters, and military-political crises have intensified. All this indicates the special role of forecasting both potential global and current individual risk phenomena in the modern world. Effective system forecasting that responds to modern challenges can make it possible to avoid or reduce the consequences of many threats and even transform them into advantages.

economic forecasting methods understand a set of techniques, assessments and methods for studying economic processes that make it possible, based on an analysis of past (retrospective) internal and external connections in the system or their changes, to provide for its possible (probable) development in the future.

The choice of forecasting method is based primarily on the need to ensure functional completeness, reliability and accuracy of the forecast, as well as the need to reduce time and cash to carry out the process of forecasting economic development. It depends on the following factors: the purpose of the forecast, its objectives; the period for which the forecast is generated; specificity of the forecasting object (namely its dynamic characteristics of the market operating environment, complexity, scale); reliability, completeness and nature of the initial information about the forecast object; limiting forecasting factors (resources, algorithms, programs, etc.); requirements for forecasting results.

All of these factors influencing the choice of a forecasting method must be considered in systemic unity and a certain sequence defined in relation to the forecasting object. But this does not mean that all the factors listed above must necessarily be taken into account. If some of them are determined to be unimportant under the conditions of a particular predicted object (phenomenon), then they can be removed from consideration or not taken into account.

systems of methods that are used in forecasting economic development lay down certain classification criteria. The most common feature is the degree of formalization, according to which forecasting methods are divided into intuitive and formalized. The second classification feature can be called general principle actions of forecasting methods (here a distinction is made between expert, factual and combined methods). According to the method of obtaining forecast information, forecasting methods are divided into statistical, expert assessments, analogies, and modeling.

The classification of methods, in addition to the necessary systematization of the forecasting procedure, should provide the opportunity comparative analysis and choosing the most appropriate forecasting method.

Expert forecasting methods used when the information array that characterizes the development of the economy in the past is insufficient or non-existent. Expert methods are based on the use of assessment (intuition) of specialist experts regarding the prospects for the development of economic processes in the future. According to the principle of operation, expert methods are divided into: a) methods of individual expert assessment and b) methods of collective expert assessment.

Methods of individual expert assessment These include forecasting methods based on the use of the judgment (assessment) of one or more expert specialists as a source of information. At the same time, experts must be competent in the field of activity subject to forecasting. Such methods include questionnaires, analytical and memos, script writing methods, morphological analysis, interviews, discussions, psycho-intellectual generation of ideas. The most common are interviews, the analogue method, and the script writing method.

Methods of individual expert assessment are still the most accessible, but not the most reliable ways to develop a forecast due to significant subjectivity and dependence on the professionalism and competence of the expert. The use of collective expert assessment methods is an attempt to increase the degree of objectivity of expert opinions and increase the reliability of group discussion.

Methods of collective expert assessment - these are methods based on identifying a generalized objective assessment of an expert group by processing individual assessments, which can be carried out through both direct (survey) and indirect (questionnaire) contact with experts. These methods include: the method of expert commissions, the method of collective generation of ideas, the method round table, Delphi method, heuristic forecasting, synoptic methods, matrix models.

Factual methods used when the information base about the past development of the object is sufficient and complete. Factual methods include statistical methods, analogy methods, predictive modeling methods.

Statistical methods are based on the analysis of time series, which make it possible to outline general trends in the development of the object. There are time series with a stable tendency (trend), with an unstable tendency, with an absent tendency. When forecasting indicators with a persistent trend, the following methods of mathematical statistics are used: predictive ectrapolation, moving average, least squares method, and adaptive exponential smoothing, harmonic weights. To predict time series with an unstable and absent trend, the most appropriate methods are correlation-regression analysis, probabilistic modeling of Markov chains, methods based on functions with a flexible structure, neuron or network.

Method of analogies consists in transferring a previously established development model of an analogue object to the predicted object. They are used when the forecasting object is considered as a physical and full-scale model of an analogous object, and the goals and objectives of forecasting correspond to the goals and objectives of the development of an analogous object. The most used analogy methods in forecasting economic and social development is a mathematical, historical and structural analogy.

Predictive modeling methods consist in constructing a conditional image of the forecasting object, reflecting its characteristics in the real environment, which are essential for the purpose of forecasting. In forecasting, such a model replaces an object that does not exist, and therefore its purpose is to build a possible information image of the predicted object and the processes occurring in it. When forecasting, the model becomes the only tool for testing the concept of the future for assumptions and determining the boundaries of possible development trajectories, that is, the model connects the information image of the modern with a theoretical reflection of the future. In the practice of economic forecasting, the following types of models are most often used: structural, simulation, network, statistical, economic-mathematical, factor, econometric.

It should be noted that in the world practice of macroeconomic forecasting, most predictive models of socio-economic development are developed using econometric tools. Economic mathematicians are working to create an analytical apparatus suitable for making forecasts. Large-scale econometric models have a multi-purpose purpose and are used both to predict economic variables (unemployment level, price growth rate, national currency exchange rate) over a certain period of time, and to identify trends in the development of the national economy as a whole. On their basis, it is possible to assess the likely consequences of certain state and political decisions and the impact of the global factor on the national economy. The most authoritative in Western countries are the models developed by specialists from the Wharton Business School (the origins of this direction were L. Klein, the Data Resources corporation (A. Ekstein) and Chase Economics (M. Bvans)). Today, the reliability of econometric modeling is complemented by non-mathematical forecasting tools, in particular expert estimates. One of the perfect forms of such assessments is the Delphi method. Another condition for the reliability of forecast models is scenario forecasting.

For a forecast to be correct, it must be reliable. The reliability of the forecast can only be assessed when the predicted phenomenon occurs. In this case, two problems arise: how to assess the quality of the forecast before its implementation and whether the forecast can be considered reliable if it did not come true. It is impossible to give an unambiguous answer to these questions, since everything depends on what decision was made based on the developed forecast and the controllability of the situation in which the object operates. To assess the reliability and accuracy (validity) of a forecast, the concept of verification and forecast quality is used. Examination is a set of criteria, methods and procedures that allow, on the basis of multilateral analysis, to evaluate the quality of the resulting forecast. The quality of a forecast is a set of characteristics of the forecast that together make it effective and useful in management, ensuring a reliable description of the object for a certain future and the possibility of reliable use of forecast results for the management procedure. The concept of forecast quality is considered in two ways: within the framework of the forecast itself and according to the results of using the forecast for management purposes.

Currently used forecast verification methods predominantly operate on purely statistical procedures, which boil down to estimating confidence intervals of calculated forecast values. In this case, two types of errors are provided: errors caused by information or description of the object, and errors in the direct choice of the forecasting method. The total forecast error can be calculated by adding all possible errors, namely: information errors (errors in obtaining and processing - Qd, errors in choosing a forecasting method, technology for its implementation (5 m), errors in computational procedures (£> 0), errors of a subjective nature (5 C), errors in the appearance of unforeseen changes in the forecasting object (5 e):

Verification is most appropriate at the final stage of forecast development. When using simple, uncomplicated methods for developing an economic forecast, expert surveys are most often used for verification. In more complex forecast calculations, it is necessary to use a special verification procedure, which covers the following actions: 1) development of a forecast using other alternative methods (direct verification); 2) comparison of forecast indicators with those obtained from other sources of information (indirect verification); 3) checking the developed forecast using a retrospective period (inverse verification); 4) analytical or logical derivation of a parallel forecast from previously obtained forecasts (sequential verification); 5) additional survey of experts and comparisons with the conclusions of competent specialists and forecasters (verification by an expert); 6) refuting critical remarks of opponents (verification by an opponent); 7) identification and recording of possible errors (error verification); 8) construction of conditional submodels equivalent to the design full model in typical situations or environments (partial target verification). It should be noted that there are no perfect forecasts from the point of view of their realism, therefore the problem of verification is relevant and important, since it makes it possible to approximate the forecast and actual (as a result of their implementation) value of the economic processes under study.

Administrative methods can be used to prevent unreasonable expenses (theft, abuse).

Opportunity costs and the problem of economic choice. Production possibility curve.

The choice of forecasting method is an auxiliary but key decision in forecasting. This solution, on the one hand, should ensure the functional completeness, reliability and accuracy of the forecast, and on the other hand, reduce the cost of time and money for forecasting.

The relevance of developing formal, including logical, procedures for selecting a type or a forecasting method itself increases under the influence of three

groups of reasons.

The first group of reasons is the growth in the number of forecasting methods generated by the variety of practical forecasting problems. Currently, the number of forecasting methods is approaching two hundred. Due to the increasing complexity of forecasting problems and conditions, the number of methods will undoubtedly increase. Therefore, even a brief acquaintance with the essence of known forecasting methods by going through them will require a lot of time and effort.

In a market economy, and even more so in a transition economy, a practicing manager may simply not have such time. Therefore, to facilitate the task of choosing a method, it is necessary to divide all forecasting methods into types. In this case, the given classifications of forecasting methods can be actively used.

The second group of reasons is that the complexity of both the problems being solved and the objects of forecasting is constantly increasing.

The third group of reasons is associated with the increasing dynamism (mobility) of the market environment, accelerating the rate of obsolescence of goods, services and the goods and services that produce them.

Therefore, the choice of forecasting method is influenced by:

1) the essence of the practical problem to be solved;

2) dynamic characteristics of the forecast object and the market environment;

3) the type and nature of the information available, a typical representation of the forecast object;

4) a combination of phases of the life cycle, market cycle or development cycle (or improvement) of a product or service, as well as the industrial enterprise that produces them;

5) the lead time period and its relationship with the expected duration of the market, life cycle, development or modification cycle of a product or service;

6) expected type of management: traditional, systemic, situational, social and ethical management;

7) requirements for forecasting results and other circumstances of a specific problem.

Moreover, all of the mentioned factors influencing the choice of a forecasting method should be considered as a systemic unity, but taking into account their significant number in a certain sequence, which does not necessarily have to coincide with the one given.

Factors recognized as unimportant under the conditions of a specific task may be excluded from consideration. For example, when developing high-tech mechanical engineering samples, the forecast lead time should be selected as the minimum from the forecast horizon of the market state and the development cycle of a new product. If there is no statistical information, then the type of forecasting method can be selected from the following set: forecasting by analogy, functional-logical forecasting, expert forecasting.

As a result of pre-forecast research, the forecaster must structure information about the forecast object, analyze it and decide on which method to use. to a greater extent corresponds to the specific conditions of the forecast and (or) plan. At the same time, it is important, at the stage of preparing a decision on choosing a forecasting method, to identify both those methods whose use is possible in the conditions of the problem being solved, and those methods that cannot be used. The latter are excluded from the number of alternatives considered.

A typical representation of the forecast object can play an important role in this. This is explained by the fact that each of the standard representations is associated with a number of elements of the methodological forecasting environment: forecasting and planning methods. This is reflected by the Boolean (logical) matrix of the presence or absence of a connection between the standard representation and the object prediction method (Table 3.1.)/3/.

In the absence of the desired connection, the type of methods or method of forecasting or planning cannot be applied when this type representation of the forecasting object. Such a connection exists if there is a “1” at the intersection of a row and a column, and does not exist if an “O” is at the intersection of a row and a column.

The rows of this matrix are numbered from 1 to 6 and correspond to:

1 - unconscious (intuitive) idea;

2 - subject representation (description) of the forecasting object in natural language (known descriptive models);

3 - functional decomposition representation;

4 - representation in the form of service contours;

5 - aggregative-decomposition representation;

6 - representation in the form of a “parameter - tolerance zone” model.

Table 3.1. Boolean (logical) matrix of the presence or absence of a connection between a typical representation and a group methods forecasting.

№

The columns of this matrix are numbered in accordance with the numbers of forecasting types:

1 - expert forecasting;

2 - functional-logical forecasting;

3 - structural forecasting;

4 - parametric forecasting;

5 - forecasting by analogy;

6 - complex forecasting systems.

You should also pay attention to what management method (type of management) is supposed to be used. This is explained by the fact that different types of management have different requirements for the type of results (qualitative or quantitative) and forecasting accuracy. It must be remembered that, as established in Chapter 1, all decisions of an entrepreneur or manager are predictive in nature, i.e. predictability is a fundamental property of any decision.

Traditional management. There are no explicit requirements for the type of results (qualitative or quantitative) and forecasting accuracy with this type of management. This is due to the fact that “by default” it is assumed that the consequences of the control action will be similar to those previously observed when managing other objects. Thus, traditional management uses forecasting by analogy.

System management involves the need to predict multiple elements of the problem or elements, solving the problem, as well as connections between them. Therefore, this type of management most often uses expert, functional-logical, structural forecasting.

Situational management implies the need to predict the consequences of decisions made. The result of such a forecast can be qualitative (worse, better or preferable, unacceptable, etc.) or quantitative. Therefore, this type of management should often use expert, functional-logical, structural or mathematical forecasting.

Social and ethical management implies the need not only to predict the consequences of decisions made, but also to assess the significance and (or) influence of these consequences on the state of objects falling within the sphere of influence of this decision. Assessing the significance of these consequences for the state of objects falling within the sphere of influence of the solution being developed allows us to classify the result of such an impact as acceptable, unacceptable, etc. The result of the forecast can be qualitative (acceptable, unacceptable, etc. state) or quantitative-qualitative in nature , when the quality of a state is determined based on the analysis of numerical values of parameters and their comparison with quantitative assessments of various types of states.

Moral and ethical management uses a forecast of personnel reactions to relevant impacts.

Stabilization management requires a prediction of the direction and rate of change in the parameters of the control object as a result of the corresponding influences.

Since the presentation of all almost 200 forecasting methods on the pages of this book is unrealistic, forecasting methods are divided into types. Each type of forecasting method is represented by the most commonly used methods. It is assumed that the scope of description of the forecasting method should provide it practical use, but does not claim to be exhaustively rigorous from a mathematical point of view. In this book, the types of methods of forecasting theory will be considered in the following sequence: expert forecasting, functional-logical and structural methods, mathematical methods of parametric forecasting, forecasting by analogy. Complex forecasting systems will also be considered here. This sequence of presentation is associated with the need to solve relevant problems in forecasting and with the information available about the forecasting object. The presentation of each type of forecasting method will be preceded by a brief description of the conditions for using the corresponding methods.

The choice of a forecasting method, on the one hand, should ensure the functional completeness, reliability and accuracy of the forecast, and on the other, reduce the cost of time and money for forecasting.

The relevance of developing formal, including logical, procedures for selecting a type or a forecasting method itself is increasing under the influence of three groups of reasons.

The first group of reasons is associated with the large number of forecasting methods generated by the variety of practical forecasting problems. Currently, the number of forecasting methods is over two hundred. Therefore, even a brief introduction to known methods predicting by simple search will require a lot of time and effort.

The second group of reasons is that the complexity of both the problems being solved and the objects of forecasting is constantly increasing. This especially applies to modern SES.

The third group is associated with increasing dynamism (mobility) of SES.

As a result of pre-forecast research, the forecaster must structure information about the forecast object, analyze it and decide which method is more suitable for the specific conditions of the forecast. At the same time, it is important, at the stage of preparing a decision on choosing a forecasting method, to identify both methods whose use is possible in the conditions of the problem being solved, and those that cannot be used. The latter are excluded from the number of alternatives considered.

The choice of forecasting method cannot be a matter of the subjective inclinations of the forecaster or group of forecasters and must be determined in accordance with objective selection criteria.

The criteria for choosing a method are:

the nature of the forecasting object, or the problem (task) solved in the forecasting process;

forecasting level, or management level (federal, sectoral, regional, municipal), for which forecasts are developed;

lead interval (long-term, long-term, medium-term, short-term);

forecast goals.

Problems vary in the degree of development and clarity of connections between the problems being studied and their consequences; identified factors and performance indicators.

There are four classes of problems that occur when solving forecasting problems.

1. Standard problems. The connections between a factor and a result are strictly determined; they can be expressed by functional equations and simple calculations (for example, labor productivity is equal to the ratio of production volume in constant prices to the number of employees).

2. Structured problems. The connections are probabilistic (stochastic) in nature, but differ high degree cramped conditions. When factors change, the result can be determined with a certain interval “from” and “to”, but it can also be determined unambiguously (for example, determining the growth rate of labor productivity depending on the rate of its capital-labor ratio). 3.

Weakly structured problems. They are characterized by a low level of close connection between the factor and the result. In this case, the effective indicator changes in a very large range of values “from” and “to”.

For example, determining the level of crop yield, which depends on factors such as weather conditions. 4.

Unstructured problems. The change in the effective indicator from the influence of the factor is difficult to predict. For example, the development of equipment and technology depending on the amount of funding, etc.

It is important to keep in mind that the class of problems depends on the object of prediction. For example, it is clear that forecasting the development of science and technology inherently refers to weakly structured problems, in contrast, for example, to forecasting the development of production. But this is in the general case. At the same time, the lead interval, i.e. forecast period can change the class of problems for the same object. Thus, forecasting changes in the volume of basic production assets depending on the volume of investment in the short term (1 year) refers to structured problems (class 2), and the same problem solved in the long term (up to 10 years) goes into the class of weakly structured problems (class 3), and in the long term period for 20 years and even more so in the long term (over 20 years) - into the class of unstructured problems.

If the problem is solved at the level of the organization (company) in the short term, it can be classified as standard problems (for example, calculating production capacity if information is available on the commissioning and withdrawal of capacities in the forecast year). It can also be classified as a weakly structured or even unstructured problem as the forecast period lengthens (lead interval) and the management level increases (for example, regional, sectoral or federal). Thus, when moving to a higher level of management and increasing the lead period, the degree of structure of the problem decreases.

To predict standard problems, identities (equalities) and economic-mathematical models are used. For structured problems, econometric and economic-mathematical models are used. For weakly structured problems - methods of expert assessments, the scenario method, it is possible to use econometric models. For unstructured problems - mainly logical methods, methods of expert assessments with a high degree of aggregation of variables, as well as simulation models.

Questions for self-control

Give a general description of the problem of choosing a forecasting method. 2.

What criteria determine the choice of method? 3.

Describe the standard problems of choosing a forecasting method. 4.

Describe structured problems of choosing a forecasting method. 5.

Describe weakly structured problems of choosing a forecasting method. 6.

Describe unstructured problems of choosing a forecasting method. 7.

Consider the algorithm of actions of a forecaster when solving a forecast problem.