Types of innovative forecasting methods

When developing an innovative strategy of the company, among other points, it is important to have the results of forecasts of scientific, technical and technological development of society and the market in a given direction.

With this in mind, the manager responsible for the decision must competently select the optimal composition of methods for forecasting scientific and technical prospects. At the same time, among the vast range of formalized and informal means of forecasting, he will have to make a choice that will be both the most economical and fully effective.

The role of predictive and its basic concepts

There is a widespread opinion that forecasts of scientific and technological development are the destiny of the state and the largest corporations. However, this is far from the case, and any company, developing its own technological and innovation strategies, should pay special attention to forecasting the future state of the economy, market, technical and technological platforms.

In the forecast, the key goal is to provide proactive information that will allow the LPR to formulate and make a strategically correct decision on the firm's package of innovative activities.

Scientific and technological progress (STP) is formed by the actions of the entire set of entities participating in its implementation, but the movers are economic entities, regardless of the form of ownership.

And a single company, being an element of the STP system, performs the development of forecasts that are part of the research stage preceding the program study of innovations and their planning. The quality and validity of the firm's strategies are directly related to the level of long-term and medium-term forecasting. Below you will find a scheme for managing the STP, divided into three stages.

Step-by-Step Control

The activity called forecasting has a theoretical basis – prognostics. By it we will understand the field of scientific knowledge, which studies and justifies the patterns, principles and methods of creating forecasts in various spheres of human activity, including technical and technological. In addition to the main tools, this field of science is also engaged in methods of using the obtained predictive results. Within the framework of this article, first of all, we will be interested in methods for forecasting innovations.

Predictive science as a science was formed in the last quarter of the last century, although its roots arose in the days of ancient Greek philosophy and logic.

The great thinker and physician of that time Hippocrates first voiced the main ideas of formulating medical diagnoses in his work with the same name "Predictive".

In the modern world, forecasting economic events, achievements of science and technology is given ever-increasing attention. The role of strategic management is also growing, and the cost of a mistake is becoming a critical factor in the life of society and business.

It should be understood that the anticipatory information and the forecasting itself are in the nature of a probabilistic development of events. Due to the randomness and stochastic nature of the processes under study, the information orienting the LPR cannot guarantee complete reliability and serve as an impeccable basis for decision-making.

Forecasts work if there is a continuous process of research, monitoring, retrospective analysis, short-, medium- and long-term observations and assumptions. Below in the diagram is a list of basic concepts and their brief definitions of predictiveness.

Classification of forecasting methods

The classification of forecasting methods serves the task of applying the optimal forecast model in terms of labor. Naturally, the goals, objectives, objects, the period of anticipation in the preparation of each forecast are different, so its typology is chosen taking into account a number of criteria that have developed over the history of prognostics.

The period of anticipation is chosen a period of time that is supposed to be sufficient to obtain information ahead of the event and make the right management decision. As a rule, the period of anticipation is the first classification feature in order among the grounds used. In addition, the following criteria are distinguished for dividing forecasting procedures by methodological types.

According to the degree of formalization, formalized and intuitive forecasting methods are distinguished.

On an information basis: factual and expert methods of forecasting. Quite often, combined methods are also used. The present base has significant overlaps with the feature in the degree of formalization, but

According to the problem-target criterion: search (research) type methods and normative methods, which are also called programmatic.

By the level of materialization of the creative result in the course of scientific and technological

The classification of forecasting methods is shown above in tabular form. In the following sections of this article, we will look at the individual methods of the major groups in this table. For innovation activity, the division of methods, determined by classification on a problem-target basis, is of particular importance. Studies based on search methods serve to establish trends in the development of the state of the object (technique, technology, level of management organization) taking into account their current dynamics.

The normative method allows you to find the features of the future states of the object in terms of existing targets. Naturally, the period of anticipation for normative and search forecasts is different, the time point of the forecasting horizon with the normative approach is located much further. According to the scale of the object under study, forecasts are divided into six categories.

NTP events of a planetary scale.
Country-regional scale of events in the development of science and technology.
The scale of cross-sectoral objects of research.
The scale of industry development.
Scale of product groups and production technologies.
The level of devices, methods of operation of the equipment and materials used.

From the point of view of the level of materialization of the results of innovative activity, the methods are divided into the following groups of forecasts:

in the field of fundamental scientific discoveries and inventions;
in the areas of application of new discoveries and inventions (forecasts of achievements of applied research);
pioneering solutions in the field of R&D and RPC results;
decisions of OCD and RPC followers and resulting diffusion innovation.
Forecasting methodology from the point of view of statistics

Mathematical methods of forecasting in the innovation process belong to the group of formalized factual (parametric) means of deriving forecasts. Statistical methods of forecasting make up the bulk of mathematical methods, so they are often considered identical, and the toolkit itself is called "mathematical-statistical".

Operating with factual information, that is, based on the numerical (quantitative) parameters of research objects, the methods of the statistical class use probability theory, the laws of large numbers and mathematical statistics themselves.

Statistical forecasting methods are divided into two large groups. The first of these includes so-called single-parameter methods that allow you to make a forecast based on time series. Such tools include:

least squares method;
exponential smoothing method;
moving average method;
autoregression models;
methods of S-shaped and envelope curves;

methods of analysis of publications (for the purpose of analyzing the dynamics of publications in the scientific and technical sphere and the dynamics of patenting).

All these tools are factual methods of forecasting, however, the last group of methods is often allocated to a separate type of toolkit. To assess the predicted state in the innovation process, extrapolation methods of trend analysis are most often used, which include the first five types of methods. Sometimes statistical forecasting methods are separated from extrapolation and interpolation agents. In such cases, one of the four main methods of mathematical statistics or a combination of them is used for statistical analysis.

Regression analysis.
Correlation models.
Factor analysis.
Analysis of variance.

Many quantitative forecasting methods use the above tools of statistical means of processing data arrays. And yet the main methods of forecasting NTP, the development of technology and technology rely on extrapolation methods of identifying trends. Exponential smoothing is one example of this approach. Consider a small example in which extrapolation is performed by the moving average method (see the diagram below).

moving average method

Example of charting to extrapolate a series of data using the moving average method
The initial information is a time series of companies that use the venture form (M) to finance innovative projects. Innovative activity, due to riskiness, often involves special forms, for example, ventures, for the success of the undertaking.

In practice, sometimes, based on the nature of the general population, it is difficult to form a representative sample, as, for example, in this case. However, extrapolation forecasts with a high degree of certainty can be formed using quantitative forecasting methods to approximate the series of dynamics, in particular, the exponential smoothing method and the moving average method.

These factual forecasting methods, together with a number of others, form a tool for smoothing experimental curves for forecasting purposes, their application depends on the specific conditions of analysis. For example, the extrapolation method based on summary curves, despite its simplicity, is limited to special parameters and forecast indicators.

A very popular way to obtain a smoothed curve is to use moving averages, which are formed by means of the same Excel. Next to the real curve, a smoothed line is built, on the basis of it and a forecast for 2018 is formed in our case.

Certain types of quantitative forecasting methods

Resorting to formalized forecasting methods based on the analysis of statistical data, we cannot but dwell on the means of regression analysis and the use of S-shaped forecast curves to assess prospects. Let me remind you that regression analysis is closely related to another type of analysis of mathematical statistics – correlation analysis.

These statistical forecasting methods are common due to the inclusion of elements of probability theory and are limited to the study of the influence of one or more factors.

And if the first is responsible for identifying the form of the relationship between the dependent and independent variables, the second is responsible for establishing the closeness of such a connection. In most cases, they do not talk about individual methods, but about a complex correlation-regression analysis.

Innovative practice in order to predict the results of NTP often operates with functional dependencies, among which the most applicable are the so-called S-shaped curves, which best correspond to the life cycles of technological systems.

But one such curve allows you to get acceptable results of the relationships of time series for one factor and within a single qualitative leap.

At the same time, forecasting often requires an assessment of the dynamics of a group of successive technological breakthroughs. In such cases, experts recommend using graphic modeling of scientific and technological development based on a series of S-shaped curves, building on their basis a summing envelope curve.

Envelope curve

Envelope curve for forecasting the prospects for increasing traffic speeds

The above is an example of the construction in the 70s of such a envelope curve to assess the prospects for the development of transport speeds by mankind. The graphical appearance of serial estimates is clear, and regression analysis can be a worthy addition.

Its tools are very useful for making predictions of the genesis of existing technical systems, taking into account previously identified factors. The constructed functions of the regression model make it possible to determine the directions of the nearest developments in the field of science, engineering and technology.

But if the reasons for the likely change in the predicted variables are unknown, the applicability of this method should be questioned, and all its advantages over the same extrapolation method come to naught.

Regression expressions formed on the basis of single factors can be combined by another formalized method called "econometrics". This method allows you to perform multi-factor analysis and create an innovative quantitative forecast based on a large number of criterion parameters. Suppose we need to perform an assessment of the autonomy potential of wearable devices in 3-4 years.

The multiplicity of parameters of the screen, the heart rate monitor sensor, and other embedded subsystems in the gadget determines the set of functions of technical progress, including the possibilities of energy intensity and battery weight.

All of them determine the composition of the equations included in a single model of cause-and-effect relationships, in which there is a combination of multifactorial regressions and variables selected according to the conditions of the problem. The econometric model is ultimately a synthetic function of the form:

Econometric formula

An example of an econometric formula for building a multifactorial forecast

Econometric techniques give the best result in conditions when it is possible to identify existing cause-and-effect relationships and, especially, when variables are very dynamic in time and it is possible to predict the direction of their changes. Despite the high laboriousness of this method, it has undoubted advantages. Chief among them is that the identified trend is stable.

Expert types of intuitive forecasting methods

In the previous sections of the article, a description of some formalized forecasting methods is given. The lion's share among them is occupied by tools for quantitative analysis of data series, namely statistical forecasting methods. We did not consider such interesting factual forecasting methods as the method of modeling or simulation, methods of analysis of patent and non-patent information, which are of great importance in innovation.

In this section, we will evaluate intuitive methods based on an expert approach in developing a vision for the intended future of STP. It is quite reasonably believed that none of the forecasting tools can do without experts and their positions. Indeed, if all the data for the forecast are selected and the level of their quality is high, then this is no longer an educated assumption, but an ordinary plan based on a simple extrapolation. In general, all intuitive forecasting methods are divided into groups of individual expert estimates in forecasting and collective.

For the purposes of the article, an expert will be understood as a highly qualified specialist in the field of scientific knowledge, engineering and technology, who is invited to analyze, evaluate, develop and express judgments on the question posed to him.

Innovative design assumes that the question and judgment are formed in the field of scientific and technological development, on the subject of pre-project research or innovation strategy. Under expert judgment, it is proposed to take an answer to the question with detailed argumentation.

Expert judgment participates in the preparation of a forecast, if the object of research is extremely simple, or, conversely, extremely complex. In the second case, it is impossible to supplement the information to a level sufficient to make a decision in any other way than by presenting a special professional opinion about the subject and object. The specific motivations for the involvement of experts are as follows.

Peer review can replace data gaps for extrapolation.
Computational processing of previously obtained information is unreasonably expensive.
The data are suitable for processing only by peer review and not otherwise.
Tolerances in the alleged error and inaccuracies of the forecast give a chance to take into account the position of the specialists involved.
Individual specialists have unique expert qualities.

The method of expert assessments assumes that when searching, selecting and selecting experts, the company's personnel service is guided by a number of rules that take into account the professional and personal characteristics of candidates and their psychological portraits. In addition, the HR service is guided by special methods for establishing the competence of the expert, among which are:

Brainstorming method.

Intuitive forecasting methods in the form of individual expert assessments differ in the following types.

Methods of interviews and questionnaires.
Analytical expert methods.
The method of morphological analysis and the construction of a morphological matrix on its basis.
Scenario method.
The method of psycho-intellectual generation of ideas.
Features of the matrix method of forecasting

In the preparation of forecasts for the development of NTP, normative forecasting methods occupy a large weight. In them, a quantitative assessment of prospects is carried out on the basis of the goals and objectives that the innovative enterprise sets itself for the forecast period.

Normative forecasting research uses the matrix forecasting method as the main method. Decision matrices can be built in two-dimensional and three-dimensional form, presented in horizontal or vertical form. The task of the matrix approach is a comprehensive comparison of forecast directions according to the degree of their importance for the company.

As a kind of normative approach, the matrix method is focused on given normative values. Thanks to the assessment of different ways to achieve the target values, the forecast information can be formed on the basis of alternative solutions embedded in the factors of the matrix format.

At the same time, problematic restrictions are introduced into the options that are at various stages of the innovation process. Thus, the ways and means of obtaining an acceptable result are most clearly manifested among hypotheses and assumptions regarding the state of NTP in the future.

The object under study in the dynamics of development depends on many factors that are in one or other interrelations. The technique proposes to break the whole set of moments into groups according to homogeneous features.

The formed groups of factors are ranked according to an agreed criterion and then, using the matrix effect, the influence of these blocks on each other and their participation in the formation of the overall predictive result are evaluated. Below is a set of formulas for a hypothetical example of factor quantitative analysis in matrix form.

Innovation as an extremely risky project places high demands on the level of probable error in forecasting. To minimize errors during the assessment of the prospects of the results of the STP, factor groups should provide the maximum range of alternatives to solving a particular alleged problem. Then the algorithm for using the matrix method is as follows.

Identification of key points that can affect the achievement of innovation goals.
Collection of identified factors into groups by homogeneous features.
Quantitative analysis and assignment of coefficient values to groups of factors according to the point principle.
Construction of matrices of the influence of factor groups on each other and on the achievement of target results.
Computational procedures for matrix analysis, digitization of the degree of influence, ranking, construction of influence graphs, formulation of predictive conclusions.
Specificity of qualitative collective expert methods

Qualitative forecasting methods based on intuitive techniques for developing and processing expert positions, as we already know, can be individual or collective in nature. Qualitative work on the formulation and testing of hypotheses of the future is determined by a methodology based on the group work of tools, which include methods:

Delphi.

The first method involves the commission work of a group of experts, during which a discussion usually unfolds about the prospects for changes in the object of study and research. In terms of organization, this method is quite simple and operational. But for a number of reasons, including psychological, it has a number of drawbacks caused by:

the presence among the subjects of communication of an authoritative figure who can influence, wittingly or unwittingly, the position of other experts;
the different persuasive abilities of experts, which can also tip the scales in favor of stronger opponents during the discussion;
the likely inclusion of conformists and/or avid debaters in the group.
The method of brainstorming or storming allows you to remove some of the shortcomings of the "commission" method. The main thing in this method is to bring collective expert dynamics into the mode of free generation of ideas with maximum disclosure of potential. To do this, a special technique is used, it assumes a number of successive stages in its algorithm.

Preparing for brainstorming. Attraction of candidates to the generating group with a number of 5-15. The optimal group size is seven participants. Of particular importance is the figure of the host or moderator of the event. He should not only conduct collegial work itself, but organize it in aspects of the tasks of the assault, the selection of participants ("generators" and experts), the creation of conditions conducive to obtaining a satisfactory result.

Actually the work of the group during the session. Before that, the host of the event informs the participants of the rules for brainstorming (refusal to criticize, maximizing the number and continuity of the ideas expressed, etc.).

Collection, evaluation and selection of ideas. This stage is carried out by a separate group of experts who are looking for rational grains in the most "wild" ideas.

Elaboration, development of the most interesting ideas to the level of acceptable conclusions.
All the non-formalized forecasting methods discussed in this section are considered intuitive, although they are carried out according to clearly developed rules for a long time of use. Another collective means of forecasting is the Delphi method. It is based on the principles of structured anonymous interaction of experts. At the same time, special questionnaires with controlled feedback are used.

The method takes a lot of time in terms of labor intensity and is very expensive, demanding on the clarity of the formulation of questions during the questionnaire.

But in the first stages of forecasting procedures, the Delphi method is very useful, since its results can provide a good basis for more detailed studies. I

t allows you to perform structured processing of responses from a large expert group and present a number of data close in volume to a representative sample. At the end of the section, I propose to consider a special case of collective expert evaluation in the form of a sequential algorithm.

forecast development algorithm
Algorithm for forecast development based on the collective expert method
Scenario and adaptive forecasting methods

We have already considered a fairly large number of methods for preparing a forecast that are used to assess the prospects of NTP objects. But there are such qualitative forecasting methods that we can not ignore in any way, and at least a few words should be said in terms of their characteristics and features. One such tool is the scenario method of forecasting.

In fact, this method has varieties. The fact is that when developing scenarios for the future, either a deductive or inductive approach is used.

The first subtype of the scenario method is based on the analysis of key factors that affect the behavior of the predicted object. Simply put, we go from a general analysis of alternatives to particular conclusions on the forecast.

The second type of scenario, on the contrary, is built by analyzing specific situations, on the basis of which a general picture of the object arises. The deductive method is applicable to the creation of both a search forecast and a normative one. Consider the stages of search forecasting by the scenario method of deduction.

Scenario variations are studied on the basis of general alternatives to changes in the object of study. An axis of polar alternatives is being built, between which various intermediate positions are located. On this axis is the so-called "zero alternative".

The system of significant factors that influence the object in its development is subject to description and ranking.

The range of possible future states of the object of study is established. Here it is important to find the exact formulations of such states, each of which defines a separate scenario branch.

With the connection of matrix techniques, the degree of mutual influence of factors is revealed.

A cognitive map of the interaction of factors that are graphically depicted on the vertices of graphs, and the connection itself is drawn up in the form of lines or arcs (see below for an example of a map).

Qualitative analysis of the map is performed and those combinations of factors that lead to the previously established scenario outcomes are revealed.

For example, a model of search scenarios for the development of the object in the field of building up the innovative potential of the company is being formed.

Example of a cognitive map of the interaction of factors

Adaptive forecasting methods allow adjusting sliding forecasts to the evolution of variable characteristics of the studied NTP objects, to flexibly adapt to the processes and events taking place. But, again, almost all predictive methods can be represented by adaptive.

Modern automation tools make it possible to achieve a short time lag of response to modify the forecast, if the relevant criteria are pre-embedded in the information model. To form an adaptive predictive image of the future, short-term forecasting methods are best suited.

This need is very sensitive when assumptions are made on the basis of statistical data in conditions of strong oscillation of dynamic series data.