A model proposal for technology forecasting

World is changing rapidly and globalization creates competitive environment which oblige companies to search new ways to stay competitive. Regarding economic growth, companies and countries need to improve technological capacities and innovation systems. Since major force in economic growth is technological change, following technological progress and staying up-to-date ensure the survival of the company. Technological progress, especially its time-based trajectory forecasting, has been the subject of numerous studies in the context of management of technological innovation and its sub-disciplines, such as technology innovation and strategy, technological forecasting and more recently roadmapping. Successful science and technology policies depend on the technology foresight and forecasting studies. The discipline of technological forecasting is directly connected to the management of technological innovation, attempting to incorporate simultaneously the principles of science-based technological progress and the diffusion of technological applications. Technology forecasting is used for analyzing the level of technology, estimates the future value of characteristics and performance of a technology. Accurate technology forecasting can improve the quality of decision making. There are many unknowns surrounding the future, so it is impossible to forecast technological progress without an uncertainty. A succesful forecast minimizes the uncertainty of the forecasts. Decisions must be taken using the best information available at the forecasting time and decision makers should use this information in the most effective way. For a better forecast the structure of the technology and technological progress need to be clear. At the very beginning, it is important to define technology. Technology is the application of a phenomenon to achieve a human purpose. It is a system that consists of one or more physical phenomena towards to a working principle. This system consists of other technologies, which contribute to the purpose of the main system. Therefore, technological progress should be studied together with the progress in subsystems. The proposed hierarchic structure composed from critical subsystems they embody and the performance characteristics they deliver to the overall product or system. The meaningful level of disaggregation is dependent on the complexity of the product and requires expert knowledge of the overall product and the technologies it embodies. The disaggregation of a system into its critical systems, driving system level performance, facilitates systematic and clearer analysis to extract technological, science-based limits. Technological forecasting on the level of subsystems provides the possibility to make the connection with the underlying natural and life sciences. Then the second important point to clarify is understanding why technologies evolve. The answer is simple, unlikely to most applications on technology forecasting literature, it is not the passage of time that leads the progress, it is the application of effort. So the realistic model of a technological progress should include the parameters of "efforts put" as inputs and parameters of "results achieved" as outputs. Therefore, measuring "efforts put" is one of the important parts of this study which explained in R&D sub-model. On the other hand, "results achieved" are the indicators of technological progress. These parameters are divided into two parts based on their emerging time. First one is patents which are known as trails of innovation. They are outputs of scientific and technological efforts which influence technological progress, respectively, are commonly accepted measures in innovation research. But not all patents creates a technology change. Patent data will obviously contain some innovations, but they will also contain inventions without any commercial value. After utilizing the useful ones which do lead innovations, performance of new technology can be measured. The performance of "technology in use" is the second type of measure of technological progress. Performance is usually characterized by multiple parameters in a complex system which are aggregates of subsystems. All these input and output parameters generates the essential components of the proposed technology forecasting model. New technologies undergo typical patterns of scientific, technological and economic developments, while the scientific-push is seen as prior to the technology-push and market-pull. More recently, strong positive relationships were identified between the inputs and outputs of scientific research, such as research expenditures and papers published, and technological innovation measured in patenting rates. The central goal of this thesis is to propose a new technology forecasting model for deriving technological progress trajectories as a function of efforts put, at a level of aggregation, regarding complex structure of technologies and dynamic environment. It is intended to accomplish by: a) illustrating the connections among R&D investments, natural science, technology and use in terms of processes and information flows; b) describing measures of technological progress that are aggregated on a system's or product level; c) offering a disaggregation model which enables the derivation of the system's critical subsystems; and d) creating a framework to connect proposed sub-models including neural networks. In order to achieve this objective, three types of submodels designed. First one is R&D sub-model in which the "efforts put" to improve technologies are analysed. Expert interviews and delphi technique will be used in determining R&D parameters. Second sub-model is system sub-model which requires to decompose main system into its critical subsystems. For each critical system, patent data be will be collected under this system sub-model. Third one is the performance sub-model at which "technology in use" performance indicators of overall and critical sub-systems will be determined. Another contribution of this thesis is, it also proposes a conceptual model for technology forecasting activities, using disaggregates of a product or system into its critical sub-systems. This model is intended to facilitate more specific and informative analysis about the causal connections between technological progress of system's critical technologies and system level performance attributes. Mapping the technological life cycles of these technologies, including the scientific-technological limits, should also have significant practical application for resource allocation decisions, technology strategy, planning, and roadmapping. After designing all sub-models, the usable data will be clear in the scope of submodels. Then next step is collecting historical time series data. These sub-models of the model will be merged to create a technology forecast using artificial neural networks (ANN) which can produce successful solutions for forecasting and classification problems with its high relation definition abilities. Considering the nonlinear nature of the technology forecasting problem, a dynamic recurrent time series NARX (Nonlinear AutoRegressive eXogenous) model with input and feedback delays is used to forecast technological progress. Finally proposed model is applied to domestic refrigerator technologies. Under the scope of the analysis, interviews and delphi questionaires are conducted to participated experts who are working in R&D department of a white goods company. ANN is applied through MATLAB 2014b. Historical data of R&D investments, patents and performance data is collected between 1980 - 2014 and forecast obtained for 2015 - 2040. As a result of the application on refrigerator technologies, the overall system appears to be mature about 2030s. But still improvement possibilities exist and manufacturers should consider further investments. When the lower levels are analyzed, the subsystems which are expected to develop and which need a radical change, are apparent. Insulation subsytem draws the attention as it stands out to be a technology need to be developed. Heat exchanger subsystem, namely condenser and evaporator, are about to reach their technological limits. Therefore they seem to be candidates as bottleneck technologies for the development of refrigerator. Since improvements are limited for these technologies, radical technology developments should be considered by the manufacturers. These investments require extensive financial resources. In compressor subsystem there are still improvement capabilities in decreasing energy efficiency. In refrigerant technologies, although environmental performances are high, still it is not very high in physical performance of the refrigerant and need to be improved. Therefore further incremental improvements on these technologies can be considered as an R&D investment strategy. To emphasize; in this proposed model, although performance is a function of technological effort, using time series data enables a time-based forecast. The multilevel analysis of main system and relevant subsystems links main system to its critical subsystems in order to delineate detailed technological progress of the system and improve decision making quality in technology investments. Overall technology progress can be observable and traceable to the lower levels. The effects of natural connections between the levels of systems can be understandable. Also utilazing artifical neural networks enables to produce a technolgy forecast without any assumption.