By analysing a significant number of patents, it might be possible to determine what explains the success of a technology and use the system to give an opinion on a new trend. Problem to which a team of researchers has come up with an answer.
Is it possible to predict by computer the technological trends that are most likely to prevail? The question has interested a group of researchers at the Center for Complex systems studies, Kalamazoo College in the United States, who are working on the evolution of the process of technological innovation.
According to the scientists, it remains difficult for companies to make technology or process adoption choices, as many are afraid that a technology adopted in advance will not be adopted in the end. And that this generates risks in terms of investment in research and development. As a result, they have developed a system based on the analysis of citations or references made in the U.S. patent citation network database containing 37 years of patents.
A similarity search
The research team took each patent and calculated the number of mentions referring to it in the bibliography of other patents, based on keywords and the USPTO (United States Patent and Trademark Office) classification of 36 technological sub-categories.
Why not enjoy unlimited reading of UP'? Subscribe from €1.90 per week.
This methodology enabled them to identify the state of the knowledge structure in different sectors and to determine technology clusters. They then used a "citation vector", a quantitative element to determine the similarities between some patents and others using keywords. This allowed them to provide a mapping that gave a representation of the innovation landscape. To determine their evolution, they performed this process for patents filed between 1975 and 2007 representing time t, and those filed thereafter (t+1).
A decision support tool?
In their view, this system would provide a tool for observing technological trends and assist companies in the decision-making process when choosing a technology. However, they recognize that this quantitative methodology has some limitations, notably on the time lag between the birth of a new technology and its appearance in patent references.
In the long term, researchers hope to be able to systematically classify these patents by group: birth, death, growth, decay, disintegration, fusion. Above all, they hope to be able to predict the emergence of a technological innovation in a more microscopic way, i.e. by identifying it as early as possible with one or more events. In this respect, statistical analysis of the information contained in social networks could make it possible to detect the birth and development of an innovation and to classify them by group.
Article published in L'Atelier.net - July 13, 2012
{Jacuzzi on}
