The problem of forecasting emerging technologies (FET) has attracted the attention of researchers in the patent-analysis community because it provides a starting point for organizations to introduce radical technology change or a new product or engage in a merger and acquisition. Many FET methods are limited, however, in their ability to identify the component subareas of emerging technologies and whether the nature of the emerging technology is that of a radical breakthrough or simply an incremental change to a previous technology. In addition, many FET methods also rely on expert opinion in the selection of inputs for forecasting models and the analysis of their outcomes. The reliance on expert opinion affects the reliability of the forecast. In this research, we propose a Hybrid Similarity Forecast (HSF) Model that identifies the technical terms and subareas of emerging technologies and provides insight into their nature. The model also minimizes dependence on expert opinion, hence increasing the reliability of results. We use patent keywords as a source of technical detail and improve keyword-based analysis by integrating keywords from their constituent cited patents (CCP). The HSF Model uses a K-means clustering algorithm to group patents based on a similarity matrix that integrates patent-based angles and CCP-based angles. Through this technique, we gain insight into the influence of CCP on the clustering outcomes. We tested the performance of HSF Model on three patent data sets related to (1) glucose biosensor, (2) personal digital assistant, and (3) thin film transistor-liquid crystal display technologies. We compared the model’s performance to a benchmark model and found the model to achieve reasonable and effective results according to evaluation measures. We also compared the performance of the model by varying a weighting scalar we assigned to the CCP-based angles. We found an optimal value to assign to the CCP-based angles for each technology tested to achieve reasonable and effective results according to clustering evaluation measures. The HSF Model integrates a unique combination of features in the keyword-based analysis, particularly the integration of keywords from the CCP. As such, this model can serve as inspiration for further studies on FET.
|In Administrative Set: