Data mining-based variable assessment methodology for evaluating the contribution of knowledge services of a public research institute to business performance of firms

Abstract

This study proposes a methodology for assessing the contribution of knowledge services (KSs) provided by a Korean public research institute to the business performance of firms. A new methodology based on a data mining-based variable assessment method in a regression model is proposed for the service-level assessment. The contribution of the KSs to firms’ business performance is analyzed using their attributes and specific business performance indicators through the conditional variable permutation method in the random forest regression. This reduces the ambiguity in variable importance caused by the correlations among input variables. The proposed methodology is applied to the survey dataset collected from firms. The survey dataset is examined 1) for the whole data and 2) for a subset of the data, namely, small- and medium-sized enterprises (SMEs). The empirical results show behavioral properties of firms with regard to the given KSs in general and SMEs in particular. Practical and user-friendly service product types increase the firms’ expectation on business performance. Also, flexibility in the service products helps firms acquire much-needed knowledge and boosts their expectation on business performance. In particular, SMEs expect better business performance from the KSs that help them create business plans and strategies.