INDUCTIVE MACHINE LEARNING FOR INSTRUMENT DEVELOPMENT

An inductive machine learning (ML) algorithm is used to discriminate b etween respondents and examine the dimensionality of an end-user compu ting satisfaction instrument. In all 616 responses were partitioned fo r training and testing purposes. Each respondent was required to asses s his or her overall satisfaction level; this was used for classificat ion of responses in two groups: satisfied and dissatisfied. Using 12 o ther survey items, the Cover Learning using Integer Linear Programming (CLILP2) algorithm correctly categorized 76% of the respondents. Reco gnition and discrepancy rates were used for instrument validation and in developing a shorter instrument.