Collaborative Filtering for Project Document Assessment of Missing Score Estimation

Abstract

     Abstract

     Measuring project document quality is usually handled by an assessment committee. Sometimes the committee may not cover all assessment issues. This can result in discrepancy of project document quality and evaluation, and thus may affect recommendations given in the project document. The researchers therefore would like to propose the collaborative filtering technique to predict scores left unevaluated by the committee. The User-based Collaborative Filtering (UBCF) was used in this research work. It consists of two main steps: similarity calculation and prediction. Experiments were conducted. There were 20 computer project documents assessed by the assessment committees and their assessment scores were collected. Some of the assessment scores were removed randomly and the UBCF was used to predict the removed scores. The experimental results showed that accuracy of the UBCF used in the experiments varied from 78 to 84 percent.        

 Keywords: Collaborative filtering, project document assessment, recommendation system, assessment score prediction