Classification Decision-making Evaluation Method Of Sci-tech Journal Quality

The self-citation of science and technology journal is a normal behavior inreference, but a number of journals do some manipulation in order to improve theimpact factor, it is false to the evaluation of the quality of journals, and it hasseriously affected the objectivity and fairness of the evaluation. How to effectivelyidentify the true state of the quality of journals is an important problem withscientific value. In order to solve this problem this paper do two aspects of the study,the method of classification decision-making of journal self-citation behavior and anew evaluation method of comprehensive assessment of the journals, the specificcontent includs the following:First, study of the method of classification decision-making of science andtechnology journals. I use decision tree classification method in pattern recognition,and collect some time distribution datas about cited and self-cited and some relevantindicator datas from17journals which belong to the biology discipline from2001to2010for ten consecutive years, and establish the model of journal self-citationpattern classification recognition, the software weka do data processing and buildthe model, and I do evaluation of classification for these journals. The results showthat decision tree algorithm on the classification of the different grades of journals isideal, the cross rate is91.1765%. Finally, I randomly selecte several groups ofjournals in biology discipline in the JCR database and do classification, and provethe effective and reliability of this method.Secondly, study of the quality evaluation of science and technology journals. Iused principal component analysis method and correlation analysis to select the keyindicators, and I excavate some information from the amount of cited and re-amendand definition of the index of discipline diffusion, then I use discipline diffusion,self-citations behavior, reaction rate, aging rate to build the model of qualityassessment of journals, at the same time, I use the AHP to determine the weight ofeach indicator, finally, I select three journals which impact factor are close and threejournals with high impact factor from biology disciplines to analyze and verify theeffective of the model.Overall, the main innovation of this paper mainly contain two things: first, thedecision tree classification algorithm apply to the information metrology researchfor the first time and solve the multi-class classification problem of the behavior ofjournal self-citations. Second, building the quality evaluation model of journalsfrom multi-dimensional perspective and doing actual verification, avoiding thelimitations of journal impact factor in evaluating. This study has very important significance in safeguarding the normal development of journal and the excellentacademic atmosphere in academia.