Research On Category Theory-based Mobile User Interface Pattern Recommendation

In order to improve the efficiency and quality of user interface development of mobile devices,pattern-based user interface development methods are widely used in mobile user interface development.By introducing mobile user interface pattern(MUIP),it not only provide design solution for complex design problems,but also help developers quickly understand and use the design knowledge and experience of development.To improve the accuracy and efficiency of finding MUIP,developers use pattern recommendation method to find MUIP from the pattern library in the pattern-based user interface development.However,when using the existing pattern recommendation methods to help developers find MUIP,there are still some problems,such as low accuracy of search and low ranking quality of recommendation list.There are two reasons are organized as follows.Firstly,the existing pattern library construction methods can not describe MUIPs and their relationships on a high level of abstraction,which leads to the lack of hierarchical structure of established pattern library and can not provide structured support for pattern recommendation methods.Secondly,the existing pattern recommendation methods do not make full use of various mapping relationships in MUIP search,and there are deficiencies in requirement information extraction,sreach strategy and similarity measurement between design problems and MUIP.To solve the above problems,the construction of pattern library and pattern recommendation method are studied.In the aspect of constructing pattern library,firstly,the user interface pattern language is studied,and category theory is used to describe MUIPs and their relationships on a high level of abstraction.Then,the user interface pattern language is used to build a pattern library with a hierarchy to provide structuralized support for pattern recommendation methods.In the research of pattern recommendation methods,firstly,in order to discover the internal structures and rules of the mapping relationships,category theory is used to categorize the mapping relationships in the search process.Then,these internal structures and rules are applied to the pattern recommendation method.The main contribution of this dissertation is given as follow:(1)In order to address the problem that the existing pattern librarys are unable to provide effective structuralized support for MUIP recommendation method,a category theory-based mobile user interface pattern language(CTBMUIPL)is proposed,which is used to build a categorized pattern library with a hierarchical structure.In CTBMUIPL,category theory is utilized to describe and organize MUIPs and their relationships on different level of abstraction.The mapping types and structural characteristics of categorized pattern library are analyzed by examples.The experimental results show that the pattern library constructed by CTBMUIPL can provide structuralized support for MUIP recommendation method.(2)In the search strategy of isolated design problem,an improved particle swarm optimization and abstract hierarchical information-based mobile user interface pattern clustering(IPSOAMC)algorithm is proposed,which uses semi-supervised kernel fuzzy c-means(SSKFCM)clustering and abstract hierarchical information.To solve the sensitive problem of clustering parameters,an improved particle swarm optimization(IPSO)algorithm is proposed to optimize clustering parameters.The experimental results show that IPSOAMC algorithm outperforms the existing clustering algorithms in terms of clustering effect and convergence.(3)In order to address the problem that the exsiting pattern recommendation methods can not provide high-quality recommendation lists.The integrity of requirement information and the similarity measurement of key attributes are studied.Firstly,in order to ensure the integrity of requirement information extraction,a question and answer-based requirement information extraction method is proposed,which uses a set of question-answer pairs to obtain requirement information of developers and obtains the key attribute information of MUIP search.Then a pullback-based same attribute search algorithm is proposed,which uses pullback to find the same part between design problem and MUIP.In order to improve the accuracy of similarity measure,the pullback-based same attribute search algorithm is applied to calculate the similarity between design problem and MUIP.The results of recommendation effect experiment show that the research on the integrity of requirement information and the similarity measurement of key attributes can improve the ranking quality of recommendation list.(4)A category theory-based mobile user interface pattern recommendation method(CTBMPR)is proposed.Firstly,in order to combine design problem with pattern library closely,category theory is used to represent the mapping relationships in the MUIP search.Then,the requirement information is extracted from the design questions by using the question and answer-based requirement information extraction method.Finally,the IPSOAMC algorithm and the categorized mapping relationship are applied to the search strategies of isolated design problems and associated design problems respectively.According to the similarity measurement results between design problems and MUIPs,a recommendation list of MUIP is provided to developer.The experimental results show that the CTBMPR outperforms the existing pattern recommendation methods in terms of search accuracy and ranking quality of recommendation list.