Research On Elicitation And Prioritization Of Feature-oriented App Requirements

With the rapid development of related technologies of Internet and Web service, there exists number of mobile Apps which have similar functions. Structural complexity of many Apps is also increasing. The number of App reviews is very large and it has been increasing. And developing App on the Internet is increasingly popular. Same as the traditional complex software system development, App development also begins with the task of requirements collection. The quality of requirements elicitation and the prior development of key features play an important role for the success of App development.Feature refers to a software characteristic specified or implied by software requirements documentation. It can also be defined as a collection which is comprised of a group of closely related individual requirements. We claim that software feature elicitation and prioritization will contribute to software requirements elicitation and prioritization. Selecting Appropriate feature elicitation techniques can not only alleviate the problems of missing requirements in the specification documents, but also meet the demand of the evolution of the App. In addition, feature prioritization from the view of structural complexity of the feature model can improve success rate of software development under the condition of given cost and time. Therefore, studies on elicitation and prioritization of App features have great significance for the work of requirements analysis of App on the Internet.This study takes App development on the Internet as the object. We study how to elicit the missing features from the description of massive similar Apps to improve the completeness of the requirements, how to compute the prioritization of the features from the view of the feature model to identify key features, and how to extract feature requests from numerous App reviews to facilitate the evolution of the App. Specifically, the research content of this paper mainly includes the following three works.(1) Feature elicitation based on feature model and collaborative filteringDeveloping App on the Internet must meet the demands of various users who are geographically distributed and it will face the problems of incomplete requirements and inaccurate expression. To this end, a feature elicitation Approach based on feature model and collaborative filtering is proposed. This Approach makes full use of feature descriptions of massive similar Apps. The relationships between the features of these historical Apps are firstly analyzed in detail to supplement partial features. Then KNN collaborative filtering recommendation algorithm is leveraged to predict whether the App has the missing features.(2) Prioritizing features based on node centrality in probability networkAiming at how to select key features to be preferentially developed from numerous features under the condition of given cost and time, an Approach for prioritizing features based on node centrality in probability network is proposed. This Approach constructs a probabilistic network by utilizing relations between features. It computes the values of all nodes in the probabilistic network. And it gets the priority of all the features based on the values of all the nodes.(3) App reviews classification and feature requests extractionIn view of how to mine valuable information from a huge number of App reviews for the evolution of the App, an Approach of App reviews classification and feature requests extraction is proposed. This Approach firstly analyses various attributes which can be used as classification features from reviews and their quantification. The performance of different classifiers under various classification attributes is discussed. Then, LDA document model is conducted to cluster the reviews on feature requests into semantic similar groups. Finally, Stanford Parser is used to extract the expected phrases as feature requests.An analysis tool for elicitation and prioritization of feature-oriented App requirements is developed based on the studies of the above three Approaches. The overall structure and functions of the tool are provided, and a study case of App features elicitation and prioritization in a simulative Wiki domain is given for simply display and analysis of the Approaches.