Mr. Ran Etgar
The common definition of 'project' is a 'one-time effort'. Yet, many projects in the R&D and software development area are actually a continuing effort (CTD - Continuing Technology Development) involving development of new units with adjustment and improvement of existing ones. This effort is divided to time intervals, at the end of each one there is a new 'release' to the market. This process is exists in almost any development of a complex system, starting with cellular phones, software development ending with weapon systems such as "Iron dome". Despite the large number of CTD projects, analysis tools for statement of work (SOW) of each of the coming releases, including setting the features to be included in each release. On one hand marketing wishes to include as many features in earlier releases as possible. On the other hand, development demands executing a set of activities (with precedence relations and resource demands) to accomplish the feature. As there are relationships among the features (different features can share common activities) the problem becomes rather complex. As there is a lack of analysis and planning tools for CTD project, determining the content is based on basic criteria, which do not encompass the entire complexity of the problem.
The proposed research aims to use Data Mining and Clustering methods in order to establish the release content. To do so, a basic algorithm to optimally schedule the activities (subject to allocation of features to releases) will be developed. Based on this algorithm, the next stage will be to develop an attribute clustering methodology that will base on similarity (and un-similarity) factors. This tool will enable to provide a near-to-optimal recommendation of the release content.
In order to examine the developed technique, an examination will be run on a wide data base generated from a well-known data base (PSPLIB) by adding features (and their attributes) and also the inter-release time periods.
The developed technique will be confronted by parallel solutions.
Beyond the obvious contribution to the field of project scheduling, this research is a the first attempt to apply data-mining techniques to the field of content release setting, thus this research is about to contribute also to the theoretical aspect as well.