Testing Schedule Prediction Based On Parameter Feedback Method-An Automation Interfacing Case Study

[[abstract]]就開發機台介面而言，由於不同人員執行能力與各機台介面複雜程度的差異性，如何提高預估測試時程的準確度，往往為專案經理人的挑戰。此外，為持續評估這些開發人員的績效，如何在多個開發機台介面中提供客觀的累積開發績效評估，從而有效激勵其人員，亦為介面供應商的難題。因此，評估機台介面開發能力的有效方法乃成為本研究之重點。 根據機台介面的測試歷程，本研究提出一參數回饋方法，以建立介面開發能力模型(Interface Developing Capability Model, CMID)以評估開發者的能力。在應用上，將從評估開發機台介面之指標建立開始，透過推估介面測試階段的執行品質與進度，從而將推估參數回饋以調整CMID，達到CMID對分析對象變化的調整功能。在評估指標的定義上，以機台介面指標為核心，包含測試內容之結構複雜度、缺失密度與缺失改進能力等指標，呈現整合機台介面執行能力的差異性。因此，在CMID的建構上，透過分析機台介面測試歷程資料推估缺失，其分析結果經由雜度評量調整後與COCOMO II連結，可建立完整之CMID。最後，在測試時程的預估上，透過機台介面歷程缺失數呈現指數分佈的特性，進而推估後續測試階段的剩餘測試時間與範圍。 由多件機台介面專案的分析結果得知，測試剩餘實際與預測時程之平均相對誤差小於50%；此外，相對於領域專家對這些發展者能力的排序而言，CMID與專家之排序誤差比為80%。因此，可知基於所建構模型下，CMID其對測試剩餘時程的預測與不同機台介面的績效評估為有效。[[abstract]]For developing equipment interface project due to affected by differences from project complexities and developer capabilities how to improve prediction accuracy of residual testing time for project manager is a challenge Moreover how to assess developer performances with objective comparisons of execution results among different projects is a problem for interface vendor Hence the evaluating method for developing capability of equipment interface is focused in this research According to equipment interfacing histories this research proposes a parameter feedback method to construct a Capability Model for Interface Developing (CMID) for evaluting developer performance In application CMID indices structural complexity defect density and refinement history are defined to characterize differences between projects To construct the CMID testing histories are analyzed to find the developing defects and the results will map to a cost model with three COCOMO II parameters personal capability personal experience and product reliability and complexity Finally for defect numbers by time revealing exponential trend in testing phase the CMID will use the exponential trend to predict residual testing time of current or next target system for the same evaluation developer(s) According to the results from several equipment interfacing projectss the ratio of mean relative error between real residual days and CMID predicts is less than 50% Relative to the ranking from expert assessments on developer capability the ranking error ratio of CMIDs’ to experts’ is 80% In general CMID is proven its effectiveness on prediction of testing residual time and assessment of different target systems using the identical evaluation mode