現今網際網路的發展日益成熟,智慧型行動裝置的網際網路使用率快速成長,越來越多使用者透過智慧型行動裝置,來取得日常生活中各式各樣的資訊與服務,應用市集平台上的應用程式App越來越多,使用者想搜尋出符合需求的應用程式是相當困難的,因此前期有學者在2014年提出基於品質考量智慧語意網路分析於最佳化應用市集軟體推薦機制,後來有學者於2015年提出運用 Skyline 方法於最佳化應用市集 App 推薦機制,但在這些學者研究中發現推薦機制的方法只採用單一多準則決策分析方法來進行推薦,可能是因為多準則決策分析的其他方法推薦的不夠準確還是不適合用在即時系統推薦上。 由於單一多準則決策分析方法無法定論為最佳解,因此本研究將探討多準則決策分析方法中,TOPSIS、VIKOR、ELECTRE、AHP、PROMETHEE、SAW等,多種多準則決策分析方法來進行分析比較,並利用準確率(Precision)、回應率(Recall)與F1指標(F1-Measure)來進行方法的評估,驗證前期學者研究中所採用的多準則決策分析方法,是否比其他多準則決策分析方法來得好,並從中找出適用於應用市集平台上使用的推薦機制,或適用於即時推薦系統,多準則決策分析方法。
As the internet becomes more sophisticated, the demand for Internet resources on smart mobile devices increases rapidly. More and more people turn to their smart mobile devices to obtain information and services in daily life. Since there are an increasing number of applications in the application markets, it has become quite difficult for users to search and identify an application that meets their requirements. Beginning in 2014, researchers sought to address this growing problem and proposed the use of smart semantic network analysis to optimize the application recommendation mechanism in application markets. In 2015, researchers suggested the use of Skyline as a potential solution. However, these researchers adopted a single multi-criteria decision-making analysis method for the recommendation mechanism. This method can result in inaccurate recommendations relative to other multi-criteria decision-making analysis methods and therefor may produce inappropriate or inaccurate recommendations. Since the single multi-criteria decision-making analysis method may produce sub-optimal results, this study used several kinds of multi-criteria decision-making analysis methods for analysis and comparison: TOPSIS, VIKOR, ELECTRE, AHP, PROMETHEE, and SAW. The study used precision, recall, and FI-measure to evaluate the various methods and verify whether the single multi-criteria decision-making analysis method adopted by previous researchers is better than other multi-criteria decision-making analysis methods and to identify which of the multi-criteria decision-making analysis methods is best for the recommendation mechanism of application markets or real-time recommendation systems.