查詢結果分析
來源資料
頁籤選單縮合
題名 | 情資優先排序技術研究=The Research of Information Priority Ranking |
---|---|
作者姓名(中文) | 郭立言; 王岳吉; | 書刊名 | 新新科技年刊 |
卷期 | 12 2016.01[民105.01] |
頁次 | 頁121-130 |
分類號 | 592.6 |
關鍵詞 | 推薦系統; 分散排序; 關係圖涵蓋; 影響力傳播模型; 節點增強式隨機漫步; Recommender system; Diversified ranking; Graph coverage; Influence diffusion model; Vertex-reinforced random walk; VRRW; |
語文 | 中文(Chinese) |
中文摘要 | 在指管通資情監偵(C4ISR)作戰環境下,掌握戰場情資為兵家首要之舉。隨著戰場環境改變,複雜及大量的各類情資紛紛湧入指揮所,透過分散排序之推薦系統,以解決情資冗餘問題,本研究以節點增強式隨機漫步加以節點影響力評估作為推薦系統演算法,並利用三種不同的關係網路驗證演算法效能。研究結果發現InfRank具有強健性,兼具PageRank與MRSP兩種不同演算法的優點,取得關聯性與分散性的平衡,獲得較好的推薦結果,可供未來情資分研系統參據。 |
英文摘要 | To retrieve information from battle field has the highest priority in C4ISR environment. With the variation of battle environment, complex, massive and heterogeneous information overwhelm command posts. To solve the problem of information redundancy, a recommender system based on diversified ranking is applied. In this research, we integrate vertex-reinforced random walks and vertex influence estimation to form a novel algorithm, and then demonstrate the performance of the algorithm on three different relation networks. From the experimental results, InfRank does not only outperform other methods with robustness but also combines the pros of the two algorithms, PageRank and MRSP. Hence, better performance resulted from finding a tradeoff between relation and diversity can be a reference for results for the prospective information analytics. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。