情报学报
情報學報
정보학보
2014年
11期
20-46
,共27页
协同过滤%推荐系统基于邻域%相似度%预测精度
協同過濾%推薦繫統基于鄰域%相似度%預測精度
협동과려%추천계통기우린역%상사도%예측정도
collaborative filtering%recommender system%neighborhood based%similarity%prediction accuracy
协同过滤是推荐系统中最流行且应用最广泛的技术。基于邻域的推荐方法作为其两种类型之一,以简单、高效、稳定和解释性强的特性被广泛应用于商业领域。相似度计算作为该方法的核心步骤,其准确性直接影响预测结果的精度。现有的相似度方法是由共同评价而不是所有评价计算得到的,反映的是局部的相似性,与实际的相似性存在偏差。评价矩阵越稀疏,偏差越大。对此,本文提出一种新的相似度计算方法JS,将整体相似度计算和原有的局部相似度计算结合,更加完整地刻画相似度,同时不增加算法的复杂度,保持其原有的简单性和高效性。并对JS进一步优化,更加细致地描述整体相似度和局部相似度的关系。实验结果表明,该方法比现有的方法更有效。
協同過濾是推薦繫統中最流行且應用最廣汎的技術。基于鄰域的推薦方法作為其兩種類型之一,以簡單、高效、穩定和解釋性彊的特性被廣汎應用于商業領域。相似度計算作為該方法的覈心步驟,其準確性直接影響預測結果的精度。現有的相似度方法是由共同評價而不是所有評價計算得到的,反映的是跼部的相似性,與實際的相似性存在偏差。評價矩陣越稀疏,偏差越大。對此,本文提齣一種新的相似度計算方法JS,將整體相似度計算和原有的跼部相似度計算結閤,更加完整地刻畫相似度,同時不增加算法的複雜度,保持其原有的簡單性和高效性。併對JS進一步優化,更加細緻地描述整體相似度和跼部相似度的關繫。實驗結果錶明,該方法比現有的方法更有效。
협동과려시추천계통중최류행차응용최엄범적기술。기우린역적추천방법작위기량충류형지일,이간단、고효、은정화해석성강적특성피엄범응용우상업영역。상사도계산작위해방법적핵심보취,기준학성직접영향예측결과적정도。현유적상사도방법시유공동평개이불시소유평개계산득도적,반영적시국부적상사성,여실제적상사성존재편차。평개구진월희소,편차월대。대차,본문제출일충신적상사도계산방법JS,장정체상사도계산화원유적국부상사도계산결합,경가완정지각화상사도,동시불증가산법적복잡도,보지기원유적간단성화고효성。병대JS진일보우화,경가세치지묘술정체상사도화국부상사도적관계。실험결과표명,해방법비현유적방법경유효。
Collaborative filtering is the most popular and widely implemented technique in recommender system. Neighborhood based method is one of two general collaborative filtering algorithms, and enjoys a huge amount of popularity, due to their simplicity, their efficiency, their stability and their explanations. Similarity computation is the key step of the method, and it can impact on the prediction accuracy deeply. Traditional similarity computation method suffers from rating data sparsity problem, and can only obtain the local similarity by measuring the common ratings. Thus, this paper propose a novel similarity computation method—JS,which can be employed to compute similarities by combined local similarity and global similarity. And an improved method is proposed to optimize the relation between local similarity and global similarity. The two methods can keep the simplicity and efficiency without additional complexity. Experimental results show that the new methods outperform traditional methods or common significance weighting methods on the prediction of ratings.