本文選題：電子商務(wù) + 商品推薦　； 參考：《山東大學(xué)學(xué)報(bào)(理學(xué)版)》2016年01期

【摘要】：數(shù)據(jù)稀疏和運(yùn)行速度慢是個(gè)性化推薦系統(tǒng)面臨的難題。為了有效利用用戶(hù)歷史行為,基于用戶(hù)的評(píng)分記錄識(shí)別出用戶(hù)感興趣的內(nèi)容,并結(jié)合用戶(hù)間的信任關(guān)系,提出使用因子分解機(jī)(factorization machine,FM)模型進(jìn)行評(píng)分預(yù)測(cè)。FM具有線性時(shí)間復(fù)雜度,并且對(duì)于稀疏的數(shù)據(jù)具有很好的學(xué)習(xí)能力,因而能進(jìn)行快速推薦。試驗(yàn)結(jié)果表明,與傳統(tǒng)方法相比,基于因子分解機(jī)的商品推薦方法的準(zhǔn)確度有明顯提高。
[Abstract]:Sparse data and slow running speed are the problems faced by personalized recommendation system. In order to utilize user's historical behavior effectively, the content of user's interest is identified based on the user's score record, and combining with the trust relationship between users, it is proposed that the factorization machine factorization (factorization machine FM) model is used to predict the score of .FM with linear time complexity. And the sparse data has the very good learning ability, therefore can carry on the fast recommendation. The experimental results show that the accuracy of the commodity recommendation method based on factor decomposition machine is obviously improved compared with the traditional method.
【作者單位】： 河池學(xué)院計(jì)算機(jī)與信息工程學(xué)院;武漢大學(xué)計(jì)算機(jī)學(xué)院;
【基金】：廣西高校科學(xué)技術(shù)研究項(xiàng)目(KY2015LX338)
【分類(lèi)號(hào)】：TP391.3
，

本文編號(hào)：2016165