Springback Research Onaerofoil Blade In Hot Die Forging

Now the collaborative filtering recommend algorithm has been applied in mostrecommend area for its simple and high precision, many websites have adopted thisalgorithm in their recommend system. However, the problem of cold-start, the sparse ofdataset makes the collaborative filtering recommend system find ways to improve itself.Nowadays the wide application of social network allows the system utilizes thetrust network information in recommendation as well as the users’ history purchasing orrating information. Based on the idea of take full advantage of resource on the internetand the reality of Birds of a feather flock together, this thesis decides to import the trustnetwork into the traditional collaborative Filtering personalized recommend algorithm.We did much research work based on the Epinions and Friendfeed dataset. The maincontributions of this thesis are as following:1. Proposed recommend algorithm based on trust network: Trust network hassignificant influence on recommend system. In daily life we used to be influencedby our friends’ advice. In this algorithm the friends’ recommendation weight willincrease the target user’s possibility in the way of numerical value. By integrate thetrust network into the traditional collaborative filtering recommend algorithm,experimental results show that this algorithm not only improved the many kinds ofprecision indexes, but also improved the diversity. What’s more, it relieved thecold-start problem to some degree.2. Analyzed recommend result influenced by different trust relationships: Differenttrust relationships have different influence on users. This thesis came up with twoother recommend algorithms based on different trust relationships. Experimentalresults show that these algorithms improved the performance. Besides, the trustsituation performs better than the trusted one, which indicates trust presents users’interest better. What’s more, when we considered the situation of trust and thetrusted, it performs different on different dataset, which indicates network functiondecides its network structure and different network suitable for different algorithms.3. Analyzed network’s structure and function influence on recommendation: We divided the network into many situations according to its structure. Then we gaveout different suggestions under different situations.4. Designed an item-friend combined recommend system:This thesis designed anitem-friend combined recommend system based on the research of trust network inrecommendation, which can recommend both friends and items to users. Itsdiversity and novelty enhanced user experience.