선택한 단어 수는 입니다.
최소 단어 이상 선택하여야 합니다.
최소 단어 이상 선택하여야 합니다.
선택한 단어 수는 30입니다.
최대 10 단어까지만 선택 가능합니다.
최대 10 단어까지만 선택 가능합니다.
제안 방법
- Experiments were carried out to predict the preference of the Social SVD++ CF and to confirm its usefulness. Experiment 1 sequentially conducted experiments for evaluating SVD, SVD++, and Social SVD++ prediction performance on the Research Data 1.
- The second data is the data used in Kim and Ahn (2010), which is rating information related to movie preference and social network information between the users. In order to optimize and generalize the parameters of SVD, the research data were divided into two parts: training set and test set, and then SVD, SVD++, and Social SVD++ were sequentially applied to the test set.
- In particular, this study will evaluate the performance of the model by reflecting the real-world user's social network information in the recommendation process.
- In this paper, we use a method to integrate social network information into collaborative filtering in order to mitigate the sparsity and scalability problems which are major limitations of typical collaborative filtering and reflect the user's qualitative and emotional information in recommendation process.
- In this study, we tried to confirm the usefulness Social SVD++ CF model, which considers social network information of real-world users as implicit data by integrating typical collaborative filtering and extended SVD++ model. Social SVD++ CF has the advantage of reflecting the user's qualitative and emotional information in the recommendation process because it can reflect the user's social network information in the recommendation process.
- In this study, we try to verify the usefulness the model by using real-world users’ product rating and social network information.
- The method used in this study reflects the social network information of the user which can be considered to contain the user's qualitative and emotional information among the implicit data that can be considered in SVD++.
대상 데이터
- Experiment 1 sequentially conducted experiments for evaluating SVD, SVD++, and Social SVD++ prediction performance on the Research Data 1. Experiment 1 consisted of 482 training sets and 120 test sets with the ratio of 6: 4 data of 32 restaurants and cafes evaluated by 39 users in Research Data 1. Experiment 2 consisted of 2816 training sets and 1877 test sets with the ratio of 6: 4 data of 90 movies evaluated by 100 users in Research Data 2.
- Experiment 1 consisted of 482 training sets and 120 test sets with the ratio of 6: 4 data of 32 restaurants and cafes evaluated by 39 users in Research Data 1. Experiment 2 consisted of 2816 training sets and 1877 test sets with the ratio of 6: 4 data of 90 movies evaluated by 100 users in Research Data 2.
- 2, there are many kinds of data was collected such as frequent visiting days, frequent visiting time, visiting purpose, taste of goods, price, atmosphere, convenience, satisfaction for restaurants, cafes in the university or nearby. Finally, total of 288 survey items were collected by defining 32 items with 9 preference evaluations per each item. In this study, only the overall satisfaction score was utilized.
- Meanwhile, research data 2 is data used by Kim and Ahn (2010), and includes ratings data and social network information of real-world users about movie preference. Research data 2 includes 100 movies rating data for 90 users and includes friendship data from social network services such as Twitter, Facebook, MySpace, Cyworld, and others. See Kim and Ahn (2010) for a detailed description of Research Data 2.
- The research data used in this study were collected through mobile and internet online surveys by building a customer rating data system. As shown in Fig.
- 3. This screen provides images of demographic characteristics, 32 restaurants and cafes. The users evaluate the point-of-interest using score from 0 to 7.
데이터처리
- For this purpose, the MAE (Mean Absolute Error), which means the absolute value of the difference between the predicted value of the product and the actual value, is calculated for each model, and the results of the SVD based collaborative filtering, the SVD++ based collaborative filtering, and the social SVD++ based collaborative filtering are compared. And we check the generalizability of the differences among models using paired sample t-test.
- In this study, to compare SVD, SVD++, and Social SVD++, the difference between the predicted preference score and actual preference score is evaluated as MAE (Mean Absolute Error), and the difference is statistically significant (Paired Sample T-Test) was performed. The MAE is a statistical measure for evaluating the prediction accuracy of the recommended performance by comparing the score predicted from the analysis with the score evaluated by the actual user.
이론/모형
- In this paper, we use a method to integrate social network information into collaborative filtering in order to mitigate the sparsity and scalability problems which are major limitations of typical collaborative filtering and reflect the user's qualitative and emotional information in recommendation process. In this paper, we use a novel recommendation algorithm which is integrated with collaborative filtering by using Social SVD++ algorithm which considers social network information in SVD++, an extension algorithm that can reflect implicit information in singular value decomposition (SVD). In particular, this study will evaluate the performance of the model by reflecting the real-world user's social network information in the recommendation process.
- Since the yj in equation (3) are controlled close to zero, their sum is normalized by # to stabilize the deviation over the range of the observed values of |R(u)|. The coefficients of the model are determined by minimizing the associated constraint squared error function through a stochastic gradient descent algorithm.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.