Dissertation > Excellent graduate degree dissertation topics show

Feature Based Informative Model for Music Recommendation

Author: ChengBing
Tutor: YuYong
School: Shanghai Jiaotong University
Course: Applied Computer Technology
Keywords: Information model Collaborative Filtering Matrix factorization Counter-example extraction
CLC: TP391.3
Type: Master's thesis
Year: 2012
Downloads: 61
Quote: 0
Read: Download Dissertation


With the rapid growth of the Internet above content providers to provide information to the explosive speed, the recommended system is also becoming increasingly important. In the past few years, the recommended system more effective in helping users obtain valuable information already showing great potential. In the actual application environment, the recommended system content providers and users will often bring great convenience, Amazon, for example, recommended to the user may be interested in the product can not only increase website sales, but also You can improve the user experience. The the KDD Cup Challenge2 game this year, Yahoo Labs open the scoring data sets by a very large user of music data. The entire game is divided into two sub-game, the first game is a score prediction, similar to the Net? Ix Prize Challenge3 goal is to narrow the predictive scoring value with the gap between the value of the actual scoring; second goal of the game is for a given user, we need to put the user played high scores three songs and the user never too much to play the three songs separate, high marks here is the scoring value greater than or equal to 80. In this paper, we solve the problems mentioned in the second game. We as a special form of Top-n recommended to treat other words, the model predicts higher scores three songs as songs played high score, and the remaining three seen as the user has never playing overly songs. In this paper, we sort of SVD (Singular Value Decomposition) model to solve this problem, and at the same time, we also adopted a number of counter-example extraction technology of counter-examples to supplement the training data. Most importantly, we have proposed to solve this problem based on characteristics of the information model, data concentration different kinds of information can be used as characteristics integrated into the model to form an integrated single model. Recommendation system, in order to improve the accuracy of the final model, usually mixed model approach. Different models to portray the information in the data, the hybrid model is often able to achieve more than a single model effect. All the results have been published in the issues discussed above, are also using this method. Hybrid model, however, often require more computational cost, no doubt, can achieve close to a single model of the hybrid model effect is more practical. In this article, we discuss this model. In this model, the different types of information, such as entry hierarchy (see Section 2.1), the the entry nearest neighbor relationships, relevant characteristics of the user as well as the implicit feedback, have been integrated into a single model. Through this model, the error rate on the test data set is only 3.1%, which is the best single-model results in all the results that have been published on this issue, even more than a lot of mixed model accuracy.

Related Dissertations

  1. Establishment and Update of Similar Users’ Cluster in Personalized Information Retrieval,TP391.3
  2. Research on Personalized Recommendation Algorithm Based on Natural Forgetting,TP311.52
  3. Research on Co-clustering and Application,TP311.13
  4. Monitoring and control in electric power remote monitoring and control system information model and security mechanisms,TP309
  5. A Study of M-commerce Personalized Recommendation System Based on the Ant Colony Algorithm,TP391.3
  6. Based on a user interested in discovering collaborative filtering,TP391.3
  7. Non-negative Local Coordinate Factorization for Image Representation,TP391.41
  8. Non-negative matrix factorization based on sparse image retrieval,TP391.41
  9. Social tagging systems based personalized recommendation engine technology,TP391.3
  10. GPU-based book recommendation system Research and Implementation,TP391.3
  11. Constraints and articles based on social context context constraints collaborative recommendation,TP391.3
  12. C2C e-commerce mode based product recommendation system applied research,F724.6
  13. Recommended based on social tagging systems research,TP391.3
  14. Research on Human Abnormal Behavior Detection and Recognition in Intelligent Video Surveillance,TP391.41
  15. Internet Oriented Sentiment Orientation Analysis of Chinese Public Opinion,TP391.1
  16. Hyperspectral Unmixing Based on Nonnegative Matrix Factorization,TP751.1
  17. Research on Personalization Recommender Technologies Based on User’s Interesting and Behavior of Browsing,TP391.3
  18. Study on Collaborative Filtering Recommendation Algorithm Based on Feature Vectors,TP391.3
  19. Research and Application of Resource Recommendation Based on Pragmatics Context,TP391.3
  20. Research of Distribution Network Model Based on IEC61970,TM769

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Retrieval machine
© 2012 www.DissertationTopic.Net  Mobile