Thesis study
Thesis Title: Indian Ying, Zhao Yuhai, Zhang Bin, WANG Guo-ren recommendation algorithm based on multi-user developer community.
- Study
Stack Overflow and related data Github - Research motivation
- Integration of user information the developer community, through the analysis of the interaction between the user and user behavior, building a developer community across the network, updating user tags. Further, to expand the scope of the problem query keywords by using the Taxonomy, on this basis , co-user matrix more accurate recommendations, and increase the range of valid users when recommended. This centralized approach from the use of the document
- Literature review
- Study Design
- Problem definition and research ideas
- Definitions, definitions should be used to identify the user's text similarity calculation, and calculating the relationship between the user and the user is applied to define the normalized linear
- Description of the problem, using the relationship with the user and the user of the tag, restart random walk by the matrix operation between obtained may represent a more accurate relationship between the user and the tag user order n × m - Matrix Label
- Research ideas, realize update algorithm based on user label and recommend restarting the random walk algorithm based on Taxonomy users to improve the accuracy and accuracy problems when the user label recommendations
-
Recommendation algorithm based on multi-user developer community
- Communities across the same user identification, answered by statistical behavior of different user communities involved in the issue, to get user preferences on the type of answering questions, and expressed in the relationship between the user and the label
- Restart random walk algorithm to update the user based on a different set of rates reboot, restart using the obtained updated user walk in many cases - Label relationship.
-
- User recommendation algorithm Taxonomy based Taxonomy by extension, the effective expansion of the scope of the label hit the weights
- Experiment and Analysis
- The experimental data sets, data collected included herein as of October 2017 the Stack Overflow and a total of about 1.4 million at 117 tags Github two communities valid theme stickers. Stickers user statistics for all subjects involved in them, and can not be ruled out visitors are recommended reference accounts and similar large-scale public account google, the cumulative effective of about 400,000 registered users two developer community id and event information in effective problem
- User identification experiment, based on the user identification tag preferences
- User label update experiment, over control to adjust the rate reboot (restart) in order to obtain a more accurate label for the number of user preferences and more accurate user preference for the right to re-label
- User recommendation experiment to calculate the value of all users are equal, to analyze the paper aside a certain number of the user's accuracy; and accuracy based on user rankings, decreasing customer value analysis recommended user ranking
- Using data sets
This article includes data collected as of October 2017 the Stack Overflow and a total of about 1.4 million at 117 tags Github two communities valid theme stickers. Stickers user statistics for all subjects involved in them, and can not be ruled out as a recommended reference tourists google account and similar large-scale public account, obtain a total of approximately 400,000 valid id registered users two developer community and its activities in the information valid question.
- Analysis conclusion
- Herein, for the extracted keyword set by semantic tree Taxonomy expansion, the user issues the label after the label and the matrix right after expansion by weight updates recommended collaborative experiments show that the method described herein for the recommended number of users increase significantly, and accuracy also have a higher increase.
- Learning experience
paper grabbed a lot of experimental data, showing that the cost of a lot of effort. Understanding the user to restart the random walk algorithm update label with recommendations based on the user algorithm Taxonomy
Thesis Title: Indian Ying, Zhao Yuhai, Zhang Bin, WANG Guo-ren recommendation algorithm based on multi-user developer community.
- Study
Stack Overflow and related data Github - Research motivation
- Integration of user information the developer community, through the analysis of the interaction between the user and user behavior, building a developer community across the network, updating user tags. Further, to expand the scope of the problem query keywords by using the Taxonomy, on this basis , co-user matrix more accurate recommendations, and increase the range of valid users when recommended. This centralized approach from the use of the document
- Literature review
- Study Design
- Problem definition and research ideas
- Definitions, definitions should be used to identify the user's text similarity calculation, and calculating the relationship between the user and the user is applied to define the normalized linear
- Description of the problem, using the relationship with the user and the user of the tag, restart random walk by the matrix operation between obtained may represent a more accurate relationship between the user and the tag user order n × m - Matrix Label
- Research ideas, realize update algorithm based on user label and recommend restarting the random walk algorithm based on Taxonomy users to improve the accuracy and accuracy problems when the user label recommendations
-
Recommendation algorithm based on multi-user developer community
- Communities across the same user identification, answered by statistical behavior of different user communities involved in the issue, to get user preferences on the type of answering questions, and expressed in the relationship between the user and the label
- Restart random walk algorithm to update the user based on a different set of rates reboot, restart using the obtained updated user walk in many cases - Label relationship.
-
- User recommendation algorithm Taxonomy based Taxonomy by extension, the effective expansion of the scope of the label hit the weights
- Experiment and Analysis
- The experimental data sets, data collected included herein as of October 2017 the Stack Overflow and a total of about 1.4 million at 117 tags Github two communities valid theme stickers. Stickers user statistics for all subjects involved in them, and can not be ruled out visitors are recommended reference accounts and similar large-scale public account google, the cumulative effective of about 400,000 registered users two developer community id and event information in effective problem
- User identification experiment, based on the user identification tag preferences
- User label update experiment, over control to adjust the rate reboot (restart) in order to obtain a more accurate label for the number of user preferences and more accurate user preference for the right to re-label
- User recommendation experiment to calculate the value of all users are equal, to analyze the paper aside a certain number of the user's accuracy; and accuracy based on user rankings, decreasing customer value analysis recommended user ranking
- Using data sets
This article includes data collected as of October 2017 the Stack Overflow and a total of about 1.4 million at 117 tags Github two communities valid theme stickers. Stickers user statistics for all subjects involved in them, and can not be ruled out as a recommended reference tourists google account and similar large-scale public account, obtain a total of approximately 400,000 valid id registered users two developer community and its activities in the information valid question.
- Analysis conclusion
- Herein, for the extracted keyword set by semantic tree Taxonomy expansion, the user issues the label after the label and the matrix right after expansion by weight updates recommended collaborative experiments show that the method described herein for the recommended number of users increase significantly, and accuracy also have a higher increase.
- Learning experience
paper grabbed a lot of experimental data, showing that the cost of a lot of effort. Understanding the user to restart the random walk algorithm update label with recommendations based on the user algorithm Taxonomy