Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 55298 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Heydarnoori, Abbass
- Abstract:
- Due to the increasing variety and quantity of programs available in each field, the choice is for the users of smart devices and one of the main solutions to solve this problem is personalization. Previous studies in this field were either specific on the therapeutic factors and features for each app or a general strategy for offering apps to users. The aim of the present study is to personalize the programs in a shelf and to rank the shelves according to different people for each user, so that the order of the shelves on a page is shown differently for each user. We also intend to have items that have business aspects to layout, can be considered for good user experience, beneficial stores and developers.In this method, after the feature engineering process, a mechanism is provided to rank each and every app in each shelf independently. For this part, we use ranking methods including MSE, Hinge loss and ListMLE. After that, using the metadata of the shelves and the programs inside them, they ranked the shelves relative to each other using the Random Forest method. Finally, taking into account the consumption of editorial layout, a page is arranged using the calculated weights.Then, he evaluated the results of the study using RSME and nDCG on the data available in the Play Store and the data downloaded from a local app store. The findings show that apps get higher ratings than games. In addition, the home page, which combines editorial and customization, had higher user ratings. In general, for the ranking of programs, about 5% increased in MRR indicators and almost 10% in nDCG indicators. The final model changed by about 30% for games and about 10% for apps. For the entire showcase page layout, the ideal composition is approximately 40% of the editorial amount as well.
- Keywords:
- Recommender System ; Machine Learning ; Mobile Application ; Application Stores ; Mobile Applications Stores ; Applications Personalization
Digital Object List
-
محتواي کتاب
- view
Bookmark