UROP Proceedings 2020-21

School of Engineering Department of Computer Science and Engineering 147 Knowledge Discovery over Database Supervisor: WONG Raymond Chi Wing / CSE Student: WU Lijia / COMP Course: UROP1100, Fall Interactive regret minimization is often used to retrieve the favourite subset of the user from a large database containing many tuples. To achieve strongly truthful results, users are required to choose his/her favourite tuple out of a small subset from the database for a small number of times. Specifically, an unknown utility function is introduced in the k-regret minimization to represent the preference of the user. With such utility function, the k-regret minimization query can reduce the regret ratio of the user. This report mainly focuses on the impacts of different forms of utility function in strongly truthful k-regret minimization query with user interactions, which includes Linear, Quadratic and Gaussian form of utility functions. Knowledge Discovery over Database Supervisor: WONG Raymond Chi Wing / CSE Student: DO Thuy Trang / DSCT Course: UROP2100, Fall The recommender system has played an important part in many online services such as e-commerce. To provide customers the most related item or service that they are likely to click, we need to predict users’ next action based on the sequence of their previous actions in a given session. From the paper: "Sessionbased Recommendation with Local Invariance", we could find two significant ideas: order of actions in a given session has a great impact on the next click of customers, and the definition of "short sub-sequence" when the products have high similarity. Based on the ideas of this paper, in this report, we propose two methods to improve the accuracy of prediction, which is Recurrent Neural Network and Bidirectional Recurrent Neural Network. Research on Mining Course Structure Supervisor: WONG Raymond Chi Wing / CSE Student: LIU Ziming / COSC Course: UROP1100, Fall In this UROP 1100 project, students should perform data mining tasks on all important components of courses on a given list to extract their structure and correlation among different courses. As a well-developed data mining application, T-Music Web System is a maintainable and scalable system to deploy T-Music algorithm, which is an innovative algorithm that can compose a series of musical notes from a given series of input lyrics. The group members of this project should fulfill the unfinished goals listed in the future work of the T-Music Web System Report, which requires the members to master basic front-end web programming skill in order to implement the new features in the web system.

RkJQdWJsaXNoZXIy NDk5Njg=