UROP Proceedings 2020-21

School of Engineering Department of Computer Science and Engineering 111 Practical ML-based Mobile Applications Supervisor: CHATZOPOULOS Dimitrios / CSE Student: ZHENG Hantao / DSCT Course: UROP1100, Fall Mobile applications assisted by machine learning(ML) algorithms can offer sophisticated functions by adapting to the individual usage patterns of the users. In this project, we developed smartphone applications based on flutter, an open-source UI software development kit developed by Google, and integrate the ML function with the help of Firebase, Google’s cloud computing and development platform. The report mainly focuses on the exploration during the term and the approaches made to solve the proposed problems, as well as the resources read for reference. We started with the basics of Flutter including widgets, navigators and other concepts for fundamental GUI design in mobile platforms. Then the powerful plugins Flutter provides such as flutter camera, image picker are introduced to our implementation. We finally tried to combine our application with the various functions supported by Google Firebase, including the cloud database service and face detection, text recognition and other powerful functions using ML algorithms. Practical ML-based Mobile Applications Supervisor: CHATZOPOULOS Dimitrios / CSE Student: KAO Shiu-hong / DSCT Course: UROP1100, Spring With the rapid development of modern technology, people often need to deal with much more information nowadays. In this report, we aim to build an app helping people categorizing articles, such as news, faster. We firstly collected hundreds of sample news on the Internet, and then used a machine learning algorithm to do multi-class classification, dividing these samples into three categories - business, engineering, and science. Finally, we built an Android app on Android Studio, uploading articles onto this app, and used the criteria we got in the machine learning to achieve our target. However, we found that the error rates in the training set and the testing set differ a lot. This may be caused by two reasons, resource collection and characteristic ignorance. We hope these problems can be fixed by considering more possible issues which may happen in the algorithm in the future. Practical ML-based Mobile Applications Supervisor: CHATZOPOULOS Dimitrios / CSE Student: TOH Magdalene Youjun / COMP Course: UROP2100, Spring Recommendations shape lives, from entertainment in videos or movies to purchases on e-commerce websites. This application of machine learning saves time by offering pre-filtered choices to the users. Then, the narrowed selection helps to make quicker decisions as users do not have to filter out irrelevant choices. Recommendations may be applied in a food recipe mobile application. Diverse food recipes are abundant online. However, for everyday cooking, users will use recipes similar in cuisine, techniques, or ingredients. This project uses natural language processing to recommend recipes with higher similarity in ingredients and equipment needed. The two stages of the recommendation are candidate generation and ranking.

RkJQdWJsaXNoZXIy NDk5Njg=