School of Engineering Department of Computer Science and Engineering 87 AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: SU Chen-yi / DSCT Course: UROP 1000, Summer This paper explores the application of diffusion models to LoRa-based localization tasks, particularly in search and rescue operations. Our preliminary results demonstrate the model’s ability to generate signal heatmaps, identifying strong signal areas to guide drones more effectively. We also discuss ongoing work in conditional training, aiming to generate complete heatmaps from limited data points. The approach shows significant potential for enhancing search and rescue efforts, providing a more accurate and automated method for locating individuals in need. AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: TSANG Kin Ho / DSCT Course: UROP 1100, Fall This report discusses a possible implementation of an ‘AI Nurse’ in the context of blood pressure prediction and issuing potential health warnings. 3 different machine-learning approaches to the problem have been constructed in Python by utilizing relevant libraries. The model was then tested with a toy dataset and results were recorded. The context of this project is to employ the models constructed to predict an upcoming blood pressure measurement based on previously measured records. The 3 machine learning models considered: Linear Regression (LR), Long-Short-Term-Memory Recurrent Neural Network (LSTM), and Auto-Regressive Integrated Moving Average (ARIMA), all possess various strengths and weaknesses in their prediction of future blood pressure measurement. AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: WANG Zekai / CPEG Course: UROP 1100, Fall The use of SSID collected from Wifi fingerprint for indoor positioning is a feasible solution, especially in Wifidense areas such as shopping malls. This can be achieved through a trained machine learning model that only requires SSID and signal strength as input for prediction, resulting in the output location. This approach reduces human effort and is easily scalable for larger areas. As a result, we decided to include SSID as an additional approach in the "Pervasive Positioning Standard for Fingerprint-based and Proximity-based Systems" to serve clients when they request positioning services. This report discusses our current approach and its limitations, followed by an examination of the "SSID Localization Approach", including the necessary preparations in the pre-collection stage and the integration into our existing mobile application system.
RkJQdWJsaXNoZXIy NDk5Njg=