School of Science Department of Chemistry 8 Methodology Development for Time Series Data Supervisor: SU Haibin / CHEM Student: DING Kangyuan / CPEG SZE Wai Yin / QFIN Course: UROP 1100, Summer UROP 1100, Summer Stock market data is considered dynamic and changeable, predicting stock prices using different technical methods has been a puzzle for people for a long time. The goal of this study is to evaluate the performance of two wide-use time series forecast methods: ARIMA, LSTM in the stock market, and compare their performance in different industry, size of company and time, trying to find underlying predictability patterns among instruments, help practitioners acknowledge which method is better to get more accurate results in certain circumstance when they want to make prediction or analysis on their data. Methodology Development for Time Series Data Supervisor: SU Haibin / CHEM Student: LAI Yi-an / QFIN Course: UROP 1100, Summer Exchange Traded Funds (ETFs) have surged in popularity due to their simplicity, cost-effectiveness, and easydiversification. Standard ETF typically track specific indices or a basket of stocks, such as QQQ by Invesco, which tracks the Nasdaq 100, and XLK by SPDR Funds, which delicately targets the technology sector. However, the needs of investors have evolved, leading to the creation of leveraged and inverse ETFs like TQQQ by ProShares, which seeks to deliver three times the return of the Nasdaq 100 using derivatives like Total Return Swaps (TRS). Despite their appeal, the U.S. Securities and Exchange Commission (SEC) advises against long-term investments in leveraged ETFs (LETF) due to Volatility Decay (or Volatility Drag). This paper delves into this concept and examines the use of machine learning to optimise LETF investments and outperform benchmarks. We provide a comparative analysis of QQQ and TQQQ, illustrate the impact of volatility decay, and propose strategies utilising technical indicators and machine learning algorithms, such as Decision Tree, Support Vector Machine, and Neural Networks. Methodology Development for Time Series Data Supervisor: SU Haibin / CHEM Student: SU Yilin Hanako / MATH-AM Course: UROP 1100, Summer In the diverse time-dependent phenomena, the one overarching theme is to develop quantitative methods which are crucial for understanding the evolution of complex systems. This highlights the importance of the time-dependent nature of chemical and physical phenomena, and the need for theoretical frameworks to explain temporal dynamics. This report aims to explore the scientific background, as well as mathematical tools involved in both kinetics and dynamics to exemplify the critical role of one key factor - time. In particular, this report will dive into the theories and principles of kinetics in the field of electrochemistry, and dynamics with multiscale character.
RkJQdWJsaXNoZXIy NDk5Njg=