Zhichong Lyu (吕志冲)

This study finds that even experienced investors struggle to recognize recycled old information from multiple sources. Using 17 million news articles from Bloomberg, it evaluates how this mechanism affects the market. Compared to direct reposting, reshuffled old information causes greater price volatility and subsequent reversals. This effect persists in terms of news sentiment, ambiguity, and investor attention. Additionally, while the overall response to old information declines over time, reactions to the reshuffled content are increasing.

This paper investigates whether ChatGPT-4, which is accessible to the internet, can provide valuable investment advice and timely assess financial information. Through field experiments, it was found that ChatGPT-4's ratings are positively correlated with future earnings announcements and stock returns. The empirical results demonstrate that ChatGPT-4 adjusts its ratings based on earnings surprises and news events, and the "attractiveness ratings" strategy generates positive returns.

This paper investigates the potential of ChatGPT and other large language models in predicting stock market returns using news headlines. By analyzing the positive, neutral, and negative impacts of news headlines on company stock prices, the scores generated by ChatGPT show a significant positive correlation with subsequent daily stock returns.

This paper designs a prompting strategy for ChatGPT to identify and analyze anomalies in financial communications, focusing on earnings call transcripts of S&P 500 companies. Large language models like ChatGPT have the potential for financial analysis, offering new insights into interpreting complex text data and its economic consequences on the market.

By using machine learning methods to classify economics papers into different fields and styles, the author analyzes significant changes in research output and citation patterns. The shift toward empirical research in economics is a phenomenon within the field, rather than a result of cross-disciplinary changes. The number of empirical papers published continues to rise, and they are appearing in more influential journals. The growth in citations of empirical papers has even surpassed the output of empirical papers themselves.

TF-IDF (Term Frequency-Inverse Document Frequency) is a widely used weighting technique in information retrieval and text mining that evaluates the importance of a word in a document relative to a collection of documents, increasing with its frequency in the document and decreasing with its frequency across the corpus.

This content explores the differences and similarities between lemmatization and stemming, two key methods for normalizing word forms in natural language processing.

From basic formatting to advanced techniques, this guide will take you into the world of Markdown, helping your documents look more professional and engaging. Whether it's for blogging, document creation, or everyday note-taking, you'll find everything you need here!💡

This paper provides an in-depth analysis of the factors contributing to the significant extension of the time required for publishing papers in top economics journals. Through time series and specific case studies, it explores multiple factors such as the democratization of the publishing process, the increasing complexity of papers, the expansion of specialized fields, and changes in social norms.

This research explores innovative deep learning algorithms, particularly Deep Reinforcement Learning (DRL) using Proximal Policy Optimization (PPO), as well as CNN and LSTM models, to address label imbalance in high-frequency trading and improve profitability, while also considering multi-dimensional data analysis, simulations, and real-time testing.