Updated DBA4761 material

This course is for students interested in financial modeling programming. Based on previous experiences, there is a great demand for students who can apply the tidy data modeling techniques covered in class.

[1] For large language models (LLMs) to be viable in public markets investing, every content generated by the AI must be attributed to its sources. In class we will use a large and clean, domain-specific data on Meta Platforms, Inc. (https://github.com/rnfermincota/academic/tree/main/research/traditional_assets/valuations/META/Notes), to fine-tune an LLM that is best suited for investment research on the company, thereby minimizing hallucination by citing sources of information. LLMs has enormous limitations and the large language model revolution brings with it newly existential and unique risk to research analysts. From the most obvious pitfalls like hallucinated details, fabricated stories and wrong assumptions, whether it be cash flows, growth and/or risk. This coincides precisely with the information retrieval and sentence-by-sentence citation model covered in class, which works effectively to pinpoint the origin of information, ensuring hallucination is minimized.?

[2] Fixedincome investing today requires a complex, interdisciplinary set of skills - some of which are necessary, others, sufficient. I created the following lectures partly in service to students who are interested in fixed income investing. The learning process starts with country risk premium (https://rpubs.com/rafael_nicolas/crp) and effective cost of debt (https://github.com/rnfermincota/academic/blob/main/research/traditional_assets/database/effective-cost-debt.pdf), and ends with fixed income portfolio engineering (https://rpubs.com/rafael_nicolas/fixed_income_portfolio_mgmt) and bond pricing (https://rpubs.com/rafael_nicolas/fixed_income_relative_value).

[3] Forecasts the return for each stock in the M6 competition and then apply the aggregated pipeline onto the portfolio and rebalance the allocation regularly: https://github.com/rnfermincota/academic/tree/main/teaching/NUS/Statistical-Learning/3-Forecasting

[4] Prediction for at least N time series and at least one model for the M5 Walmart competition: https://github.com/rnfermincota/academic/tree/main/teaching/NUS/Statistical-Learning/3-Forecasting

[5] With the tidymodels framework and a lot of feature engineering one could build systematic trading strategies: https://github.com/rnfermincota/academic/tree/main/teaching/NUS/Statistical-Learning/2-Cryptocurrencies

[6] A sentiment analysis is conducted on the Federal Reserve Board BeigeBook to compare the sentiments of reports across time. We will be focusing on the use of BERT models in performing the said analysis by leveraging on various Python packages. https://rpubs.com/qyingtong/Beige_Book_BERT

[7] Download the letters from Berkshire Hathaway, WarrenBuffett's company and then implement a sentiment analysis: https://rpubs.com/rafael_nicolas/berkshire_sentiment

[8] Treasure trove of data science recipes to help students achieve operational excellence, and develop competitive edge in the privateequity industry: https://rpubs.com/rafael_nicolas/nus_pe_ds_course

[9] Airbnb Data Analysis for Singapore: Separate, clean, upload and analyze host and listing data taken from Airbnb for Singapore. https://rpubs.com/rafael_nicolas/airbnb_singapore_data_analysis

[10] Formula 1 database of historical racing data (1950-Present) using R/tidyverse: https://rpubs.com/rafael_nicolas/f1_db

[11] Estimate the optimal capital structure and excess return on capital for #Singaporean companies across industries: https://rpubs.com/rafael_nicolas/singapore_extracting_data

[12] Applying the tidy data principles to the Saudi aramco IPO Valuation: https://rpubs.com/rafael_nicolas/aramco