Stock Research Platform 

Problem/Opportunities:

Research for personal investing is time-consuming, with information being overloaded from online sites. Tedious to consolidate all scattered information and summarise it.

​

Solution:

Create an online platform framework to streamline the process by turning complex multi-step scripts into an interactive application. Tackle bottlenecks caused by overloaded information, and merge and summarise it into actionable insights.

​

Approach:

Adopt a systematic top-down approach and allow procedural steps to be performed in sequence during the research process.

Industry trend analysis

Leverage the LLM capabilities to identify the top three industries impacted by a user-defined scenario (e.g. AI chip demands, decrease in interest rates, etc.) and have the model suggest three top stocks with an explanation. The logic is to craft the prompt, coded in the backend, to search for a few websites to retrieve the information before summarising it into insights. The result will be output online in Gradio as well as saved as a copy in Markdown. 

​​

Stock Analysis

This allows users to input the stock ticker and then get the full analysis on the stock, including snapshots, financial forecasts, news & trends, etc. Also included in the output will be the reference sites where that information is being cited. Output in Gradio as well as saving as a hard copy

​​

SWOT Analysis

Strengths, weaknesses, opportunities and threats analysis will be done on the input stock. A quadrant diagram will be drafted online on Gradio, as well as saving a copy for later referral.

Current Price and Info (Display the current price, historical price, and the price set for the last 10 days, P/E ratio, as well as the debt-to-equity ratio.)

​

Challenges:

LLM has its own limitations, especially in terms of time. Therefore, I am using a web scraping method to get the real-time information. As for the latest analysis on the stock, I am using an LLM model that can retrieve live data. As results may vary in the next run, I am keeping a hard copy on file with a date timestamp as part of the filenames and using a strict prompt to avoid hallucination.

Conclusion:

A good end-to-end exercise for stock research. This demonstrates how we can enhance the LLM’s ability to serve as a tool that would help us to solve business problems.

​​

Future considerations:

Consider including more performance indicators and metrics for technical analysis and improving output with graphs for research representation. Enrich with more research functionalities, to include research on EFTs, mutual funds, etc. Host all code in the cloud.

​​

Application of use:

As currently indicated, this can be used for personal investment research and can be extended to be used in brokerage firms and wealth management. The program logic turns complex multi-step scripts into interactive applications, and the methodology can be applied to any area that needs to run a fixed set of separate scripts/queries with user-defined variables to pull relevant data; such as performing market research for any sector, automated document processing, etc.

*Special Note* AI can be used as a tool to help us to make better decisions, but we cannot fully rely on it to make speculative ones. It is not intended to replace a financial advisor, as there are limitations on using this method. (Knowledge not up to date, can hallucinate and produce biased information. Having said that, it is good to help us with the research, summarise news, transcripts, etc., to help us to come up with a better financial decision.