Generative AI x Investment Analysis: 3 steps to automate information collection with Python

Why automate investment analysis with generative AI?

Conclusion: Information processing speed and comprehensiveness are greatly improved

One word explanation

Generation AI = AI that can understand, summarize, and generate sentences

Assignment

• Too much information to disclose
• I can't keep up with the news
• Human power causes bias.

Solved

• Automatic collection → automatic organization → automatic summary
• Humans can concentrate on “judgment”

Overall configuration: Basic design of RAG

Conclusion: Combination of search + generation is optimal

One word explanation

RAG = AI answers after searching for necessary information

Processing flow

1. Data collection (API/scraping)
2. Text division (chunking)
3. Vectorization (for search)
4. Similar search
5. Summary generation

Points

• "Passing the correct information" is the key to accuracy
• Less false information than AI alone

Step 1: Automate information collection

Conclusion: First, create a system to collect data

Method

• RSS feed
• API (News/Disclosure)
• Web scraping

Python example (simple)

import feedparser

feed = feedparser.parse("https://example.com/rss")

for entry in feed.entries:
    print(entry.title, entry.link)

Practical points

• Decide on update frequency (e.g. every hour)
• Also save noise information temporarily

Step 2: Keyword extraction and filtering

Conclusion: Narrow down to only the necessary information

One word explanation

Filtering = Extract only information that meets the conditions

Example

• “Profit increase” “Upward revision”
• "M&A" "Share buyback"

Python image

keywords = ["profit growth", "Key point"]

filtered = [text for text in texts if any(k in text for k in keywords)]

Points

• Review keywords regularly
• Beware of excessive filters

Step 3: Summarize with generation AI

Conclusion: Get the essence in a short time

One word explanation

Summary = Extract only the important parts

How to use it

• News summary
• Key points for disclosure information
• Positive/negative judgment

Output example

• 3 main points
• Risk factors
• Investment decision materials

Practical usage (important)

Conclusion: Use as an analysis aid

###NG

• Trust the AI output as is

OK

• Use as a tool to organize judgment materials

Role division

Points to note when installing

• Check the authenticity of your data
• Test with historical data
• Manage costs (API/calculation)

Common mistakes

• Aiming for perfect automation
• Too much noise removal
• Take the AI summary with a grain of salt

Summary

• Generative AI greatly increases the efficiency of information processing
• Accuracy and reliability can be improved with RAG
• The final decision is always made by humans

Action steps

• ① Automate information collection using RSS and API
• ② Filter by keyword
• ③ Summarize with AI and use for judgment

---