Stock analysis model based on graph database

Some thoughts

I actually have great ambitions for this project, but I can only make what I have now due to limited resources.

Before I get the data, my ultimate goal is to make a business analysis AI that can automatically trade and make profits based on global real-time news data and knowledge pedigree.

When I first saw the title of the dataset, GDELT Open Intelligence (Geopolitical), I thought it was a global geopolitical conflict data, and I could make a global geopolitical conflict analysis AI to help business analysis.

But after a few rounds of queries, I went to the database server UI interface and looked carefully, only to find that this thing actually only has data on Africa, and there are only 22 countries. It is seriously inconsistent with the title.

What it does

At present, I only have geopolitical data on Africa, and most of them are armed conflicts. At most, I can only do a very simple consulting AI for Africa. Moreover, the data in this dataset is pitifully small and not timely, so it has no commercial value.

What we learned

I can divert the upcoming AI by preprocessing the input information, which gave me great inspiration. In other words, I can create several use case schemas to divert the upcoming operations of AI. This inspired me deeply.

Because I know that the cost and expense of training an LLM are very outrageous. If the ability to search the database is added to the use of LLM, it will significantly improve the intelligence of LLM at a very low cost.

In fact, this graph database also limits my ability to play, because relational database is the most common way to store data. Most of the data is in relational database, and some are very difficult to convert into graph database. This makes it very difficult for me to use other data or prepare my own data.

However, the efficient retrieval capability of graph database is indeed the future, especially when dealing with complex relationships, because it is obvious that graph queries are more efficient than traditional relational databases, and the computational overhead of sql join in the face of complex relationships is very large, from O (n^n) to O (1) or O(n).

This is the ultimate goal which is the business analysis AI, I will finish as soon as I have enough data and a powerful computer.

Built With

• arangedb
• python