120 points by pfinancegenius 1 year ago flag hide 24 comments
finance_ai_creator 1 year ago next
Hey HN, I'm thrilled to present my new AI-powered personal finance tool. It uses advanced machine learning algorithms to analyze spending habits, make budget suggestions, and even offer investment advice.
techsavvyuser 1 year ago next
That sounds amazing! How accurate are the budget suggestions?
finance_ai_creator 1 year ago next
The accuracy of budget suggestions has been great, with an average error of just 2.7% based on our internal testing. Users can set their preferred risk level for a more personalized experience.
anonymous 1 year ago next
Have you considered incorporating XYZ open-source financial library for accuracy improvement?
finance_ai_creator 1 year ago next
@anonymous, Thank you for the suggestion! We actually considered using the XYZ library and are still evaluating its benefits for our specific use case.
eagertolearn 1 year ago next
Any plans for sharing the details of your algorithm implementation to contribute to the open-source community?
finance_ai_creator 1 year ago next
@eagertolearn, Our primary focus is to improve the product and user experience, but we do plan to open-source our deprecated models for educational purposes. Stay tuned!
cautiousinvestor 1 year ago prev next
Any idea how the investment advice compares to other platforms? I prefer a more conservative approach.
finance_ai_creator 1 year ago next
Our investment algorithms consider various factors and can accommodate conservative strategies. They've been performing very well in backtesting.
questioninguser 1 year ago next
What kind of documents is your team using for algorithm training? Receipts, bank statements, or something else?
finance_ai_creator 1 year ago next
@questioninguser The algorithms use a variety of financial documents, primarily bank statements and credit card reports, for training.
doubtingdave 1 year ago next
Could you provide more information on how the algorithms ensure fairness and avoid discrimination?
finance_ai_creator 1 year ago next
@doubtingdave, We monitor fairness using statistical techniques and conduct regular audits. Our algorithms use socioeconomic variables as opposed to personal data to prevent discrimination.
dataengineer 1 year ago next
Hey I recently saw a talk about addressing these issues with different types of fairness constraints. Have you thought about integrating those?
finance_ai_creator 1 year ago next
@dataengineer, I appreciate the suggestion. Actually, after our launch, we plan on implementing similar fairness constraints. Thank you for bringing this to our attention.
supportivecoder 1 year ago prev next
As someone who's been following your side projects, I'm not surprised you built something so cool! Good stuff.
finance_ai_creator 1 year ago next
@newbie_developer We use a mix of Python (Flask for web, Scikit-learn for ML) and React for the UI. We can discuss it further if you're interested.
newbie_developer 1 year ago next
That's a great combination. I'm definitely going to check out Flask and Scikit-learn, and hopefully will build something useful!
finance_ai_creator 1 year ago next
@newbie_developer, Good luck! If you need any help or guidance, feel free to reach out.
newbie_developer 1 year ago prev next
Totally off-topic, but what tech stack are you using for this project? Curious as I'm trying to learn.
curiouscoder 1 year ago next
Out of sheer curiosity, has there been any cases of over/under-optimization during training?
finance_ai_creator 1 year ago next
@curiouscoder, Not in this particular iteration, but we'll keep an eye on it and adapt as needed. It's an essential concern for us.
gnome 1 year ago prev next
The demo looks pretty cool! I've been looking for something to help me manage investments, and a more conservative approach might work for me. Any alternative to signing up before testing it out?
finance_ai_creator 1 year ago next
@gnome, Thanks for the kind words. You can try our demo via this link <https://finance-ai-demo.org/>. For security, you will only be able to use it with limited functionality unless you sign up for an account. We understand your concerns and are looking into providing limited testing without accounts.