Beginner questions about buy/hold/sell signals, or other labels for ML.

Hi!

I wanted to make a trading bot as a hobby project, but I've read on this subreddit that it is not quite good to try to predict stock prices using ML. Given historical stock data, I would like to predict whether I should buy/sell/hold stock in the next day. I have few questions regarding this problem:

1. Is buy/sell/hold labels for classification a good way to formulate this problem?
2. Assuming that classification is a good way to predict actions, how I should generate labels for historical data? Using some technical analysis and using lets say RSI, MACD or other indicators? How I should approach this?
3. What are other ways to model this problem, if I want to use it later in trading bot (I assume that actions would be taken daily granurality)?

Also I would be grateful for any tips/resources on how to start this project. Sorry for newbie post, but things are not clear for me yet 🙂