I wanted to make this project for Stock Price Prediction which is still in progress but I’m stuck at LSTM part of it since I don’t understand what amount of loss would be good for this kind of model.

I have used 12 years worth of daily APPLE stock price data from yfinance and coupled it with some technical indicators data to feed to the LSTM with a window of 30 (now 45). I tried training the model with fewer lstm units and fewer epochs but they fail to either capture the trend or lag a lot behind the actual prices or severely underestimate the price (lower predicted prices). That’s why I have increased the amount of LSTM units just this time for experiment and this is the best result I’ve got. I don’t have someone to seek guidance so I’d appreciate if someone could confirm if this much lag and error is acceptable and what else I can do to improve the accuracy.
This is for my third year’s college project, I’ll try to attach the github repo of the colab notebook.

I’m new at this platform so please excuse if there’s any mistake in the asking of the question.

Github repo: https://github.com/baka-onii/Project-Ritchie/tree/main).
I hope it is accessible.