I don't have a quant/algo trading background. My only experience in this area is from several years ago when I dabbled with QuantConnect (IIRC, at that time they didn't have option price history).

However, I have always wondered if one could use stock price history and option price history to do the following:

1. Select an underlying asset (or small set of underlying assets) to focus on

2. Select a simple strategy over the underlying asset such as BUY CALL + BUY PUT at the same price, same expiry date, holding the position until a threshold gain is attained and then liquidating the position.

Of course, you would lose the cost of the CALL + PUT if the position expires without hitting the threshold gain. But part of the algorithm's parameter choice would be to determine the threshold gain% for a specific underlying asset and selecting an underlying asset with a very high probability of exceeding the threshold gain% and thereby a very high probability of positive return on investment.

Arguably, one could train a neural network to do Threshold gain% estimation for an asset based on asset price history and option price history and automate the order placements (both for buy and sell).

For the end-user, the only input they would provide is the investment amount and a target ROI and the system could show a preview of all orders using this strategy that would meet their criteria and execute them at a click of a button (and where permitted, execute them automatically).

Buying Call and Put at same strike and expiration is just simple Long Straddle position, typically used near Earnings announcements, when you are expecting underlying to move significantly but not sure of direction. You don't need a neural network to do this. Look up books by Lawrence McMillan and Sheldon Nutenberg to learn more about options trading.