Artificial intelligence, neural networks, machine learning
- Thread starter irniger
- Start date
Yes. I built my own C# SDK which uses supervised machine learning techniques. Its strength is it can test thousands of feature combinations very quickly where a feature (i.e. Close related to LowerBollingerBand) can be dynamic or static. A dynamic feature will be calculated for each point in time while a static feature (i.e. Close < LowerBollingerBand) will remain so and has the effect of eliminating the need to backtest strategies created on the fly where that condition is not true. The technique is very useful in that I can test thousands of strategies in under a minute, filter those strategies by result characteristics and then retest those filtered strategies on out of sample data. I might start with 10,000 strategies and the end result after filtering and retesting is a handful of strategies that warrant further investigation. I should have a beta version of the SDK around the end of the month.
Well see for yourself. Here is a Yahoo Finance chart comparing the AIEQ (The AI Powered Equity ETF) vs SPY, DIA and QQQ for the past 3 months.
https://goo.gl/LbJZfr
https://goo.gl/LbJZfr
I just uploaded the historical dataset into a neural network and it modelled it automatically. Then I hooked up the model to my broker through a basic execution algorithm. Loaded up the account with money and I have way more time to answer questions on this forum now while watching my account lose money.
In fairness, AIEQ had a terrible first month but has been gaining ground on the indices since.
I heard that Tyrannosaurus rex went extinct because his front limbs were too short and they weren’t even able to pleasure themselves
On the topic, my sense is that so far nobody is successful extracting alpha using AI methods alone. There are a few successful funds but almost all of them use “ML+“ model.
Yes since the beginning of the year, AIEQ has really performed well. I wonder why it suddenly started doing so well. Perhaps they rebalanced.