Nowadays I want to use machine learning on trading; I want to use "Neural Networks" to select some situations(up%,vol and so on) in which a big swing is coming; shall I get some tips about this？
I've got a tip for you. There is no holy grail. Also, if you have no background in algo-trading or programming, then the chances of you being able to "get into" machine learning is pretty slim.
thanks for your reminding; I have programming background and I can scan all stocks of the market each day to select stocks whose current movings fit my some models; Now I want to use ML to add another filter to get a better win ratio.
You can start with Coursera's course on Machine Learning which will take you 12 weeks, assuming you have basics of calculus/algebra (otherwise start with calculus course that takes 8-10 weeks). If you have no knowledge of programming that's probably going to be the hardest, count a good 2-5 months before you're fluent in Python and/or C#.
Machine Learning can forecast anything from volatility to how long it takes me to make spaghetti. The question of "reliability" is more important when dealing with financial data.
I want to use "Neural Networks" to select some situations(up%,vol and so on) in which a big swing is coming; _____________________________________________________________________________ Market internals are in my opinion the most powerful tells a market reveals that can get you into the big trend days and more importantly can get you in synch with the institutional buy programs that kick in during major capitulation at major swing lows. If you look hard enough you will see specific market intrenal numbers that repeat when a reversal occurs which kicks in the buy programs sending the market rocketing up on major up legs lasting days/weeks/months.
Why do people bother replying why they have zero to add, obviously no one has any credentials. Try NuPic, they have a stock based demo and is one of the better ones. All this "knowledge" on these forums and no one knows a thing. Is ML worth it, if you can find an advantage against the millions of other traders by using it, of course.
All system based & AI is rooted in the programmers discretionary thinking - this chart illustrates that concept. At the and of the day you still have to have the skills/instinct, courage, risk management. etc to get anywhere. AI goes way back and has not over taken discretionary for the reason already stated. Trying to over compensate with AI/deep learning without a strong skill set as a discretionary trader is wishful thinking that you computer will do for you what you can not do for yourself. Trading is like dieting - everyone knows how to make $ - buy low, sell high but the masses sell at lows and buy at tops. Everyone knows how to loose weight, yet few ever do it. Instead they search feverishly for the next pill or fad. Markets moves follow a non-normal distribution curve highlighted by large tails, narrow mid-ranges and a high peak - it's been proven in academia. Simple, yet few can ever capitalize on this.
I agree with your first statement that there is no holy grail. If all it took was applying machine learning to the stock market, then programmers would be making all the money. I disagree with your second statement. We all started out with no background in anything. Most people can pick up programming if they have the interest. It's probably one of the most accessible fields out there since there are so many free resources.