Dies ist eine Verbesserung der Skalierten Normalisierten Vektorstrategie von drkhodakarami, die hauptsächlich Aktivierungsfunktionen hinzufügt, um die Leistung der Strategie zu verbessern.
Lösungen:
Auf der Grundlage der Arbeit von drkhodakarami
/*backtest start: 2023-01-15 00:00:00 end: 2024-01-21 00:00:00 period: 1d basePeriod: 1h exchanges: [{"eid":"Futures_Binance","currency":"BTC_USDT"}] */ //@version=4 // author: capissimo strategy("Scaled Normalized Vector Strategy, ver.4", precision=2, overlay=false) // This is a modification of my Scaled Normalized Vector Strategy // original: Drkhodakarami (https://www.tradingview.com/script/Fxv2xFWe-Normalized-Vector-Strategy-By-Drkhodakarami-Opensource/) price = input(close, "Price Data") tf = input(18, "Timeframe", minval=1, maxval=1440) thresh = input(14., "Threshold", minval=.1, step=.1) div = input(1000000,"Divisor", options=[1,10,100,1000,10000,100000,1000000,10000000,100000000]) mmx = input(233, "Minimax Lookback", options=[1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233, 377, 610, 987, 1597, 2584]) showVol = input(false, "Volume") useold = input(true, "Use Old System") method = input("Swish", "Activation", options=["Step", "LReLU", "Swish", "None"]) scaleMinimax(X, p, min, max) => hi = highest(X, p), lo = lowest(X, p) (max - min) * (X - lo)/(hi - lo) + min getdiff(prc, tf) => prev = scaleMinimax((useold ? security(syminfo.tickerid, tostring(tf), prc[1], barmerge.gaps_off, barmerge.lookahead_on) : security(syminfo.tickerid, tostring(tf), prc[1])), tf, 0, 1) curr = scaleMinimax((useold ? security(syminfo.tickerid, tostring(tf), hlc3, barmerge.gaps_off, barmerge.lookahead_on) : security(syminfo.tickerid, tostring(tf), hlc3)), tf, 0, 1) (curr/prev) - 1 relu(x) => max(x, 0) lrelu(x, alpha) => relu(x) - alpha * relu(-x) step(x) => x >= 0 ? 1 : -1 log2(x) => log(x) / log(2) sigmoid(x) => 1 / (1 + exp(-x)) swish(x) => x * sigmoid(x) f(m) => method==m vol = useold ? security(syminfo.tickerid, tostring(tf), volume, barmerge.gaps_off, barmerge.lookahead_on) : security(syminfo.tickerid, tostring(tf), volume) obv = cum(change(price) > 0 ? vol : change(price) < 0 ? -vol : 0*vol) prix = showVol ? obv : price x = getdiff(prix, tf) p = f("Swish") ? swish(x) : f("Step") ? step(x) : f("LReLU") ? lrelu(x, .8) : x th = thresh/div long = crossover(p, th) short= crossunder(p, -th) lime = color.new(color.lime, 10), fuchsia = color.new(color.fuchsia, 10), black = color.new(color.black, 100), gray = color.new(color.gray, 50) bg = long ? lime : short ? fuchsia : black cl = p > th ? color.green : p < -th ? color.red : color.silver bgcolor(bg, editable=false) plot(scaleMinimax(th, mmx, -1, 1), color=lime, editable=false, transp=0) hline(0, linestyle=hline.style_dotted, title="base line", color=gray, editable=false) plot(scaleMinimax(-th, mmx, -1, 1), color=fuchsia, editable=false, transp=0) plot(scaleMinimax(p, mmx, -1, 1), color=cl, style=plot.style_histogram, transp=70, editable=false) plot(scaleMinimax(p, mmx, -1, 1), color=cl, style=plot.style_linebr, title="prediction", transp=0, editable=false) strategy.entry("L", true, 1, when=long) strategy.entry("S", false, 1, when=short) alertcondition(long, title='Long', message='Long Signal!') alertcondition(short, title='Short', message='Short Signal!') //*** Karobein Oscillator per = input(8, "Karobein Osc Lookback") prix2 = ema(price, per) a = ema(prix2 < prix2[1] ? prix2/prix2[1] : 0, per) b = ema(prix2 > prix2[1] ? prix2/prix2[1] : 0, per) c = (prix2/prix2[1])/(prix2/prix2[1] + b) d = 2*((prix2/prix2[1])/(prix2/prix2[1] + c*a)) - 1 plot(scaleMinimax(d, mmx, -1, 1), color=color.orange, transp=0)