AI Buy/Sell SIgnals by price prediction

4 jam yang lalu

//version=5

indicator("AI Buy/Sell SIgnals by price prediction", overlay=true)

learning_times = input.int(200, "Learning times")

ema_length = input.int(1, "EMA length")

learn_filter_length = input.int(5, "SMA Filter length")

learning_block = input.bool(title="Filter Learning data", defval=true)

reaction = input.int(1, "Reaction (1-3)")

a = close



var input_list = array.new_float(100)

var weights = array.new_float(100)

var outt = array.new_float(2)



//def info table

var tab = label.new(bar_index, high, ".", style=label.style_label_left, color=color.white)

infotable = table.new(position=position.top_right, columns=3, rows=3, bgcolor=color.new(color.white, 0))

label.delete(tab)



get_errg(input_array, weights_array, len_of_both) =>

out = 0

for x = 0 to len_of_both

out += int(array.get(weights_array, x) * array.get(input_array, x))

out



//getting inputs

array.set(input_list, 0, ta.valuewhen(bar_index, close, 10))

array.set(input_list, 1, ta.valuewhen(bar_index, close, 20))

array.set(input_list, 2, ta.valuewhen(bar_index, close, 30))

array.set(input_list, 3, ta.valuewhen(bar_index, close, 40))

array.set(input_list, 4, ta.valuewhen(bar_index, close, 50))

array.set(input_list, 5, ta.valuewhen(bar_index, close, 60))

array.set(input_list, 6, ta.valuewhen(bar_index, close, 70))

array.set(input_list, 7, ta.valuewhen(bar_index, close, 80))

array.set(input_list, 8, ta.valuewhen(bar_index, close, 90))

array.set(input_list, 9, ta.valuewhen(bar_index, close, 100))

array.set(input_list, 10, ta.valuewhen(bar_index, open, 10))

array.set(input_list, 11, ta.valuewhen(bar_index, open, 20))

array.set(input_list, 12, ta.valuewhen(bar_index, open, 30))

array.set(input_list, 13, ta.valuewhen(bar_index, open, 40))

array.set(input_list, 14, ta.valuewhen(bar_index, open, 50))

array.set(input_list, 15, ta.valuewhen(bar_index, open, 60))

array.set(input_list, 16, ta.valuewhen(bar_index, open, 70))

array.set(input_list, 17, ta.valuewhen(bar_index, open, 80))

array.set(input_list, 18, ta.valuewhen(bar_index, open, 90))

array.set(input_list, 19, ta.valuewhen(bar_index, open, 100))



// teaching neural network

sma = ta.sma(ta.ema(close, 10), learn_filter_length)

if math.abs(ta.valuewhen(bar_index, sma, 1) - sma) > close / 10000 or not learning_block

for rn = 0 to learning_times

for list_number = 0 to 19

if rn == 0

array.set(weights, list_number, 1) // Initialisiere die Gewichte mit 1

else

target_output = close[50]

current_output = get_errg(input_list, weights, 19)

current_input = array.get(input_list, list_number)

target_input = target_output / current_output * current_input // Berechne die entsprechende Eingabe für das Gewicht

array.set(weights, list_number, target_input)



// getting new output

array.set(outt, 0, get_errg(input_list, weights, 19))



var col = #ff1100

var table_i_col = ''

var pcol = #ff1100



// getting signals

if ta.ema(ta.valuewhen(bar_index, array.get(outt, 0), 1), ema_length) < ta.sma(ta.ema(array.get(outt, 0), ema_length), 10)

col := #39ff14

table_i_col := 'AI Up'

if ta.ema(ta.valuewhen(bar_index, array.get(outt, 0), 1), ema_length) > ta.sma(ta.ema(array.get(outt, 0), ema_length), 10)

col := #ff1100

table_i_col := 'AI down'



if ta.valuewhen(bar_index, col, 50) == col and ta.valuewhen(bar_index, col, 10) == ta.valuewhen(bar_index, col, 20) and ta.valuewhen(bar_index, col, 30) == ta.valuewhen(bar_index, col, 40) and reaction == 1

pcol := col



if ta.valuewhen(bar_index, col, 50) == col and ta.valuewhen(bar_index, col, 10) == ta.valuewhen(bar_index, col, 20) and reaction == 2

pcol := col



if ta.valuewhen(bar_index, col, 50) == col and reaction == 3

pcol := col



// plotting all info

plot(0, "plot2", col, offset=50)

plot(ta.sma(ta.ema(close, 10), 10), color=ta.valuewhen(bar_index, pcol, 50), linewidth=2)



table.cell(infotable, 0, 0, str.tostring(float(array.get(outt, 0))))

table.cell(infotable, 0, 1, str.tostring(float(ta.valuewhen(bar_index, array.get(outt, 0), 50))))

table.cell(infotable, 0, 2, str.tostring(table_i_col))