L6

simulator.py

@@ -0,0 +1,115 @@
+import requests as req
+import matplotlib.pyplot as plot
+import numpy as np
+from datetime import datetime
+import random
+
+MAX_NN = 11
+NUM_OF_SIMS = 100
+currPairs = ["ltcusd", "btcusd", "ethusd"]
+noiseChance = 0.1
+noisePercent = 0.01
+currency = currPairs[0]
+
+def fetch_data(currency, startTime, endTime, step):
+    limit = min(1000, (endTime-startTime)//step)
+    data = req.get("https://www.bitstamp.net/api/v2/ohlc/{0}/".format(currency), params={'start': startTime, 'end': endTime, 'step': step, 'limit': limit})
+    return data.json()
+
+def parse_data(rawData):
+    rawData = rawData['data']['ohlc']
+    output = []
+    for d in rawData:
+        entry = {}
+        entry['timestamp'] = int(d['timestamp'])
+        entry['close'] = float(d['close'])
+        entry['open'] = float(d['open'])
+        entry['volume'] = float(d['volume'])
+        output.append(entry)
+    return output
+
+def distance(dataRow1, dataRow2):
+    dist = 0.0
+    dist+=(dataRow1['timestamp'] - dataRow2['timestamp'])**2
+    dist+=(dataRow1['close'] - dataRow2['close'])**2
+    dist+=(dataRow1['open'] - dataRow2['open'])**2
+    dist+=(dataRow1['volume'] - dataRow2['volume'])**2
+    return dist**(1/2)
+
+def nearest_neighbours(data, predictRow):
+    distances = []
+    for dataRow in data:
+        distances.append((dataRow, distance(dataRow, predictRow)))
+    distances.sort(key=lambda tup: tup[1])
+    neighbours = []
+    for i in range(min(len(distances), MAX_NN)):
+        neighbours.append(distances[i][0])
+    return neighbours
+
+def predict_avg(neighbours):
+    valueDif = []
+    volumeAvg = []
+    for n in neighbours:
+        valueDif.append((n['close']-n['open'])/n['close'])
+        volumeAvg.append(n['volume'])
+    return np.mean(valueDif), np.mean(volumeAvg)
+
+def simulate_single(data):
+    predictionsData = []
+    predictionsData.append(data[0])
+    predictions = []
+    for i in range(1, len(data)): #len(data) = simulation for historical period data
+        nn = nearest_neighbours(data, predictionsData[i-1])
+        valueDif, volumeAvg = predict_avg(nn)
+        entry = {}
+        entry['timestamp'] = data[i]['timestamp']
+        entry['volume'] = volumeAvg
+        prevClose = predictionsData[i-1]['close']
+        entry['open'] = prevClose
+        entry['close'] = prevClose + prevClose * valueDif
+        if random.random() <= noiseChance:
+            entry['close']+=entry['close']*noisePercent
+        predictions.append(entry['close'])
+        predictionsData.append(entry)
+
+    return predictions
+
+def plotter(single, mean, data):
+    data.pop()
+    dates = []
+    actualValues = []
+    for d in data:
+        dates.append(datetime.fromtimestamp(d['timestamp']))
+        actualValues.append(d['close'])
+    plot.title(currency)
+    plot.plot(dates, single, label = "Single simulation")
+    plot.plot(dates, mean, label = "100 simulations")
+    plot.plot(dates, actualValues, label = "Actual data")
+    plot.legend()
+    plot.show()
+
+def simulation(currency, startTime, endTime, step):
+    data = fetch_data(currency, startTime, endTime, step)
+    data = parse_data(data)
+    singleSim = simulate_single(data)
+    hundredSims = []
+    for i in range(NUM_OF_SIMS):
+        print(f"Simulation {i+1}...")
+        hundredSims.append(simulate_single(data))
+    meanSim = []
+    limit = len(hundredSims[0])
+    for i in range(limit):
+        meanSim.append(0)
+        for j in range(NUM_OF_SIMS):
+            meanSim[i] += hundredSims[j][i]
+        meanSim[i] = meanSim[i] / NUM_OF_SIMS
+
+    avg = np.average(meanSim)
+    median = np.median(meanSim)
+    std = np.std(meanSim)
+    print(f"Average: {avg}, Median: {median}, Standard Deviation: {std}")
+    plotter(singleSim, meanSim, data)
+
+
+if __name__ == "__main__":
+    simulation(currency, 1561040415, 1591440415, 86400)#currency, starting timestamp, ending timestamp, step in seconds - supported: 60, 180, 300, 900, 1800, 3600, 7200, 14400, 21600, 43200, 86400, 259200