Initial commit

Author: calebheinzman

lib/DataRetrieval.py

@@ -0,0 +1,115 @@
+from Historic_Crypto import HistoricalData
+from datetime import datetime
+import yfinance as yf
+import talib as ta
+
+'''
+All data goes from earliest to latest and is stored in a data frame.
+You can access individual values by doing: data['High'], data['Low']...
+'''
+
+
+'''
+This functions takes a symbol for a crypto and returns it's historic data.
+
+Input:
+    symbol (Required) - Takes the exchanged pairs: ETH-USD or BTC_ETH, or BTC-USD
+    interval (optional) - Takes the interval of prices: 1d, 1h, 5m or 1m.
+    start_date (optional) - The earliest historic data you want in form YYYY-MM-DD-HH-SS: 2000-01-01-00-00
+    end_date (optional) - The latest historic data you want in form YYYY-MM-DD-HH-SS: 2000-01-01-00-00
+
+Output:
+    A data frame with the High, Low, Open, Close and Volume prices. This is the historic prices. By default it returns
+    all daily prices unless otherwise specified.
+    
+Examples:
+    get_crypto_data('BTC-USD',interval='1h',start_date='2021-04-01-00-00')
+    get_crypto_data('BTC-USD',interval='1h',start_date='2021-04-01-00-00',end_date='2021-06-01-00-00')
+    get_crypto_data('ETH-USD')
+'''
+def get_crypto_data(symbol,**kwargs):
+    interval = kwargs.get('interval','1d')
+    start_date = kwargs.get('start_date', '2000-01-01-00-00')
+    end_date = kwargs.get('end_date', datetime.today().strftime('%Y-%m-%d-%H-%M'))
+    seconds = 0
+    if interval == '1d':
+        seconds = 86400
+    elif interval == '1h':
+        seconds = 3600
+    elif interval == '5m':
+        seconds = 300
+    elif interval == '1m':
+        seconds = 60
+    data = HistoricalData(symbol,seconds,start_date,end_date).retrieve_data()
+    data = data.rename(columns={"open": "Open", "high": "High","low": "Low", "close": "Close",'volume':'Volume'})
+    return data
+
+
+'''
+This functions takes a symbol for a stock and returns it's historic data.
+
+Input:
+    symbol (Required) - Takes the exchanged pairs: ETH-USD or BTC_ETH, or BTC-USD
+    interval (optional) - Takes the interval of prices: 1d, 1h, 5m or 1m.
+    start_date (optional) - The earliest historic data you want in form YYYY-MM-DD: 2000-01-01
+    end_date (optional) - The latest historic data you want in form YYYY-MM-DD: 2000-01-01
+    period (optional) - Instead of specifying an end_date you can specify a period: 1d, 1y, 1m, 2y
+
+Output:
+    A data frame with the High, Low, Open, Close and Volume prices. This is the historic prices. By default it returns
+    all daily prices unless otherwise specified.
+
+** Note: yFinance is limited in what it can return in terms of intraday data so be aware.
+
+Examples:
+    data_source = DR.get_stock_data('DIA',interval='1d',period='2y')
+    data_source = DR.get_stock_data('AAPL',start_date='2021-04-01',end_date='2021-06-01')
+'''
+def get_stock_data(symbol,**kwargs):
+    interval = kwargs.get('interval','1d')
+    start_date = kwargs.get('start_date', '2000-01-01')
+    end_date = kwargs.get('end_date', datetime.today().strftime('%Y-%m-%d'))
+    period = kwargs.get('period', None)
+
+    stock = yf.Ticker(symbol)
+    if period != None:
+        return stock.history(period=period,interval=interval)
+
+    return stock.history(interval=interval,start=start_date,end=end_date)
+
+'''
+This functions takes a data frame of historic data and adds the SMA to it.
+
+Input:
+    data (required) - A dataframe of historic data that SMA will be added to.
+
+Output:
+    A data frame with the SMA added.
+Examples:
+    add_SMA(data_source)
+'''
+def add_SMA(data):
+    # Adding SMA
+    sma = ta.SMA(data['Close'])
+    data['SMA'] = sma
+    data= data.iloc[29:, :]
+    return data
+
+'''
+This functions takes a data frame of historic data and adds the EMA to it.
+
+Input:
+    data (required) - A dataframe of historic data that EMA will be added to.
+
+Output:
+    A data frame with the EMA added.
+Examples:
+    add_EMA(data_source)
+'''
+def add_EMA(data):
+    # Adding EMA
+    ema = ta.EMA(data['Close'])
+    data['EMA'] = ema
+    data= data.iloc[29:, :]
+    return data
+

lib/Metrics.py

@@ -0,0 +1,61 @@
+import sklearn.metrics as skm
+
+def calculate_classification(test_x,test_y,model,label_scaler):
+    # Getting predictions for the test set.
+    pred_y = model(test_x)
+
+    # Transforming the labels back to original space
+    orig_y = label_scaler.inverse_transform(test_y.numpy()).tolist()
+    pred_y = label_scaler.inverse_transform(pred_y.detach().numpy()).tolist()
+
+    # Transforming labels to be returns instead of price
+    rtrn_y = [orig_y[i][0] - orig_y[i - 1][0] for i in range(1, len(orig_y))]
+    rtrn_pred = [pred_y[i][0] - pred_y[i - 1][0] for i in range(1, len(pred_y))]
+
+    # Getting Binary Classification
+    bin_y = [1 if i > 0 else 0 for i in rtrn_y]
+    bin_pred = [1 if i > 0 else 0 for i in rtrn_pred]
+    return bin_y, bin_pred
+
+def get_acc(bin_y,bin_pred):
+    # Getting accuracy
+    acc = skm.accuracy_score(bin_y,bin_pred)
+
+    # Accuracy if you randomly guessed
+    rndm_acc = sum(bin_y) / len(bin_y)
+    rndm_acc = max(1 - rndm_acc, rndm_acc)
+
+    return rndm_acc, acc
+
+# Output:
+#   true neg, false pos
+#   false neg, true pos
+def get_confusion_matrix(bin_y,bin_pred):
+    cfn_mtrx = skm.confusion_matrix(bin_y, bin_pred)
+    tp = cfn_mtrx[1][1]
+    fp = cfn_mtrx[0][1]
+    tn = cfn_mtrx[0][0]
+    fn = cfn_mtrx[1][0]
+    return cfn_mtrx
+
+def get_precision(cfn_mtrx):
+    precision = cfn_mtrx[1][1] / (cfn_mtrx[0][1] + cfn_mtrx[1][1])
+    return precision
+
+def print_metrics(test_x,test_y,model,label_scaler):
+    # Getting binary classification
+    bin_y, bin_pred = calculate_classification(test_x, test_y, model, label_scaler)
+
+    rndm_acc, acc = get_acc(bin_y, bin_pred)
+    cfn_mtrx = get_confusion_matrix(bin_y,bin_pred)
+    prec = get_precision(cfn_mtrx)
+
+    # Printing Accuracy
+    print("Random: " + str(rndm_acc))
+    print("Accuracy: " + str(acc))
+
+    # Printing Confusion Matrix
+    print("Confusion Matrix: ")
+    print(cfn_mtrx)
+
+    print("Precision: " + str(prec))