quantra-go-algo

def optimal_d(DF, alpha=0.035, minimum = 0, maximum = 1):
    # Copy the dataframe
    df = DF.copy()
    # Estimate the best d based on the basic d range
    out = d_estimates_db(df,minimum,maximum)
    out.to_excel('out1.xlsx')
    # A try-except block to handle erros while getting a d with better decimals
    try:
        # Get the d value which critical value is closest to and higher than alpha
        d1 = out[out['pVal']>alpha].index.values[-1]

quantra-go-algo

# Get d estimation summary
out = d_estimates_db(data,0,1)

quantra-go-algo

def d_estimates_db(DF,minimum,maximum):
    # Create the output dataframe
    out=pd.DataFrame(columns=['adfStat','pVal','lags',\
                              'nObs','95% conf','corr'])
    # Copy the dataframe in order to not change it
    df0=DF.copy()
    # Create a range of posible d values and loop throughout them
    for d in np.linspace(minimum,maximum,11):
        # Create the log of close prices
        df1=np.log(df0['Close'])

quantra-go-algo

def fracDiff_FFD(series,d,thres=1e-5, w = None):
    """
    Constant width window (new solution)
    Note 1: thres determines the cut-off weight for the window
    Note 2: d can be any positive fractional, not necessarily bounded [0,1].
    """
    # 1) Compute weights for the longest series
    # Get the length of the series
    length = len(series)
    # In case we don't have the weights, estimate them with the previous function

quantra-go-algo

# Function to get the ARFIMA weights
def getWeights_FFD(d, thres, lim):
    # Set w as a list and k as one
    w, k = [1.], 1
    ctr = 0
    while True:
        # Create the new weight value
        w_ = -w[-1] / k * (d - k + 1)
        # End the loop in case the threshold is breached
        if abs(w_) < thres:

quantra-go-algo

# Get the classification report
print(classification_report(y_test, y_pred))

quantra-go-algo

%%time
#Create an object of XGBClassifier with the below input parameters
model1 = XGBClassifier(booster='gbtree', num_leaves = 80, max_depth=5, learning_rate = 0.01, n_estimators=200)
model1.fit(x_train, y_train)

quantra-go-algo

%%time
#Create an object of LGBMClassifier with the below input parameters
model = lgb.LGBMClassifier(num_leaves = 80, max_depth=5, learning_rate = 0.01, n_estimators=200)
model.fit(x_train, y_train)

quantra-go-algo

#Plot confusion matrix for training and testing data
from sklearn.metrics import confusion_matrix, precision_score, accuracy_score
plot_confusion_matrix(model, x_train, y_train)
plt.title("Confusion matrix of training data")
plot_confusion_matrix(model, x_test, y_test)  
plt.title("Confusion matrix of testing data")

quantra-go-algo

#To obtain the training and testing accuracy
print('Training accuracy: {0:0.2f}'.format(model.score(x_train,y_train)))
print('Testing accuracy: {0:0.2f}'.format(model.score(x_test,y_test)))