Python_code_part3.py

# Use drop method to drop the columns
X = df.drop(['Close', 'Signal', 'High',
             'Low', 'Volume', 'Ret'], axis=1)

# Create a variable which contains all the 'Signal' values
y = df['Signal']

# Test variables for 'c' and 'g'
c = [10, 100, 1000, 10000]
g = [1e-2, 1e-1, 1e0]

# Intialise the parameters
parameters = {'svc__C': c,
              'svc__gamma': g,
              'svc__kernel': ['rbf']
              }

# Create the 'steps' variable with the pipeline functions
steps = [('scaler', StandardScaler()), ('svc', SVC())]

# Pass the 'steps' to the Pipeline function
pipeline = Pipeline(steps)

# Call the RandomizedSearchCV function and pass the parameters
rcv = RandomizedSearchCV(pipeline, parameters, cv=TimeSeriesSplit(n_splits=2))

# Call the 'fit' method of rcv and pass the train data to it
rcv.fit(X.iloc[:split], y.iloc[:split])

# Call the 'best_params_' method to obtain the best parameters of C
best_C = rcv.best_params_['svc__C']

# Call the 'best_params_' method to obtain the best parameters of kernel
best_kernel = rcv.best_params_['svc__kernel']

# Call the 'best_params_' method to obtain the best parameters of gamma
best_gamma = rcv.best_params_['svc__gamma']

# Create a new SVC classifier
cls = SVC(C=best_C, kernel=best_kernel, gamma=best_gamma)

# Instantiate the StandardScaler
ss1 = StandardScaler()

# Pass the scaled train data to the SVC classifier
cls.fit(ss1.fit_transform(X.iloc[:split]), y.iloc[:split])

# Pass the test data to the predict function and store the values into 'y_predict'
y_predict = cls.predict(ss1.transform(X.iloc[split:]))

# Initiate a column by name, 'Pred_Signal' and assign 0 to it
df['Pred_Signal'] = 0

# Save the predicted values for the train data
df.iloc[:split, df.columns.get_loc('Pred_Signal')] = pd.Series(
    cls.predict(ss1.transform(X.iloc[:split])).tolist())

# Save the predicted values for the test data
df.iloc[split:, df.columns.get_loc('Pred_Signal')] = y_predict

# Calculate strategy returns and store them in 'Ret1' column
df['Ret1'] = df['Ret']*df['Pred_Signal']

# Calculate the confusion matrix
cm = confusion_matrix(y[split:], y_predict)
cm

# Calculate the classification report
cr = classification_report(y[split:], y_predict)
print(cr)

#declare figure
fig = go.Figure()

#Set up traces
fig.add_trace(go.Scatter(x=df.index, y= (df['Ret'][split:]+1).cumprod(),line=dict(color='royalblue', width=.8), name = 'stock_returns'))
fig.add_trace(go.Scatter(x=df.index, y= (df['Ret1'][split:]+1).cumprod(),line=dict(color='orange', width=.8), name = 'strategy_returns'))

# Add titles
fig.update_layout(
    title='Support Vector Machine Strategy',
    yaxis_title='Stock return (% Return)')

fig.show()