Partial example of utilizing adaptive filter as real-time model prediction

nlms_stock_production.py

s_future = ts['Close'].values.flatten()[1000: ] 
filter_min_error = pa_list[np.argmin(error_list)]
print(f'Selected the filter with mu of {filter_min_error.mu}')
print(f'with avg eror of {error_list[np.argmin(error_list)]}') 
x_future = pa.input_from_history(s_future, n)
d_future = np.zeros(len(x_future))
N_future = len(x_future)

for i, k in enumerate(range((n-1), N_future)):
    d_future[i] = s_future[k+1]

actual_list = np.zeros(N_future)
predict_list = np.zeros(N_future)

# Simulate real-time prediction and update the filter
for k in range(N_future):
    x_input = x_future[k]
    y = filter_min_error.predict(x_input)
    d_actual = d_future[k]
    filter_min_error.adapt(d_actual, x_input)
    actual_list[k] = d_actual
    predict_list[k] = y
    
## show results
avg_error = np.round(np.mean(10*np.log10((actual_list -predict_list)**2)), 4)
plt.figure(figsize=(15,9))
plt.subplot(211)
plt.title("Adaptive Filter on future data (simulated)")
plt.xlabel("Samples [k]")
plt.plot(actual_list[n: (len(actual_list)-n)], "b", label="actual")
plt.plot(predict_list[n: (len(predict_list)-n)], "g", label="predict")
plt.legend()
plt.subplot(212)
plt.title(f'Filter error (avg: {avg_error})')
plt.xlabel("Samples [k]")
plt.plot(10*np.log10((actual_list[n: (len(actual_list)-n)] - \
                      predict_list[n: (len(predict_list)-n)])**2), 
         "r", label="Error")
plt.legend()
plt.tight_layout()
plt.show()