def plot_sma_forecast():
#get dates to be shown on graph
df = data.get_data()
dates = list(df['date'])
date_show = [x for x in range(len(dates)) if x%(365//2) == 0]
date = []
for x in date_show:
date.append(dates[x])
#plot the graph
data_forplot = data.get_hlc_average()
style.use('seaborn-whitegrid')
fig = plt.figure()
plt.plot(data_forplot, label = "HLC average")
# add the forecasted data, remove the comments below to compare the sma forecast
#to the Double Exponential Smoothing and Weighted Moving Average
plt.plot(algo.double_exponential_smoothing(), label = 'Double Exponential Smoothing Forecast')
plt.plot(algo.weighted_moving_average(), label = 'Weighted Moving Average')
plt.plot(algo.simple_moving_average(), label = 'Simple Moving Average' )
plt.xlabel('time')
plt.xticks(date_show, date, rotation = 50)
plt.yticks([0,50,100,150,200], ['0 PHP', '50 PHP', '100 PHP', \
'150 PHP', '200 PHP'])
plt.ylabel('high,low, and close average')
plt.title('Stock Prices of JFC in PSE ')
plt.legend()
plt.show()
def simple_moving_average():
"""
Simple Moving Average
a forecasting technique where it takes the last n actual values
then takes the average of it as the forecast. This is not a good
forecasting technique, this shall only be used for comparing to the
other forecasting techniques
"""
#take values from the data module
actual_data = list(data.get_hlc_average())
sma_forecast = []
#simple moving average
for x in range(len(actual_data)):
if x<3:
sma_forecast.append(actual_data[x])
else:
forecast = (actual_data[x - 3] + actual_data[x - 2] + actual_data[x - 1])/3
sma_forecast.append(forecast)
#test accuracy here
errors, total_error = test.test_accuracy(actual_data, sma_forecast)
print('the error percentage of the simple moving average forecast is: '
, total_error)
return sma_forecast
def weighted_moving_average():
"""
Weighted Moving Average
an improved version of the simple moving average,
the Weighted Moving Average makes forecast based on
previous actual data, depending on how far a particular
entry's time is from the time of the forecast
"""
#take values from the data module
actual_data = list(data.get_hlc_average())
wma_forecast = []
#weighted moving average
for x in range(len(actual_data)):
if x<4:
#first four elements are going to take in actual data
wma_forecast.append(actual_data[x])
else:
#this is where the forecast begins
wma_forecast.append(
((actual_data[x - 4] * 0.4) + (actual_data[x - 3] * 0.3) +
(actual_data[x - 2] * 0.2) + (actual_data[x - 1] * 0.1))
)
#test accuracy here
errors, total_error = test.test_accuracy(actual_data, wma_forecast)
print('the error percentage of the weighted moving average forecast is: '
, total_error)
return wma_forecast
def plot_hlc_average():
data_forplot = data.get_hlc_average()
#data_forplot = data.get_ohlc() - should only use if
# you can do OHLC candlestick
style.use('seaborn-whitegrid')
fig=plt.figure()
plt.plot( data_forplot ,label = "HLC average")
plt.xlabel('hlc average')
plt.ylabel('')
plt.title('Stock Prices of JFC in PSE ')
plt.legend()
plt.show()
def double_exponential_smoothing():
"""
Double Exponential Smoothing
using the exponential smoothing with
respect to the trend in the historical
data
"""
#take values from the data module
actual_data = list(data.get_hlc_average())
#set the alpha and beta values
alpha, beta = 0.35, 0.35
#initialization of lists
forecast_data = [actual_data[0],]
estimated_trend = [0]
trend_adjusted_forecast = [actual_data[0]]
for x in range(len(actual_data)):
if x == 0 :
pass
else:
#calculating level/trend
tt = (beta * (actual_data[x - 1] - forecast_data[x - 1])) + \
((1 - beta) * estimated_trend[x - 1] )
estimated_trend.append(tt)
#calculating forecast
ft = (alpha * actual_data[x - 1]) + ((1 - alpha) * \
(forecast_data[x - 1] + estimated_trend[x - 1]))
forecast_data.append(ft)
#adjusting the forecast with the trend
aft = ft + tt
trend_adjusted_forecast.append(aft)
#test accuracy here
errors, total_error = test.test_accuracy(actual_data, trend_adjusted_forecast)
print('the error percentage of the second-order exponential smoothing '
'forecast is: ', total_error)
return trend_adjusted_forecast