petur.print_evaluation(y_test, predictions)
df_predict = pd.DataFrame({'Predictions': predictions, 'True_label': y_test})
time_End = time.time()
print("Seconds to run:", time_End - time_Start)
# In[Genetic algorithm - Classification]
# Generate Genetic algorithm output == TEST DATA
#==============================================================================
time_Start = time.time()
train_data = df_train.copy()
test_data = df_test.copy()
features = petur.get_tickers()
features.append('OMXIPI')
# Train test split
y_train = train_data.Target
X_train = train_data[features]
y_test = test_data.Target
X_test = test_data[features]
tpot = TPOTClassifier(generations=50,
population_size=50,
verbosity=2,
scoring='f1',
n_jobs=-1,
cv=5)
benchmark_name = "Scores/BENCHMARK.csv"
df_benchmark = pd.read_csv(benchmark_name, header=0)
first_column = df_benchmark.columns[0]
df_benchmark = df_benchmark.set_index(first_column, drop=True)
df_benchmark.dropna(axis=1, inplace=True)
importance_dic = pickle.load(open("Pickles/importance_dic.p", "rb"))
#==============================================================================
#In[Feature Engineering]
#==============================================================================
df_fe = df_raw.copy()
# Get list of tickers
tickers = petur.get_tickers()
tickers.append('OMXIPI')
#tickers = ['OMXIPI','MARL', 'OSSRu'] #45% of market cap + index
# Create MAs
sma_days = [5, 10, 20, 50]
fibbonaci = [1, 2, 3, 5, 8, 13, 21]
both = [1, 2, 3, 5, 8, 10, 13, 20, 21, 30, 50]
df_fe = petur.make_sma(df_fe, tickers, sma_days)
df_fe = petur.make_ewma(df_fe, tickers, sma_days)
df_fe = df_fe.dropna() # Maybe do at end of labels
#==============================================================================
#In[Create labels]
#==============================================================================
import pandas as pd
import time
from tpot import TPOTClassifier
from sklearn.svm import SVC
import petur_functions as petur
# In[Load data and prepare]
# Load data from csv
df_name = "Working_Dataset/MAIN_FULL.csv"
df = pd.read_csv(df_name, index_col='Date', header=0)
# Define tickers to predict and days
tickers = petur.get_tickers()
tickers.append('OMXIPI')
#tickers = ['OMXIPI','MARL', 'OSSRu']
days = [1, 3, 5, 10]
sma_days = [5, 10, 20, 50]
fibbonaci = [1, 2, 3, 5, 8, 13, 21]
# Feature engineering and train
df = petur.make_sma(df, tickers, sma_days)
df = petur.make_ewma(df, tickers, sma_days)
df = df.dropna()
# Train/test split
df_train, df_test = petur.tr_ts_split(df)