def performFeatureSelection(maxlag):
import functions
import datetime
target = 'CLASSIFICATION'
lags = range(2, maxlag)
print 'Maximum time lag applied', max(lags)
print ''
for maxdelta in range(3,12):
#datasets = functions.loadDatasets('/home/francesco/Dropbox/DSR/StocksProject/longdatasets')
#start = datetime.datetime(1990, 1, 1)
#end = datetime.datetime(2014, 8, 31)
#out = functions.getStock('AAPL', start, end)
datasets = functions.loadDatasets('/home/francesco/Dropbox/DSR/StocksProject/longdatasets')
#datasets.insert(0, out)
delta = range(2,maxdelta)
print 'Delta days accounted: ', max(delta)
for dataset in datasets:
columns = dataset.columns
adjclose = columns[-2]
returns = columns[-1]
for n in delta:
functions.addFeatures(dataset, adjclose, returns, n)
#dataset = dataset.iloc[max(delta):,:]
finance = functions.mergeDataframes(datasets, 6, target)
#finance = finance.ix[max(delta):]
print 'Size of data frame: ', finance.shape
print 'Number of NaN after merging: ', functions.count_missing(finance)
finance = finance.interpolate(method='time')
print 'Number of NaN after time interpolation: ', functions.count_missing(finance)
finance = finance.fillna(finance.mean())
print 'Number of NaN after mean interpolation: ', functions.count_missing(finance)
back = -1
finance.Return_Out = finance.Return_Out.shift(back)
finance = functions.applyTimeLag(finance, lags, delta, back, target)
print 'Number of NaN after temporal shifting: ', functions.count_missing(finance)
print 'Size of data frame after feature creation: ', finance.shape
if target == 'CLASSIFICATION':
start_test = datetime.datetime(2014,4,1)
X_train, y_train, X_test, y_test = functions.prepareDataForClassification(finance, start_test)
acc = functions.performCV(X_train, y_train, 10, 'GTB', [])
print ''
print 'Mean Accuracy for (%d, %d): %f' % (max(lags), max(delta), acc)
#print functions.performClassification(X, y, X_val, y_val, 'ADA', [100, 1])
print '============================================================================'
def final():
target = "CLASSIFICATION"
lags = range(2, 3)
print "Maximum time lag applied", max(lags)
start = datetime.datetime(1990, 1, 1)
end = datetime.datetime(2014, 8, 31)
out = functions.getStock("GE", start, end)
datasets = functions.loadDatasets("/home/francesco/Dropbox/DSR/StocksProject/longdatasets")
datasets.insert(0, out)
delta = range(2, 5)
print "Max Delta days accounted: ", max(delta)
for dataset in datasets:
columns = dataset.columns
adjclose = columns[-2]
returns = columns[-1]
for n in delta:
functions.addFeatures(dataset, adjclose, returns, n)
# dataset = dataset.iloc[max(delta):,:]
finance = functions.mergeDataframes(datasets, 6, target)
# finance = finance.ix[max(delta):]
print "Size of data frame: ", finance.shape
print "Number of NaN after merging: ", functions.count_missing(finance)
print "% of NaN after merging: ", (
functions.count_missing(finance) / float(finance.shape[0] * finance.shape[1])
) * 100, "%"
finance = finance.interpolate(method="time")
print "Number of NaN after time interpolation: ", functions.count_missing(finance)
finance = finance.fillna(finance.mean())
print "Number of NaN after mean interpolation: ", functions.count_missing(finance)
back = -1
# finance.Return_SP500 = finance.Return_SP500.shift(back)
finance.Return_Out = finance.Return_Out.shift(back)
finance = functions.applyTimeLag(finance, lags, delta, back, target)
# finance = functions.mergeSentimenToStocks(finance)
# print finance.columns
print "Number of NaN after temporal shifting: ", functions.count_missing(finance)
print "Size of data frame after feature creation: ", finance.shape
if target == "CLASSIFICATION":
start_test = datetime.datetime(2014, 4, 1)
X_train, y_train, X_test, y_test = functions.prepareDataForClassification(finance, start_test)
print ""
# print 'Performing CV...'
# grid = {'n_estimators': [80, 100, 150], 'learning_rate': [0.01, 0.1, 1, 10]}
# grid = {'n_estimators': [50, 80, 100, 1000]}
# functions.performTimeSeriesSearchGrid(finance, 4, 0.8, features, 'ADA', grid)
print functions.performClassification(X_train, y_train, X_test, y_test, "RF", [])
path_manure_nonpts = 'GIS_data/final_data/manure_nonpts.csv' path_AD_pts = 'GIS_data/final_data/AD_pts.csv' path_COMB_pts = 'GIS_data/final_data/COMB_pts.csv' path_W2E_pts = 'GIS_data/final_data/W2E_pts.csv' path_DES_CBG_pts = 'GIS_data/final_data/DES_CBGcntrd.csv' path_PROC_ZC_pts = 'GIS_data/final_data/PROC_ZCcntrd.csv' path_msw_CBGcntrd_pts = 'GIS_data/final_data/DES_CBGcntrd.csv' path_crp2016_pts = 'GIS_data/final_data/DES_CBGcntrd.csv' path_crp2020_pts = 'GIS_data/final_data/DES_CBGcntrd.csv' path_crp2050_pts = 'GIS_data/final_data/DES_CBGcntrd.csv' #path_proc_pts = 'GIS_data/final_data/proc_pts.csv' path_proc_nonpts = 'GIS_data/final_data/proc_nonpts.csv' path_thermal = 'GIS_data/final_data/DES_CBGcntrd.csv' # Load all data as geodataframes counties_df = loadDatasets(path_counties) biomass_pts = loadDatasets(path_manure) AD_pts = loadDatasets(path_AD_pts) COMB_pts = loadDatasets(path_COMB_pts) W2E_pts = loadDatasets(path_W2E_pts) DES_CBG_pts = loadDatasets(path_DES_CBG_pts) PROC_ZC_pts = loadDatasets(path_PROC_ZC_pts) msw_CBGcntrd_pts = loadDatasets(path_msw_CBGcntrd_pts) crp2016_pts = loadDatasets(path_crp2016_pts) crp2020_pts = loadDatasets(path_crp2020_pts) crp2050_pts = loadDatasets(path_crp2050_pts) manure_nonpts = pd.read_csv(path_manure_nonpts) proc_nonpts = pd.read_csv(path_proc_nonpts) thermal = loadDatasets(path_thermal) # Set the preferred resulted columns for each dataset
def performFeatureSelection(maxlag):
import functions
import datetime
target = 'CLASSIFICATION'
lags = range(2, maxlag)
print 'Maximum time lag applied', max(lags)
print ''
for maxdelta in range(3, 12):
#datasets = functions.loadDatasets('/home/francesco/Dropbox/DSR/StocksProject/longdatasets')
#start = datetime.datetime(1990, 1, 1)
#end = datetime.datetime(2014, 8, 31)
#out = functions.getStock('AAPL', start, end)
datasets = functions.loadDatasets(
'/home/francesco/Dropbox/DSR/StocksProject/longdatasets')
#datasets.insert(0, out)
delta = range(2, maxdelta)
print 'Delta days accounted: ', max(delta)
for dataset in datasets:
columns = dataset.columns
adjclose = columns[-2]
returns = columns[-1]
for n in delta:
functions.addFeatures(dataset, adjclose, returns, n)
#dataset = dataset.iloc[max(delta):,:]
finance = functions.mergeDataframes(datasets, 6, target)
#finance = finance.ix[max(delta):]
print 'Size of data frame: ', finance.shape
print 'Number of NaN after merging: ', functions.count_missing(finance)
finance = finance.interpolate(method='time')
print 'Number of NaN after time interpolation: ', functions.count_missing(
finance)
finance = finance.fillna(finance.mean())
print 'Number of NaN after mean interpolation: ', functions.count_missing(
finance)
back = -1
finance.Return_Out = finance.Return_Out.shift(back)
finance = functions.applyTimeLag(finance, lags, delta, back, target)
print 'Number of NaN after temporal shifting: ', functions.count_missing(
finance)
print 'Size of data frame after feature creation: ', finance.shape
if target == 'CLASSIFICATION':
start_test = datetime.datetime(2014, 4, 1)
X_train, y_train, X_test, y_test = functions.prepareDataForClassification(
finance, start_test)
acc = functions.performCV(X_train, y_train, 10, 'GTB', [])
print ''
print 'Mean Accuracy for (%d, %d): %f' % (max(lags), max(delta),
acc)
#print functions.performClassification(X, y, X_val, y_val, 'ADA', [100, 1])
print '============================================================================'
def final():
target = 'CLASSIFICATION'
lags = range(2, 3)
print 'Maximum time lag applied', max(lags)
start = datetime.datetime(1990, 1, 1)
end = datetime.datetime(2014, 8, 31)
out = functions.getStock('GE', start, end)
datasets = functions.loadDatasets(
'/home/francesco/Dropbox/DSR/StocksProject/longdatasets')
datasets.insert(0, out)
delta = range(2, 5)
print 'Max Delta days accounted: ', max(delta)
for dataset in datasets:
columns = dataset.columns
adjclose = columns[-2]
returns = columns[-1]
for n in delta:
functions.addFeatures(dataset, adjclose, returns, n)
#dataset = dataset.iloc[max(delta):,:]
finance = functions.mergeDataframes(datasets, 6, target)
#finance = finance.ix[max(delta):]
print 'Size of data frame: ', finance.shape
print 'Number of NaN after merging: ', functions.count_missing(finance)
print '% of NaN after merging: ', (
functions.count_missing(finance) /
float(finance.shape[0] * finance.shape[1])) * 100, '%'
finance = finance.interpolate(method='time')
print 'Number of NaN after time interpolation: ', functions.count_missing(
finance)
finance = finance.fillna(finance.mean())
print 'Number of NaN after mean interpolation: ', functions.count_missing(
finance)
back = -1
#finance.Return_SP500 = finance.Return_SP500.shift(back)
finance.Return_Out = finance.Return_Out.shift(back)
finance = functions.applyTimeLag(finance, lags, delta, back, target)
#finance = functions.mergeSentimenToStocks(finance)
#print finance.columns
print 'Number of NaN after temporal shifting: ', functions.count_missing(
finance)
print 'Size of data frame after feature creation: ', finance.shape
if target == 'CLASSIFICATION':
start_test = datetime.datetime(2014, 4, 1)
X_train, y_train, X_test, y_test = functions.prepareDataForClassification(
finance, start_test)
print ''
#print 'Performing CV...'
#grid = {'n_estimators': [80, 100, 150], 'learning_rate': [0.01, 0.1, 1, 10]}
#grid = {'n_estimators': [50, 80, 100, 1000]}
#functions.performTimeSeriesSearchGrid(finance, 4, 0.8, features, 'ADA', grid)
print functions.performClassification(X_train, y_train, X_test, y_test,
'RF', [])
