def skill_score_murphy(predicted,reference):
'''
Calculate non-dimensional skill score (SS) between two variables using
definition of Murphy (1988)
Calculates the non-dimensional skill score (SS) difference between two
variables PREDICTED and REFERENCE. The skill score is calculated using
the formula:
SS = 1 - RMSE^2/SDEV^2
where RMSE is the root-mean-squre error between the predicted and
reference values
(RMSE)^2 = sum_(n=1)^N (p_n - r_n)^2/N
and SDEV is the standard deviation of the reference values
SDEV^2 = sum_(n=1)^N [r_n - mean(r)]^2/(N-1)
where p is the predicted values, r is the reference values, and
N is the total number of values in p & r. Note that p & r must
have the same number of values.
Input:
PREDICTED : predicted field
REFERENCE : reference field
Output:
SS : skill score
Reference:
Allan H. Murphy, 1988: Skill Scores Based on the Mean Square Error
and Their Relationships to the Correlation Coefficient. Mon. Wea.
Rev., 116, 2417-2424.
doi: http//dx.doi.org/10.1175/1520-0493(1988)<2417:SSBOTM>2.0.CO;2
Author: Peter A. Rochford
Symplectic, LLC
www.thesymplectic.com
[email protected]
Created on Dec 7, 2016
'''
utils.check_arrays(predicted, reference)
# Calculate RMSE
rmse2 = rmsd(predicted,reference)**2
# Calculate standard deviation
sdev2 = np.std(reference,ddof=1)**2
#% Calculate skill score
ss = 1 - rmse2/sdev2
return ss
def taylorPlot():
obsCol = df.iloc[:, iObs]
stds = []
rmses = []
coefs = []
labels = []
stds.append(obsCol.std())
rmses.append(0)
coefs.append(1)
labels.append('Fluxnet')
for i in range(colN):
if i != iObs:
simLabel = df.columns[i]
thisDF = df.dropna(subset=[simLabel, df.columns[iObs]])
simCol = thisDF[simLabel]
obsCol = thisDF[df.columns[iObs]]
if simCol.any():
std = simCol.std()
rmse = sm.rmsd(simCol, obsCol)
coef = np.corrcoef(simCol, obsCol)[0, 1]
stds.append(std)
rmses.append(rmse)
coefs.append(coef)
labels.append(simLabel)
else:
pass
else:
pass
sm.taylor_diagram(np.array(stds),
np.array(rmses),
np.array(coefs),
markerLabel=labels,
markerLabelColor='r',
markerSize=6,
markerLegend='on',
colOBS='g',
styleOBS='-',
markerobs='o',
showlabelsRMS='on',
titleRMS='on',
titleOBS='Fluxnet',
rmslabelformat=':.1f')
plt.title(metricName,
y=1.06,
fontsize='large',
loc='center',
horizontalalignment='center')
plt.savefig(argv['outputPath'])
plt.close('all')
def getStatisticalIndex():
result = {
'means': [],
'stds': [],
'rmses': [],
'coefs': [],
'nses': [],
'r2s': [],
'labels': []
}
for i in range(colN):
if i != iObs:
simLabel = df.columns[i]
thisDF = df.dropna(subset=[simLabel, df.columns[iObs]])
simCol = thisDF[simLabel]
obsCol = thisDF[df.columns[iObs]]
std = simCol.std()
mean = simCol.mean()
if simCol.any():
rmse = sm.rmsd(simCol, obsCol)
coef = np.corrcoef(simCol, obsCol)[0, 1]
nse = 1 - sum((simCol - obsCol)**2) / sum(
(obsCol - obsCol.mean())**2)
r2 = coef**2
else:
rmse = np.NaN
coef = np.NaN
nse = np.NaN
r2 = np.NaN
result['means'].append(mean)
result['labels'].append(simLabel)
result['stds'].append(std)
result['rmses'].append(rmse)
result['coefs'].append(coef)
result['nses'].append(nse)
result['r2s'].append(r2)
else:
result['means'].append(obsCol.mean())
result['labels'].append('Fluxnet')
result['stds'].append(obsCol.std())
result['rmses'].append(0)
result['coefs'].append(1)
result['nses'].append(0)
result['r2s'].append(1)
jsonStr = json.dumps(result).replace('NaN', 'null')
print(jsonStr)
# Calculate various skill metrics, writing results to screen
# and Excel file. Use an ordered dictionary so skill metrics are
# saved in the Excel file in the same order as written to screen.
stats = OrderedDict()
# Read data from pickle file
data = load_obj('target_data')
pred = data.pred1['data']
ref = data.ref['data']
# Get bias
stats['bias'] = sm.bias(pred, ref)
print('Bias = ' + str(stats['bias']))
# Get Root-Mean-Square-Deviation (RMSD)
stats['rmsd'] = sm.rmsd(pred, ref)
print('RMSD = ' + str(stats['rmsd']))
# Get Centered Root-Mean-Square-Deviation (CRMSD)
stats['crmsd'] = sm.centered_rms_dev(pred, ref)
print('CRMSD = ' + str(stats['crmsd']))
# Get Standard Deviation (SDEV)
stats['sdev'] = np.std(pred)
print('SDEV = ' + str(stats['sdev']))
# Get correlation coefficient (r)
ccoef = np.corrcoef(pred, ref)
stats['ccoef'] = ccoef[0, 1]
print('r = ' + str(stats['ccoef']))
observationData = observationData.reshape(observationData.size)
stds = []
rmses = []
coefs = []
stds.append(observationData.std())
rmses.append(0)
coefs.append(1)
for i, nc in enumerate(cmip.ncPaths):
if i == 0:
continue
data = cmip.getData(i, allTime=False, timeIndex=0)
data = data.reshape(data.size)
std = data.std()
rmse = sm.rmsd(data, observationData)
coef = np.ma.corrcoef(data, observationData)[0, 1]
stds.append(std)
rmses.append(rmse)
coefs.append(coef)
# intervalsCOR = np.concatenate((np.arange(0,1.0,0.2), [0.9, 0.95, 0.99, 1]))
sm.taylor_diagram(
np.array(stds),
np.array(rmses),
np.array(coefs),
markerLabel=cmip.markerLabels,
# tickRMS = np.arange(0,25,10),
# tickSTD = np.arange(9,20,5),
# tickCOR = intervalsCOR,
rmslabelformat=':.1f')
####### Plot Nov
plt.scatter(fechan, realn, color='black', label='Data') # datos iniciales
plt.plot(fechan, testn, color='red', label='Modelo RBF') # RBF kernel
plt.title('Tunal (Nov 22 y 23)')
#plt.plot(fechas,svr_lin.predict(fechas), color= 'green', label= 'Modelo Lineal') # lineal kernel
#plt.plot(fechas,svr_poly.predict(fechas), color= 'blue', label= 'Modelo Polinomial') # Polinomial kernel
plt.xlabel('Horas')
plt.ylabel('Concentracion de PM_25')
plt.legend()
plt.savefig('Tun_Nov_22-23.png')
plt.show()
plt.close()
print('------ Febrero 14 y 15 de 2019 ------')
print(np.corrcoef(real, test))
print(sm.rmsd(test, np.array(real)))
print(sm.bias(test, np.array(real)))
print('------ Noviembre 22 y 23 de 2018 ------')
print(np.corrcoef(realn, testn))
print(sm.rmsd(testn, np.array(realn)))
print(sm.bias(testn, np.array(realn)))
##### Create files
## Feb
soda = {'SVR_tun': test.tolist()}
df = pd.DataFrame(soda, columns=['SVR_tun'])
df.to_csv('tun_feb_svr.csv')
## Nov
soda = {'SVR_tun': test.tolist()}