def main():
yr = 2015
train_years = range(1995, 2015)
clf, mean_norm, stdev_norm, sind, tind, train_labels, train_features = cpt.fit_model(train_years, False)
train_measured = train_labels
train_predicted = clf.predict(train_features)
bfit = cpt.calc_bias_function(train_measured, train_predicted)
# Predict crab population for left out year
sdata, hdata, wdata, seddata, savdata, ldata = cpt.get_data()
features, featureNames = cpt.get_features(-1, sdata, hdata, wdata, seddata, savdata, ldata, [yr])
features = features[:, sind]
featureNames = featureNames[sind]
features = (features - mean_norm) / stdev_norm
features = features[:, tind]
featureNames = featureNames[tind]
# Predict values for next year from that data
predicted_t_per_area_corr = clf.predict(features)
# Save individual results
predicted = predicted_t_per_area_corr
predicted = cpt.apply_bias_correction(predicted, bfit)
# Calculate total number of crabs in the bay
area_bay = 9812.0 # per square kilometer
area_bay = area_bay*((1e3**2)) # per square meter
total_crabs = np.mean(predicted)*area_bay
fn = 'pickle-predictCrabs2015.pickle'
with open(fn, 'wb') as f:
pickle.dump([predicted, total_crabs, clf, features, featureNames, sind, tind, mean_norm, stdev_norm, bfit], f)
def main():
yr = 2016
# Get model that you trained for 2015
fn = 'pickle-predictCrabs2015.pickle'
with open(fn, 'rb') as f:
predicted, total_crabs, clf, features, featureNames, sind, tind, mean_norm, stdev_norm, bfit = pickle.load(f)
# Get features
sdata, hdata, wdata, seddata, savdata, ldata = cpt.get_data()
features, featureNames = cpt.get_features(-1, sdata, hdata, wdata, seddata, savdata, ldata, [2016])
features = features[:, sind]
featureNames = featureNames[sind]
features = (features - mean_norm) / stdev_norm
features = features[:, tind]
featureNames = featureNames[tind]
# Predict values for next year from that data
predicted_t_per_area_corr = clf.predict(features)
# Save individual results
predicted = predicted_t_per_area_corr
predicted = cpt.apply_bias_correction(predicted, bfit)
# Calculate total number of crabs in the bay
area_bay = 9812.0 # per square kilometer
area_bay = area_bay*((1e3**2)) # per square meter
total_crabs = np.mean(predicted)*area_bay
fn = 'pickle-predictCrabs2016.pickle'
with open(fn, 'wb') as f:
pickle.dump([predicted, total_crabs, features, featureNames], f)
def main():
#
# Get total number of crabs per year in bay from Dredge survey
#
db = MySQLdb.connect(host = "localhost", user="******", passwd="crabby", db="crabs")
ddata = pd.read_sql('SELECT Year,Total_Crabs FROM DREDGE;', con=db)
db.close()
ddata['Predict'] = -1
for yr in range(1995, 2015):
train_years = range(1991, 2015)
train_years = [yy for yy in train_years if yy < yr]
#train_years = range(1995, 2015)
#train_years.remove(yr)
clf, mean_norm, stdev_norm, sind, tind, train_labels, train_features = cpt.fit_model(train_years, False)
# Get bias correction
train_measured = train_labels
train_predicted = clf.predict(train_features)
bfit = cpt.calc_bias_function(train_measured, train_predicted)
# Predict crab population for left out year
sdata, hdata, wdata, seddata, savdata, ldata = cpt.get_data()
# Get features for this year
labels, label_data = cpt.get_labels(sdata, [yr])
features, featureNames = cpt.get_features(label_data, sdata, hdata, wdata, seddata, savdata, ldata, [yr])
features = features[:, sind]
featureNames = featureNames[sind]
features = (features - mean_norm) / stdev_norm
features = features[:, tind]
featureNames = featureNames[tind]
# Predict values for next year from that data
predicted_t_per_area_corr = clf.predict(features)
# Save individual results
measured = labels
predicted = predicted_t_per_area_corr
predicted = cpt.apply_bias_correction(predicted, bfit)
# Calculate total number of crabs in the bay
area_bay = 9812.0 # per square kilometer
area_bay = area_bay*((1e3**2)) # per square meter
total_crabs = np.mean(predicted)*area_bay
ddata.loc[ddata.Year == yr, 'Predict'] = total_crabs
print(str(yr)+' '+str(ddata[ddata.Year == yr].Total_Crabs)+' '+str(total_crabs))
fn = 'pickle-predictCrabs_'+str(yr)+'.pickle'
with open(fn, 'wb') as f:
pickle.dump([label_data, measured, predicted, clf, features, featureNames], f)
fn = 'pickle-predictCrabs.pickle'
with open(fn, 'wb') as f:
pickle.dump([ddata], f)
