Created
July 13, 2019 17:26
-
-
Save ikanez/7a11cb83636443a7906d6ca4efd6f6d4 to your computer and use it in GitHub Desktop.
Find best parameter for sarimax model and log useful stuff into MLflow
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
# Initial approximation of parameters | |
Qs = range(0, 2) | |
qs = range(0, 3) | |
Ps = range(0, 3) | |
ps = range(0, 3) | |
D=1 | |
d=1 | |
parameters = product(ps, qs, Ps, Qs) | |
parameters_list = list(parameters) | |
len(parameters_list) | |
# Model Selection | |
results = [] | |
best_aic = float("inf") | |
warnings.filterwarnings('ignore') | |
for param in parameters_list: | |
# start mlflow run | |
with mlflow.start_run(run_name='sarimax_param'): | |
# log parameters | |
mlflow.log_param('order-Qs', param[0]) | |
mlflow.log_param('order-qs', param[1]) | |
mlflow.log_param('seasonal-order-Ps', param[2]) | |
mlflow.log_param('seasonal-order-ps', param[3]) | |
try: | |
# model = SARIMAX(btc_month.close_box, order=(param[0], d, param[1]), seasonal_order=(param[2], D, param[3], 12)).fit(disp=-1) | |
model = SARIMAX(btc_month.close_box, order=(param[0], d, param[1]), seasonal_order=(param[2], D, param[3], 4)).fit(disp=-1) | |
except ValueError: | |
print('bad parameter combination:', param) | |
continue | |
aic = model.aic | |
if aic < best_aic: | |
best_model = model | |
best_aic = aic | |
best_param = param | |
results.append([param, model.aic]) | |
# log metric | |
mlflow.log_metric('aic',aic) | |
mlflow.log_metric('dickey-fuller-test',adfuller(model.resid[13:])[1]) | |
# log artifact: model summary | |
plt.rc('figure', figsize=(12, 7)) | |
plt.text(0.01, 0.05, str(model.summary()), {'fontsize': 10}, fontproperties = 'monospace') | |
plt.axis('off') | |
plt.tight_layout() | |
summary_fn = 'model_sarimax_summary_{}_{}_{}_{}.png'.format(param[0],param[1],param[2],param[3]) | |
plt.savefig(summary_fn) | |
mlflow.log_artifact(summary_fn) # logging to mlflow | |
plt.close() | |
# log artifact: diagnostics plot | |
model.plot_diagnostics(figsize=(15, 12)) | |
fig1_fn = 'figure_diagnostics_{}_{}_{}_{}.png'.format(param[0],param[1],param[2],param[3]) | |
plt.savefig(fig1_fn) | |
mlflow.log_artifact(fig1_fn) # logging to mlflow | |
plt.close() | |
# log artifact: residuals and pacf plot | |
plt.subplot(211) | |
model.resid[13:].plot() | |
plt.ylabel(u'Residuals') | |
ax = plt.subplot(212) | |
plot_acf(model.resid[13:].values.squeeze(), lags=12, ax=ax) | |
plt.tight_layout() | |
fig2_fn = 'figure_res_pacf_{}_{}_{}_{}.png'.format(param[0],param[1],param[2],param[3]) | |
plt.savefig(fig2_fn) | |
mlflow.log_artifact(fig2_fn) # logging to mlflow | |
plt.close() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
product isn't defined