Skip to content

Fix auto-select on model failure #998

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Nov 11, 2024
Merged

Fix auto-select on model failure #998

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
c1731e6
Fix auto-select on model failure
prasankh Nov 8, 2024
a010890
Merge branch 'main' into fix/ODSC-63260
prasankh Nov 8, 2024
31295b5
Merge branch 'main' into fix/ODSC-63260
ahosler Nov 11, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
45 changes: 25 additions & 20 deletions ads/opctl/operator/lowcode/forecast/model_evaluator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -121,23 +121,26 @@ def run_all_models(self, datasets: ForecastDatasets, operator_config: ForecastOp
from .model.factory import ForecastOperatorModelFactory
metrics[model] = {}
for i in range(len(cut_offs)):
backtest_historical_data = train_sets[i]
backtest_additional_data = additional_data[i]
backtest_test_data = test_sets[i]
backtest_operator_config = self.create_operator_config(operator_config, i, model,
backtest_historical_data,
backtest_additional_data,
backtest_test_data)
datasets = ForecastDatasets(backtest_operator_config)
ForecastOperatorModelFactory.get_model(
backtest_operator_config, datasets
).generate_report()
test_metrics_filename = backtest_operator_config.spec.test_metrics_filename
metrics_df = pd.read_csv(
f"{backtest_operator_config.spec.output_directory.url}/{test_metrics_filename}")
metrics_df["average_across_series"] = metrics_df.drop('metrics', axis=1).mean(axis=1)
metrics_average_dict = dict(zip(metrics_df['metrics'].str.lower(), metrics_df['average_across_series']))
metrics[model][i] = metrics_average_dict[operator_config.spec.metric]
try:
backtest_historical_data = train_sets[i]
backtest_additional_data = additional_data[i]
backtest_test_data = test_sets[i]
backtest_operator_config = self.create_operator_config(operator_config, i, model,
backtest_historical_data,
backtest_additional_data,
backtest_test_data)
datasets = ForecastDatasets(backtest_operator_config)
ForecastOperatorModelFactory.get_model(
backtest_operator_config, datasets
).generate_report()
test_metrics_filename = backtest_operator_config.spec.test_metrics_filename
metrics_df = pd.read_csv(
f"{backtest_operator_config.spec.output_directory.url}/{test_metrics_filename}")
metrics_df["average_across_series"] = metrics_df.drop('metrics', axis=1).mean(axis=1)
metrics_average_dict = dict(zip(metrics_df['metrics'].str.lower(), metrics_df['average_across_series']))
metrics[model][i] = metrics_average_dict[operator_config.spec.metric]
except:
logger.warn(f"Failed to calculate metrics for {model} and {i} backtest")
return metrics

def find_best_model(self, datasets: ForecastDatasets, operator_config: ForecastOperatorConfig):
Expand All @@ -147,10 +150,12 @@ def find_best_model(self, datasets: ForecastDatasets, operator_config: ForecastO
model = SupportedModels.Prophet
logger.error(f"Running {model} model as auto-select failed with the following error: {e.message}")
return model
avg_backtests_metrics = {key: sum(value.values()) / len(value.values()) for key, value in metrics.items()}
best_model = min(avg_backtests_metrics, key=avg_backtests_metrics.get)
nonempty_metrics = {model: metric for model, metric in metrics.items() if metric != {}}
avg_backtests_metric = {model: sum(value.values()) / len(value.values())
for model, value in nonempty_metrics.items()}
best_model = min(avg_backtests_metric, key=avg_backtests_metric.get)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What if the metric is set to accuracy "r-squared"? Do we prevent this?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The r-squared metric is not present in forecast/schema.yaml file as an allowed optimization metric, so it is not used for selecting the best model. Although it gets calculated when test data is provided only for reporting purpose.

logger.info(f"Among models {self.models}, {best_model} model shows better performance during backtesting.")
backtest_stats = pd.DataFrame(metrics).rename_axis('backtest')
backtest_stats = pd.DataFrame(nonempty_metrics).rename_axis('backtest')
backtest_stats.reset_index(inplace=True)
output_dir = operator_config.spec.output_directory.url
backtest_report_name = "backtest_stats.csv"
Expand Down
7 changes: 3 additions & 4 deletions ads/opctl/operator/lowcode/forecast/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,10 +19,9 @@
mean_absolute_percentage_error,
mean_squared_error,
)
try:
from scipy.stats import linregress
except:
from sklearn.metrics import r2_score

from scipy.stats import linregress
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is scipy in the pyproject.toml?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not explicitly, statsmodels is there and it is dependent on scipy.

from sklearn.metrics import r2_score

from ads.common.object_storage_details import ObjectStorageDetails
from ads.dataset.label_encoder import DataFrameLabelEncoder
Expand Down