Forecast a single time series with a univariate model

Create a dataset

Create a BigQuery dataset to store your ML model:

1. In the Google Cloud console, go to the BigQuery page.

   Go to the BigQuery page

2. In the Explorer pane, click your project name.

3. Click more_vert View actions > Create dataset.

   

4. On the Create dataset page, do the following:

   • For Dataset ID, enter bqml_tutorial.

   • For Location type, select Multi-region, and then select US (multiple regions in United States).

     The public datasets are stored in the US multi-region. For simplicity, store your dataset in the same location.

   • Leave the remaining default settings as they are, and click Create dataset.

     

Visualize the input data

Before creating the model, you can optionally visualize your input time series data to get a sense of the distribution. You can do this by using Looker Studio.

Follow these steps to visualize the time series data:

import bigframes.pandas as bpd

# Start by loading the historical data from BigQuerythat you want to analyze and forecast.
# This clause indicates that you are querying the ga_sessions_* tables in the google_analytics_sample dataset.
# Read and visualize the time series you want to forecast.
df = bpd.read_gbq("bigquery-public-data.google_analytics_sample.ga_sessions_*")
parsed_date = bpd.to_datetime(df.date, format="%Y%m%d", utc=True)
visits = df["totals"].struct.field("visits")
total_visits = visits.groupby(parsed_date).sum()

# Expected output: total_visits.head()
# date
# 2016-08-01 00:00:00+00:00    1711
# 2016-08-02 00:00:00+00:00    2140
# 2016-08-03 00:00:00+00:00    2890
# 2016-08-04 00:00:00+00:00    3161
# 2016-08-05 00:00:00+00:00    2702
# Name: visits, dtype: Int64

total_visits.plot.line()

Create the time series model

Create a time series model to forecast total site visits as represented by totals.visits column, and train it on the Google Analytics 360 data.

In the following query, the OPTIONS(model_type='ARIMA_PLUS', time_series_timestamp_col='date', ...) clause indicates that you are creating an ARIMA-based time series model. The auto_arima option of the CREATE MODEL statement defaults to TRUE, so the auto.ARIMA algorithm automatically tunes the hyperparameters in the model. The algorithm fits dozens of candidate models and chooses the best model, which is the model with the lowest Akaike information criterion (AIC). The data_frequency option of the CREATE MODEL statements defaults to AUTO_FREQUENCY, so the training process automatically infers the data frequency of the input time series. The decompose_time_series option of the CREATE MODEL statement defaults to TRUE, so that information about the time series data is returned when you evaluate the model in the next step.

Follow these steps to create the model:

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, paste in the following query and click Run:

   CREATE OR REPLACE MODEL `bqml_tutorial.ga_arima_model`
   OPTIONS
   (model_type = 'ARIMA_PLUS',
     time_series_timestamp_col = 'parsed_date',
     time_series_data_col = 'total_visits',
     auto_arima = TRUE,
     data_frequency = 'AUTO_FREQUENCY',
     decompose_time_series = TRUE
   ) AS
   SELECT
   PARSE_TIMESTAMP("%Y%m%d", date) AS parsed_date,
   SUM(totals.visits) AS total_visits
   FROM
   `bigquery-public-data.google_analytics_sample.ga_sessions_*`
   GROUP BY date;

   The query takes about 4 seconds to complete, after which the ga_arima_model model appears in the Explorer pane. Because the query uses a CREATE MODEL statement to create a model, you don't see query results.

Evaluate the candidate models

Evaluate the time series models by using the ML.ARIMA_EVALUATE function. The ML.ARIMA_EVALUATE function shows you the evaluation metrics of all the candidate models evaluated during the process of automatic hyperparameter tuning.

Follow these steps to evaluate the model:

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, paste in the following query and click Run:

   SELECT
    *
   FROM
    ML.ARIMA_EVALUATE(MODEL `bqml_tutorial.ga_arima_model`);

   The results should look similar to the following:

   

   The non_seasonal_p, non_seasonal_d, non_seasonal_q, and has_drift output columns define an ARIMA model in the training pipeline. The log_likelihood, AIC, and varianceoutput columns are relevant to the ARIMA model fitting process.

   The auto.ARIMA algorithm uses the KPSS test to determine the best value for non_seasonal_d, which in this case is 1. When non_seasonal_d is 1, the auto.ARIMA algorithm trains 42 different candidate ARIMA models in parallel. In this example, all 42 candidate models are valid, so the output contains 42 rows, one for each candidate ARIMA model; in cases where some of the models aren't valid, they are excluded from the output. These candidate models are returned in ascending order by AIC. The model in the first row has the lowest AIC, and is considered the best model. The best model is saved as the final model and is used when you call functions such as ML.FORECAST on the model

   The seasonal_periods column contains information about the seasonal pattern identified in the time series data. It has nothing to do with the ARIMA modeling, therefore it has the same value across all output rows. It reports a weekly pattern, which agrees with the results you saw if you chose to visualize the input data.

   The has_holiday_effect, has_spikes_and_dips, and has_step_changes columns are only populated when decompose_time_series=TRUE. These columns also reflect information about the input time series data, and are not related to the ARIMA modeling. These columns also have the same values across all output rows.

   The error_message column shows any errors that incurred during the auto.ARIMA fitting process. One possible reason for errors is when the selected non_seasonal_p, non_seasonal_d, non_seasonal_q, and has_drift columns are not able to stabilize the time series. To retrieve the error message of all the candidate models, set the show_all_candidate_models option to TRUE when you create the model.

   For more information about the output columns, see ML.ARIMA_EVALUATE function.

Inspect the model's coefficients

Inspect the time series model's coefficients by using the ML.ARIMA_COEFFICIENTS function.

Follow these steps to retrieve the model's coefficients:

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, paste in the following query and click Run:

   SELECT
    *
   FROM
    ML.ARIMA_COEFFICIENTS(MODEL `bqml_tutorial.ga_arima_model`);

   The results should look similar to the following:

   

   The ar_coefficients output column shows the model coefficients of the autoregressive (AR) part of the ARIMA model. Similarly, the ma_coefficients output column shows the model coefficients of the moving-average (MA) part of the ARIMA model. Both of these columns contain array values, whose lengths are equal to non_seasonal_p and non_seasonal_q, respectively. You saw in the output of the ML.ARIMA_EVALUATE function that the best model has a non_seasonal_p value of 2 and a non_seasonal_q value of 3. Therefore, in the ML.ARIMA_COEFFICIENTS output, the ar_coefficients value is a 2-element array and the ma_coefficients value is a 3-element array. The intercept_or_drift value is the constant term in the ARIMA model.

   For more information about the output columns, see ML.ARIMA_COEFFICIENTS function.

Use the model to forecast data

Forecast future time series values by using the ML.FORECAST function.

In the following GoogleSQL query, the STRUCT(30 AS horizon, 0.8 AS confidence_level) clause indicates that the query forecasts 30 future time points, and generates a prediction interval with a 80% confidence level.

Follow these steps to forecast data with the model:

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, paste in the following query and click Run:

   SELECT
    *
   FROM
    ML.FORECAST(MODEL `bqml_tutorial.ga_arima_model`,
                STRUCT(30 AS horizon, 0.8 AS confidence_level));

   The results should look similar to the following:

   

   The output rows are in chronological order by the forecast_timestamp column value. In time series forecasting, the prediction interval, as represented by the prediction_interval_lower_bound and prediction_interval_upper_bound column values, is as important as the forecast_value column value. The forecast_value value is the middle point of the prediction interval. The prediction interval depends on the standard_error and confidence_level column values.

   For more information about the output columns, see ML.FORECAST function.

Explain the forecasting results

You can get explainability metrics in addition to forecast data by using the ML.EXPLAIN_FORECAST function. The ML.EXPLAIN_FORECAST function forecasts future time series values and also returns all the separate components of the time series.

Similar to the ML.FORECAST function, the STRUCT(30 AS horizon, 0.8 AS confidence_level) clause used in the ML.EXPLAIN_FORECAST function indicates that the query forecasts 30 future time points and generates a prediction interval with 80% confidence.

Follow these steps to explain the model's results:

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, paste in the following query and click Run:

   SELECT
    *
   FROM
    ML.EXPLAIN_FORECAST(MODEL `bqml_tutorial.ga_arima_model`,
      STRUCT(30 AS horizon, 0.8 AS confidence_level));

   The results should look similar to the following:

   

   The output rows are ordered chronologically by the time_series_timestamp column value.

   For more information about the output columns, see ML.EXPLAIN_FORECAST function.

   If you would like to visualize the results, you can use Looker Studio as described in the Visualize the input data section to create a chart, using the following columns as metrics:

   • time_series_data
   • prediction_interval_lower_bound
   • prediction_interval_upper_bound
   • trend
   • seasonal_period_weekly
   • step_changes