-
Notifications
You must be signed in to change notification settings - Fork 1
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request #27 from fredmontet/22-detailed-plot-of-anomalies
22 detailed plot of anomalies
- Loading branch information
Showing
12 changed files
with
1,143 additions
and
701 deletions.
There are no files selected for viewing
Large diffs are not rendered by default.
Oops, something went wrong.
Large diffs are not rendered by default.
Oops, something went wrong.
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
210 changes: 210 additions & 0 deletions
210
notebooks/docs/0_core/0.7-anomaly-time-series-plot.ipynb
Large diffs are not rendered by default.
Oops, something went wrong.
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
Large diffs are not rendered by default.
Oops, something went wrong.
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
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1 +1,2 @@ | ||
| from . import plots | ||
| from . import anomaly_plot |
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,212 @@ | ||
| import pandas as pd | ||
|
|
||
| from ..time_series import TimeSeries, BinaryTimeSeries | ||
| import altair as alt | ||
|
|
||
|
|
||
| class AnomalyPlot: | ||
| _DATA_COLUMN_NAME = "data" | ||
|
|
||
| @staticmethod | ||
| def plot_anomalies( | ||
| data: TimeSeries, | ||
| point_anomalies: BinaryTimeSeries | None = None, | ||
| contextual_anomalies: BinaryTimeSeries | None = None, | ||
| collective_anomalies: BinaryTimeSeries | None = None, | ||
| seasonal_anomalies: BinaryTimeSeries | None = None, | ||
| cyclical_anomalies: BinaryTimeSeries | None = None, | ||
| ): | ||
| """ | ||
| Plot the anomalies of the given time series. | ||
| each kind of anomaly is plotted in a different subplot. | ||
| :param data: TimeSeries data | ||
| :param point_anomalies: BinaryTimeSeries of point anomalies. Will be plotted as points on the graph. | ||
| :param contextual_anomalies: BinaryTimeSeries of contextual anomalies. Will be plotted as a line on the graph. | ||
| :param collective_anomalies: BinaryTimeSeries of collective anomalies. Will be plotted as an area on the graph. | ||
| :param seasonal_anomalies: BinaryTimeSeries of seasonal anomalies. Will be plotted as a line on the graph. | ||
| :param cyclical_anomalies: BinaryTimeSeries of cyclical anomalies. Will be plotted as a line on the graph. | ||
| :return: Return an altair chart with the current time series and the anomalies drawn. Data are plot in blue and | ||
| anomalies in red. | ||
| """ | ||
| data_df = data.pd_dataframe() | ||
|
|
||
| chart = ( | ||
| alt.Chart(data_df.reset_index()) | ||
| .mark_line() | ||
| .encode( | ||
| x="time:T", | ||
| y=alt.Y( | ||
| f"{AnomalyPlot._DATA_COLUMN_NAME}:Q", | ||
| axis=alt.Axis(title="Values"), | ||
| ), | ||
| ) | ||
| .properties( | ||
| title="Chart representing the data and the anomalies over the time", | ||
| width=600, | ||
| height=400, | ||
| ) | ||
| ) | ||
|
|
||
| if point_anomalies is not None: | ||
| anomalies_df = point_anomalies.pd_dataframe() | ||
| anomalies_df["anomalies_y"] = ( | ||
| anomalies_df["anomalies"] * data_df[AnomalyPlot._DATA_COLUMN_NAME] | ||
| ) | ||
|
|
||
| anomalies_chart = ( | ||
| alt.Chart(anomalies_df.reset_index()) | ||
| .mark_circle(color="red", size=100) | ||
| .encode( | ||
| x="time:T", | ||
| y="anomalies_y:Q", | ||
| ) | ||
| .transform_filter(alt.datum.anomalies == 1) | ||
| ) | ||
|
|
||
| chart += anomalies_chart | ||
|
|
||
| if contextual_anomalies is not None: | ||
| chart = AnomalyPlot._make_line_chart(data_df, chart, contextual_anomalies) | ||
|
|
||
| if collective_anomalies is not None: | ||
| chart = AnomalyPlot._make_area_chart(data_df, chart, collective_anomalies) | ||
|
|
||
| if seasonal_anomalies is not None: | ||
| chart = AnomalyPlot._make_line_chart(data_df, chart, seasonal_anomalies) | ||
|
|
||
| if cyclical_anomalies is not None: | ||
| chart = AnomalyPlot._make_line_chart(data_df, chart, cyclical_anomalies) | ||
|
|
||
| return chart | ||
|
|
||
| @staticmethod | ||
| def _make_line_chart( | ||
| data: pd.DataFrame, actual_chart: alt.Chart, anomalies: BinaryTimeSeries | ||
| ) -> alt.Chart: | ||
| """ | ||
| Make a line chart with the given anomalies. | ||
| :param data: TimeSeries within the data that are plotted in the chart. | ||
| :param actual_chart: Chart that will be updated with the anomalies. | ||
| :param anomalies: BinaryTimeSeries of anomalies that will be plotted in red on the chart. | ||
| :return: Return an altair chart with the current time series and the anomalies drawn. | ||
| """ | ||
| array_anomalies_df = AnomalyPlot.split_continuous_series(anomalies) | ||
| chart_total = actual_chart | ||
| for anomalies_df in array_anomalies_df: | ||
| anomalies_df["anomalies_y"] = ( | ||
| anomalies_df["anomalies"] * data[AnomalyPlot._DATA_COLUMN_NAME] | ||
| ) | ||
| chart = ( | ||
| alt.Chart(anomalies_df.reset_index()) | ||
| .mark_line(color="red", strokeWidth=2.5) | ||
| .encode( | ||
| x="time:T", | ||
| y="anomalies_y:Q", | ||
| ) | ||
| ) | ||
| chart_total += chart | ||
|
|
||
| return chart_total | ||
|
|
||
| @staticmethod | ||
| def _make_area_chart( | ||
| data_df: pd.DataFrame, actual_chart: alt.Chart, anomalies: BinaryTimeSeries | ||
| ) -> alt.Chart: | ||
| """ | ||
| Adding background to the chart with the given anomalies. | ||
| :param data_df: TimeSeries within the data that are plotted in the chart. | ||
| :param actual_chart: Chart that will be updated with the anomalies. | ||
| :param anomalies: BinaryTimeSeries of anomalies that will be used to color the background in red on the chart. | ||
| :return: Return an altair chart with the current time series and the anomalies drawn. | ||
| """ | ||
|
|
||
| y_max = data_df[AnomalyPlot._DATA_COLUMN_NAME].max() | ||
| y_min = data_df[AnomalyPlot._DATA_COLUMN_NAME].min() | ||
|
|
||
| delta = (y_max - y_min) * 0.1 | ||
|
|
||
| array_anomalies_df = AnomalyPlot.split_continuous_series(anomalies) | ||
| chart_total = None | ||
| for anomalies_df in array_anomalies_df: | ||
| anomalies_df["max"] = y_max + delta | ||
| anomalies_df["min"] = y_min - delta | ||
|
|
||
| chart = ( | ||
| alt.Chart(anomalies_df.reset_index()) | ||
| .mark_area(color="red", opacity=0.3) | ||
| .encode( | ||
| x="time:T", | ||
| y="min:Q", | ||
| y2="max:Q", | ||
| ) | ||
| ) | ||
|
|
||
| if chart_total is None: | ||
| chart_total = chart | ||
| else: | ||
| chart_total += chart | ||
|
|
||
| if chart_total is not None: | ||
| # Put our chart as background | ||
| return chart_total + actual_chart | ||
| return actual_chart | ||
|
|
||
| @staticmethod | ||
| def split_continuous_series(anomalies: BinaryTimeSeries) -> list[pd.DataFrame]: | ||
| """ | ||
| Split a continuous series of anomalies into multiple series of anomalies. The response covert all the anomalies | ||
| in the BinaryTimeSeries but change its structure to have a list with one df by continuous series of 1. Zeros are | ||
| not represented anymore. | ||
| :param anomalies: BinaryTimeSeries of anomalies that will be split. | ||
| :return: Return a list of DataFrame of anomalies. | ||
| """ | ||
| anomalies_df = anomalies.pd_dataframe() | ||
|
|
||
| # Initialize variables | ||
| result_dfs = [] | ||
| current_value = 0 | ||
| current_df = None | ||
|
|
||
| # Iterate through each row of the base_df | ||
| for idx, row in anomalies_df.iterrows(): | ||
| value = row["anomalies"] | ||
|
|
||
| # If value changes | ||
| if value != current_value: | ||
| if current_df is not None: | ||
| # Save the current DataFrame into the array | ||
| current_df = current_df.rename_axis("time") | ||
| result_dfs.append(current_df) | ||
| current_df = None | ||
|
|
||
| else: | ||
| # Create a new DataFrame with the same structure as base_df | ||
| current_df = pd.DataFrame(columns=anomalies_df.columns) | ||
| # Save the current row in the new DataFrame | ||
| current_df = pd.concat( | ||
| [current_df, pd.DataFrame([row], index=[idx])] | ||
| ) | ||
|
|
||
| current_value = value | ||
|
|
||
| else: | ||
| # If value does not change | ||
| if current_df is not None: | ||
| # Save the row in the current DataFrame | ||
| current_df = pd.concat( | ||
| [current_df, pd.DataFrame([row], index=[idx])] | ||
| ) | ||
|
|
||
| if current_value == 1: | ||
| current_df = current_df.rename_axis("time") | ||
| result_dfs.append(current_df) | ||
|
|
||
| return result_dfs |
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,6 +1,5 @@ | ||
| import xarray as xr | ||
| import numpy as np | ||
|
|
||
| from .resticted_time_series import RestrictedTimeSeries | ||
|
|
||
|
|
||
|
|
||
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,6 +1,5 @@ | ||
| import xarray as xr | ||
| import numpy as np | ||
|
|
||
| from .resticted_time_series import RestrictedTimeSeries | ||
|
|
||
|
|
||
|
|
||
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