mirror of
https://github.com/VictoriaMetrics/VictoriaMetrics.git
synced 2025-01-20 07:19:17 +01:00
update wiki pages
Vika
parent
86d0534654
commit
4f3ab79de1
@ -14,7 +14,7 @@ aliases:
|
||||
# FAQ - VictoriaMetrics Anomaly Detection
|
||||
|
||||
## What is VictoriaMetrics Anomaly Detection (vmanomaly)?
|
||||
VictoriaMetrics Anomaly Detection, also known as `vmanomaly`, is a service for detecting unexpected changes in time series data. Utilizing machine learning models, it computes and pushes back an ["anomaly score"](/anomaly-detection/components/models/models.html#vmanomaly-output) for user-specified metrics. This hands-off approach to anomaly detection reduces the need for manual alert setup and can adapt to various metrics, improving your observability experience.
|
||||
VictoriaMetrics Anomaly Detection, also known as `vmanomaly`, is a service for detecting unexpected changes in time series data. Utilizing machine learning models, it computes and pushes back an ["anomaly score"](/anomaly-detection/components/models.html#vmanomaly-output) for user-specified metrics. This hands-off approach to anomaly detection reduces the need for manual alert setup and can adapt to various metrics, improving your observability experience.
|
||||
|
||||
Please refer to [our guide section](/anomaly-detection/#practical-guides-and-installation) to find out more.
|
||||
|
||||
@ -32,10 +32,10 @@ Respective config is defined in a [`reader`](/anomaly-detection/components/reade
|
||||
`vmanomaly` operates on data fetched from VictoriaMetrics using [MetricsQL](https://docs.victoriametrics.com/MetricsQL.html) queries, so the initial data quality can be fine-tuned with aggregation, grouping, and filtering to reduce noise and improve anomaly detection accuracy.
|
||||
|
||||
## Output produced by vmanomaly
|
||||
`vmanomaly` models generate [metrics](/anomaly-detection/components/models/models.html#vmanomaly-output) like `anomaly_score`, `yhat`, `yhat_lower`, `yhat_upper`, and `y`. These metrics provide a comprehensive view of the detected anomalies. The service also produces [health check metrics](/anomaly-detection/components/monitoring.html#metrics-generated-by-vmanomaly) for monitoring its performance.
|
||||
`vmanomaly` models generate [metrics](/anomaly-detection/components/models.html#vmanomaly-output) like `anomaly_score`, `yhat`, `yhat_lower`, `yhat_upper`, and `y`. These metrics provide a comprehensive view of the detected anomalies. The service also produces [health check metrics](/anomaly-detection/components/monitoring.html#metrics-generated-by-vmanomaly) for monitoring its performance.
|
||||
|
||||
## Choosing the right model for vmanomaly
|
||||
Selecting the best model for `vmanomaly` depends on the data's nature and the types of anomalies to detect. For instance, [Z-score](anomaly-detection/components/models/models.html#z-score) is suitable for data without trends or seasonality, while more complex patterns might require models like [Prophet](anomaly-detection/components/models/models.html#prophet).
|
||||
Selecting the best model for `vmanomaly` depends on the data's nature and the types of anomalies to detect. For instance, [Z-score](anomaly-detection/components/models.html#z-score) is suitable for data without trends or seasonality, while more complex patterns might require models like [Prophet](anomaly-detection/components/models.html#prophet).
|
||||
|
||||
Please refer to [respective blogpost on anomaly types and alerting heuristics](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-2/) for more details.
|
||||
|
||||
|
||||
@ -16,7 +16,7 @@ aliases:
|
||||
|
||||
# VictoriaMetrics Anomaly Detection
|
||||
|
||||
In the dynamic and complex world of system monitoring, VictoriaMetrics Anomaly Detection, being a part of our [Enterprise offering](https://victoriametrics.com/products/enterprise/), stands as a pivotal tool for achieving advanced observability. It empowers SREs and DevOps teams by automating the intricate task of identifying abnormal behavior in time-series data. It goes beyond traditional threshold-based alerting, utilizing machine learning techniques to not only detect anomalies but also minimize false positives, thus reducing alert fatigue. By providing simplified alerting mechanisms atop of [unified anomaly scores](/anomaly-detection/components/models/models.html#vmanomaly-output), it enables teams to spot and address potential issues faster, ensuring system reliability and operational efficiency.
|
||||
In the dynamic and complex world of system monitoring, VictoriaMetrics Anomaly Detection, being a part of our [Enterprise offering](https://victoriametrics.com/products/enterprise/), stands as a pivotal tool for achieving advanced observability. It empowers SREs and DevOps teams by automating the intricate task of identifying abnormal behavior in time-series data. It goes beyond traditional threshold-based alerting, utilizing machine learning techniques to not only detect anomalies but also minimize false positives, thus reducing alert fatigue. By providing simplified alerting mechanisms atop of [unified anomaly scores](/anomaly-detection/components/models.html#vmanomaly-output), it enables teams to spot and address potential issues faster, ensuring system reliability and operational efficiency.
|
||||
|
||||
## Practical Guides and Installation
|
||||
Begin your VictoriaMetrics Anomaly Detection journey with ease using our guides and installation instructions:
|
||||
|
||||
484
anomaly-detection/components/models.md
Normal file
484
anomaly-detection/components/models.md
Normal file
@ -0,0 +1,484 @@
|
||||
---
|
||||
title: Models
|
||||
weight: 1
|
||||
# sort: 1
|
||||
menu:
|
||||
docs:
|
||||
identifier: "vmanomaly-models"
|
||||
parent: "vmanomaly-components"
|
||||
weight: 1
|
||||
# sort: 1
|
||||
aliases:
|
||||
- /anomaly-detection/components/models.html
|
||||
- /anomaly-detection/components/models/custom_model.html
|
||||
- /anomaly-detection/components/models/models.html
|
||||
---
|
||||
|
||||
# Models
|
||||
|
||||
This section describes `Model` component of VictoriaMetrics Anomaly Detection (or simply [`vmanomaly`](/vmanomaly.html)) and the guide of how to define a respective section of a config to launch the service.
|
||||
vmanomaly includes various [built-in models](#built-in-models) and you can integrate your custom model with vmanomaly see [custom model](#custom-model-guide)
|
||||
|
||||
|
||||
## Built-in Models
|
||||
|
||||
### Overview
|
||||
VM Anomaly Detection (`vmanomaly` hereinafter) models support 2 groups of parameters:
|
||||
|
||||
- **`vmanomaly`-specific** arguments - please refer to *Parameters specific for vmanomaly* and *Default model parameters* subsections for each of the models below.
|
||||
- Arguments to **inner model** (say, [Facebook's Prophet](https://facebook.github.io/prophet/docs/quick_start.html#python-api)), passed in a `args` argument as key-value pairs, that will be directly given to the model during initialization to allow granular control. Optional.
|
||||
|
||||
**Note**: For users who may not be familiar with Python data types such as `list[dict]`, a [dictionary](https://www.w3schools.com/python/python_dictionaries.asp) in Python is a data structure that stores data values in key-value pairs. This structure allows for efficient data retrieval and management.
|
||||
|
||||
|
||||
**Models**:
|
||||
|
||||
* [ARIMA](#arima)
|
||||
* [Holt-Winters](#holt-winters)
|
||||
* [Prophet](#prophet)
|
||||
* [Rolling Quantile](#rolling-quantile)
|
||||
* [Seasonal Trend Decomposition](#seasonal-trend-decomposition)
|
||||
* [Z-score](#z-score)
|
||||
* [MAD (Median Absolute Deviation)](#mad-median-absolute-deviation)
|
||||
* [Isolation forest (Multivariate)](#isolation-forest-multivariate)
|
||||
* [Custom model](#custom-model)
|
||||
|
||||
### [ARIMA](https://en.wikipedia.org/wiki/Autoregressive_integrated_moving_average)
|
||||
Here we use ARIMA implementation from `statsmodels` [library](https://www.statsmodels.org/dev/generated/statsmodels.tsa.arima.model.ARIMA.html)
|
||||
|
||||
*Parameters specific for vmanomaly*:
|
||||
|
||||
* `class` (string) - model class name `"model.arima.ArimaModel"`
|
||||
|
||||
* `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculating boundaries to define anomaly score. Defaults to `2.5`.
|
||||
|
||||
* `provide_series` (list[string], optional) - List of columns to be produced and returned by the model. Defaults to `["anomaly_score", "yhat", "yhat_lower" "yhat_upper", "y"]`. Output can be **only a subset** of a given column list.
|
||||
|
||||
* `resample_freq` (string, optional) - Frequency to resample input data into, e.g. data comes at 15 seconds resolution, and resample_freq is '1m'. Then fitting data will be downsampled to '1m' and internal model is trained at '1m' intervals. So, during inference, prediction data would be produced at '1m' intervals, but interpolated to "15s" to match with expected output, as output data must have the same timestamps.
|
||||
|
||||
*Default model parameters*:
|
||||
|
||||
* `order` (list[int]) - ARIMA's (p,d,q) order of the model for the autoregressive, differences, and moving average components, respectively.
|
||||
|
||||
* `args` (dict, optional) - Inner model args (key-value pairs). See accepted params in [model documentation](https://www.statsmodels.org/dev/generated/statsmodels.tsa.arima.model.ARIMA.html). Defaults to empty (not provided). Example: {"trend": "c"}
|
||||
|
||||
*Config Example*
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```yaml
|
||||
model:
|
||||
class: "model.arima.ArimaModel"
|
||||
# ARIMA's (p,d,q) order
|
||||
order: [1, 1, 0]
|
||||
z_threshold: 2.7
|
||||
resample_freq: '1m'
|
||||
# Inner model args (key-value pairs) accepted by statsmodels.tsa.arima.model.ARIMA
|
||||
args:
|
||||
trend: 'c'
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
### [Holt-Winters](https://en.wikipedia.org/wiki/Exponential_smoothing)
|
||||
Here we use Holt-Winters Exponential Smoothing implementation from `statsmodels` [library](https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing.html). All parameters from this library can be passed to the model.
|
||||
|
||||
*Parameters specific for vmanomaly*:
|
||||
|
||||
* `class` (string) - model class name `"model.holtwinters.HoltWinters"`
|
||||
|
||||
* `frequency` (string) - Must be set equal to sampling_period. Model needs to know expected data-points frequency (e.g. '10m'). If omitted, frequency is guessed during fitting as **the median of intervals between fitting data timestamps**. During inference, if incoming data doesn't have the same frequency, then it will be interpolated. E.g. data comes at 15 seconds resolution, and our resample_freq is '1m'. Then fitting data will be downsampled to '1m' and internal model is trained at '1m' intervals. So, during inference, prediction data would be produced at '1m' intervals, but interpolated to "15s" to match with expected output, as output data must have the same timestamps. As accepted by pandas.Timedelta (e.g. '5m').
|
||||
|
||||
* `seasonality` (string, optional) - As accepted by pandas.Timedelta.
|
||||
|
||||
* If `seasonal_periods` is not specified, it is calculated as `seasonality` / `frequency`
|
||||
Used to compute "seasonal_periods" param for the model (e.g. '1D' or '1W').
|
||||
|
||||
* `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculating boundaries to define anomaly score. Defaults to 2.5.
|
||||
|
||||
|
||||
*Default model parameters*:
|
||||
|
||||
* If [parameter](https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing.html#statsmodels.tsa.holtwinters.ExponentialSmoothing-parameters) `seasonal` is not specified, default value will be `add`.
|
||||
|
||||
* If [parameter](https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing.html#statsmodels.tsa.holtwinters.ExponentialSmoothing-parameters) `initialization_method` is not specified, default value will be `estimated`.
|
||||
|
||||
* `args` (dict, optional) - Inner model args (key-value pairs). See accepted params in [model documentation](https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing.html#statsmodels.tsa.holtwinters.ExponentialSmoothing-parameters). Defaults to empty (not provided). Example: {"seasonal": "add", "initialization_method": "estimated"}
|
||||
|
||||
*Config Example*
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```yaml
|
||||
model:
|
||||
class: "model.holtwinters.HoltWinters"
|
||||
seasonality: '1d'
|
||||
frequency: '1h'
|
||||
# Inner model args (key-value pairs) accepted by statsmodels.tsa.holtwinters.ExponentialSmoothing
|
||||
args:
|
||||
seasonal: 'add'
|
||||
initialization_method: 'estimated'
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
Resulting metrics of the model are described [here](#vmanomaly-output).
|
||||
|
||||
### [Prophet](https://facebook.github.io/prophet/)
|
||||
Here we utilize the Facebook Prophet implementation, as detailed in their [library documentation](https://facebook.github.io/prophet/docs/quick_start.html#python-api). All parameters from this library are compatible and can be passed to the model.
|
||||
|
||||
*Parameters specific for vmanomaly*:
|
||||
|
||||
* `class` (string) - model class name `"model.prophet.ProphetModel"`
|
||||
* `seasonalities` (list[dict], optional) - Extra seasonalities to pass to Prophet. See [`add_seasonality()`](https://facebook.github.io/prophet/docs/seasonality,_holiday_effects,_and_regressors.html#modeling-holidays-and-special-events:~:text=modeling%20the%20cycle-,Specifying,-Custom%20Seasonalities) Prophet param.
|
||||
* `provide_series` (dict, optional) - model resulting metrics. If not specified [standard metrics](#vmanomaly-output) will be provided.
|
||||
|
||||
**Note**: Apart from standard vmanomaly output Prophet model can provide [additional metrics](#additional-output-metrics-produced-by-fb-prophet).
|
||||
|
||||
**Additional output metrics produced by FB Prophet**
|
||||
Depending on chosen `seasonality` parameter FB Prophet can return additional metrics such as:
|
||||
- `trend`, `trend_lower`, `trend_upper`
|
||||
- `additive_terms`, `additive_terms_lower`, `additive_terms_upper`,
|
||||
- `multiplicative_terms`, `multiplicative_terms_lower`, `multiplicative_terms_upper`,
|
||||
- `daily`, `daily_lower`, `daily_upper`,
|
||||
- `hourly`, `hourly_lower`, `hourly_upper`,
|
||||
- `holidays`, `holidays_lower`, `holidays_upper`,
|
||||
- and a number of columns for each holiday if `holidays` param is set
|
||||
|
||||
*Config Example*
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```yaml
|
||||
model:
|
||||
class: "model.prophet.ProphetModel"
|
||||
seasonalities:
|
||||
- name: 'hourly'
|
||||
period: 0.04166666666
|
||||
fourier_order: 30
|
||||
# Inner model args (key-value pairs) accepted by
|
||||
# https://facebook.github.io/prophet/docs/quick_start.html#python-api
|
||||
args:
|
||||
# See https://facebook.github.io/prophet/docs/uncertainty_intervals.html
|
||||
interval_width: 0.98
|
||||
country_holidays: 'US'
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
Resulting metrics of the model are described [here](#vmanomaly-output)
|
||||
|
||||
### [Rolling Quantile](https://en.wikipedia.org/wiki/Quantile)
|
||||
|
||||
*Parameters specific for vmanomaly*:
|
||||
|
||||
* `class` (string) - model class name `"model.rolling_quantile.RollingQuantileModel"`
|
||||
* `quantile` (float) - quantile value, from 0.5 to 1.0. This constraint is implied by 2-sided confidence interval.
|
||||
* `window_steps` (integer) - size of the moving window. (see 'sampling_period')
|
||||
|
||||
*Config Example*
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```yaml
|
||||
model:
|
||||
class: "model.rolling_quantile.RollingQuantileModel"
|
||||
quantile: 0.9
|
||||
window_steps: 96
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
Resulting metrics of the model are described [here](#vmanomaly-output).
|
||||
|
||||
### [Seasonal Trend Decomposition](https://en.wikipedia.org/wiki/Seasonal_adjustment)
|
||||
Here we use Seasonal Decompose implementation from `statsmodels` [library](https://www.statsmodels.org/dev/generated/statsmodels.tsa.seasonal.seasonal_decompose.html). Parameters from this library can be passed to the model. Some parameters are specifically predefined in vmanomaly and can't be changed by user(`model`='additive', `two_sided`=False).
|
||||
|
||||
*Parameters specific for vmanomaly*:
|
||||
|
||||
* `class` (string) - model class name `"model.std.StdModel"`
|
||||
* `period` (integer) - Number of datapoints in one season.
|
||||
* `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculating boundaries to define anomaly score. Defaults to `2.5`.
|
||||
|
||||
|
||||
*Config Example*
|
||||
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```yaml
|
||||
model:
|
||||
class: "model.std.StdModel"
|
||||
period: 2
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
Resulting metrics of the model are described [here](#vmanomaly-output).
|
||||
|
||||
**Additional output metrics produced by Seasonal Trend Decomposition model**
|
||||
* `resid` - The residual component of the data series.
|
||||
* `trend` - The trend component of the data series.
|
||||
* `seasonal` - The seasonal component of the data series.
|
||||
|
||||
### [MAD (Median Absolute Deviation)](https://en.wikipedia.org/wiki/Median_absolute_deviation)
|
||||
The MAD model is a robust method for anomaly detection that is *less sensitive* to outliers in data compared to standard deviation-based models. It considers a point as an anomaly if the absolute deviation from the median is significantly large.
|
||||
|
||||
*Parameters specific for vmanomaly*:
|
||||
|
||||
* `class` (string) - model class name `"model.mad.MADModel"`
|
||||
* `threshold` (float, optional) - The threshold multiplier for the MAD to determine anomalies. Defaults to `2.5`. Higher values will identify fewer points as anomalies.
|
||||
|
||||
*Config Example*
|
||||
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```yaml
|
||||
model:
|
||||
class: "model.mad.MADModel"
|
||||
threshold: 2.5
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
Resulting metrics of the model are described [here](#vmanomaly-output).
|
||||
|
||||
### [Z-score](https://en.wikipedia.org/wiki/Standard_score)
|
||||
*Parameters specific for vmanomaly*:
|
||||
|
||||
* `class` (string) - model class name `"model.zscore.ZscoreModel"`
|
||||
* `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculation boundaries and anomaly score. Defaults to `2.5`.
|
||||
|
||||
*Config Example*
|
||||
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```yaml
|
||||
model:
|
||||
class: "model.zscore.ZscoreModel"
|
||||
z_threshold: 2.5
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
Resulting metrics of the model are described [here](#vmanomaly-output).
|
||||
|
||||
### [Isolation forest](https://en.wikipedia.org/wiki/Isolation_forest) (Multivariate)
|
||||
Detects anomalies using binary trees. The algorithm has a linear time complexity and a low memory requirement, which works well with high-volume data. It can be used on both univatiate and multivariate data, but it is more effective in multivariate case.
|
||||
|
||||
**Important**: Be aware of [the curse of dimensionality](https://en.wikipedia.org/wiki/Curse_of_dimensionality). Don't use single multivariate model if you expect your queries to return many time series of less datapoints that the number of metrics. In such case it is hard for a model to learn meaningful dependencies from too sparse data hypercube.
|
||||
|
||||
Here we use Isolation Forest implementation from `scikit-learn` [library](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.IsolationForest.html). All parameters from this library can be passed to the model.
|
||||
|
||||
*Parameters specific for vmanomaly*:
|
||||
|
||||
* `class` (string) - model class name `"model.isolation_forest.IsolationForestMultivariateModel"`
|
||||
|
||||
* `contamination` (float or string, optional) - The amount of contamination of the data set, i.e. the proportion of outliers in the data set. Used when fitting to define the threshold on the scores of the samples. Default value - "auto". Should be either `"auto"` or be in the range (0.0, 0.5].
|
||||
|
||||
* `args` (dict, optional) - Inner model args (key-value pairs). See accepted params in [model documentation](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.IsolationForest.html). Defaults to empty (not provided). Example: {"random_state": 42, "n_estimators": 100}
|
||||
|
||||
*Config Example*
|
||||
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```yaml
|
||||
model:
|
||||
# To use univariate model, substitute class argument with "model.isolation_forest.IsolationForestModel".
|
||||
class: "model.isolation_forest.IsolationForestMultivariateModel"
|
||||
contamination: "auto"
|
||||
args:
|
||||
n_estimators: 100
|
||||
# i.e. to assure reproducibility of produced results each time model is fit on the same input
|
||||
random_state: 42
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
Resulting metrics of the model are described [here](#vmanomaly-output).
|
||||
|
||||
## vmanomaly output
|
||||
|
||||
When vmanomaly is executed, it generates various metrics, the specifics of which depend on the model employed.
|
||||
These metrics can be renamed in the writer's section.
|
||||
|
||||
The default metrics produced by vmanomaly include:
|
||||
|
||||
- `anomaly_score`: This is the *primary* metric.
|
||||
- It is designed in such a way that values from 0.0 to 1.0 indicate non-anomalous data.
|
||||
- A value greater than 1.0 is generally classified as an anomaly, although this threshold can be adjusted in the alerting configuration.
|
||||
- The decision to set the changepoint at 1 was made to ensure consistency across various models and alerting configurations, such that a score above 1 consistently signifies an anomaly.
|
||||
|
||||
- `yhat`: This represents the predicted expected value.
|
||||
|
||||
- `yhat_lower`: This indicates the predicted lower boundary.
|
||||
|
||||
- `yhat_upper`: This refers to the predicted upper boundary.
|
||||
|
||||
- `y`: This is the original value obtained from the query result.
|
||||
|
||||
**Important**: Be aware that if `NaN` (Not a Number) or `Inf` (Infinity) values are present in the input data during `infer` model calls, the model will produce `NaN` as the `anomaly_score` for these particular instances.
|
||||
|
||||
|
||||
## Healthcheck metrics
|
||||
|
||||
Each model exposes [several healthchecks metrics](./../monitoring.html#models-behaviour-metrics) to its `health_path` endpoint:
|
||||
|
||||
|
||||
## Custom Model Guide
|
||||
|
||||
Apart from vmanomaly predefined models, users can create their own custom models for anomaly detection.
|
||||
|
||||
Here in this guide, we will
|
||||
- Make a file containing our custom model definition
|
||||
- Define VictoriaMetrics Anomaly Detection config file to use our custom model
|
||||
- Run service
|
||||
|
||||
**Note**: The file containing the model should be written in [Python language](https://www.python.org/) (3.11+)
|
||||
|
||||
### 1. Custom model
|
||||
|
||||
We'll create `custom_model.py` file with `CustomModel` class that will inherit from vmanomaly `Model` base class.
|
||||
In the `CustomModel` class there should be three required methods - `__init__`, `fit` and `infer`:
|
||||
* `__init__` method should initiate parameters for the model.
|
||||
|
||||
**Note**: if your model relies on configs that have `arg` [key-value pair argument](./models.md#section-overview), do not forget to use Python's `**kwargs` in method's signature and to explicitly call
|
||||
```python
|
||||
super().__init__(**kwargs)
|
||||
```
|
||||
to initialize the base class each model derives from
|
||||
* `fit` method should contain the model training process.
|
||||
* `infer` should return Pandas.DataFrame object with model's inferences.
|
||||
|
||||
For the sake of simplicity, the model in this example will return one of two values of `anomaly_score` - 0 or 1 depending on input parameter `percentage`.
|
||||
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```python
|
||||
import numpy as np
|
||||
import pandas as pd
|
||||
import scipy.stats as st
|
||||
import logging
|
||||
|
||||
from model.model import Model
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class CustomModel(Model):
|
||||
"""
|
||||
Custom model implementation.
|
||||
"""
|
||||
|
||||
def __init__(self, percentage: float = 0.95, **kwargs):
|
||||
super().__init__(**kwargs)
|
||||
self.percentage = percentage
|
||||
self._mean = np.nan
|
||||
self._std = np.nan
|
||||
|
||||
def fit(self, df: pd.DataFrame):
|
||||
# Model fit process:
|
||||
y = df['y']
|
||||
self._mean = np.mean(y)
|
||||
self._std = np.std(y)
|
||||
if self._std == 0.0:
|
||||
self._std = 1 / 65536
|
||||
|
||||
|
||||
def infer(self, df: pd.DataFrame) -> np.array:
|
||||
# Inference process:
|
||||
y = df['y']
|
||||
zscores = (y - self._mean) / self._std
|
||||
anomaly_score_cdf = st.norm.cdf(np.abs(zscores))
|
||||
df_pred = df[['timestamp', 'y']].copy()
|
||||
df_pred['anomaly_score'] = anomaly_score_cdf > self.percentage
|
||||
df_pred['anomaly_score'] = df_pred['anomaly_score'].astype('int32', errors='ignore')
|
||||
|
||||
return df_pred
|
||||
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
|
||||
### 2. Configuration file
|
||||
|
||||
Next, we need to create `config.yaml` file with VM Anomaly Detection configuration and model input parameters.
|
||||
In the config file `model` section we need to put our model class `model.custom.CustomModel` and all parameters used in `__init__` method.
|
||||
You can find out more about configuration parameters in vmanomaly docs.
|
||||
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```yaml
|
||||
scheduler:
|
||||
infer_every: "1m"
|
||||
fit_every: "1m"
|
||||
fit_window: "1d"
|
||||
|
||||
model:
|
||||
# note: every custom model should implement this exact path, specified in `class` field
|
||||
class: "model.model.CustomModel"
|
||||
# custom model params are defined here
|
||||
percentage: 0.9
|
||||
|
||||
reader:
|
||||
datasource_url: "http://localhost:8428/"
|
||||
queries:
|
||||
ingestion_rate: 'sum(rate(vm_rows_inserted_total)) by (type)'
|
||||
churn_rate: 'sum(rate(vm_new_timeseries_created_total[5m]))'
|
||||
|
||||
writer:
|
||||
datasource_url: "http://localhost:8428/"
|
||||
metric_format:
|
||||
__name__: "custom_$VAR"
|
||||
for: "$QUERY_KEY"
|
||||
model: "custom"
|
||||
run: "test-format"
|
||||
|
||||
monitoring:
|
||||
# /metrics server.
|
||||
pull:
|
||||
port: 8080
|
||||
push:
|
||||
url: "http://localhost:8428/"
|
||||
extra_labels:
|
||||
job: "vmanomaly-develop"
|
||||
config: "custom.yaml"
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
### 3. Running custom model
|
||||
Let's pull the docker image for vmanomaly:
|
||||
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```sh
|
||||
docker pull us-docker.pkg.dev/victoriametrics-test/public/vmanomaly-trial:latest
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
Now we can run the docker container putting as volumes both config and model file:
|
||||
|
||||
**Note**: place the model file to `/model/custom.py` path when copying
|
||||
<div class="with-copy" markdown="1">
|
||||
|
||||
```sh
|
||||
docker run -it \
|
||||
--net [YOUR_NETWORK] \
|
||||
-v [YOUR_LICENSE_FILE_PATH]:/license.txt \
|
||||
-v $(PWD)/custom_model.py:/vmanomaly/src/model/custom.py \
|
||||
-v $(PWD)/custom.yaml:/config.yaml \
|
||||
us-docker.pkg.dev/victoriametrics-test/public/vmanomaly-trial:latest /config.yaml \
|
||||
--license-file=/license.txt
|
||||
```
|
||||
|
||||
</div>
|
||||
|
||||
Please find more detailed instructions (license, etc.) [here](/vmanomaly.html#run-vmanomaly-docker-container)
|
||||
|
||||
|
||||
### Output
|
||||
As the result, this model will return metric with labels, configured previously in `config.yaml`.
|
||||
In this particular example, 2 metrics will be produced. Also, there will be added other metrics from input query result.
|
||||
|
||||
```
|
||||
{__name__="custom_anomaly_score", for="ingestion_rate", model="custom", run="test-format"}
|
||||
|
||||
{__name__="custom_anomaly_score", for="churn_rate", model="custom", run="test-format"}
|
||||
```
|
||||
@ -50,7 +50,7 @@ Future updates will introduce additional export methods, offering users more fle
|
||||
<td><code>__name__: "vmanomaly_$VAR"</code></td>
|
||||
<td rowspan="4">Metrics to save the output (in metric names or labels). Must have <code>__name__</code> key. Must have a value with <code>$VAR</code> placeholder in it to distinguish between resulting metrics. Supported placeholders:
|
||||
<ul>
|
||||
<li><code>$VAR</code> -- Variables that model provides, all models provide the following set: {"anomaly_score", "y", "yhat", "yhat_lower", "yhat_upper"}. Description of standard output is <a href="/anomaly-detection/components/models/models.html#vmanomaly-output">here</a>. Depending on <a href="/anomaly-detection/components/models/models.html">model type</a> it can provide more metrics, like "trend", "seasonality" etc.</li>
|
||||
<li><code>$VAR</code> -- Variables that model provides, all models provide the following set: {"anomaly_score", "y", "yhat", "yhat_lower", "yhat_upper"}. Description of standard output is <a href="/anomaly-detection/components/models.html#vmanomaly-output">here</a>. Depending on <a href="/anomaly-detection/components/models.html">model type</a> it can provide more metrics, like "trend", "seasonality" etc.</li>
|
||||
<li><code>$QUERY_KEY</code> -- E.g. "ingestion_rate".</li>
|
||||
</ul>
|
||||
Other keys are supposed to be configured by the user to help identify generated metrics, e.g., specific config file name etc.
|
||||
@ -130,7 +130,7 @@ __name__: PREFIX1_$VAR
|
||||
for: PREFIX2_$QUERY_KEY
|
||||
```
|
||||
|
||||
* for `__name__` parameter it will name metrics returned by models as `PREFIX1_anomaly_score`, `PREFIX1_yhat_lower`, etc. Vmanomaly output metrics names described [here](anomaly-detection/components/models/models.html#vmanomaly-output)
|
||||
* for `__name__` parameter it will name metrics returned by models as `PREFIX1_anomaly_score`, `PREFIX1_yhat_lower`, etc. Vmanomaly output metrics names described [here](anomaly-detection/components/models.html#vmanomaly-output)
|
||||
* for `for` parameter will add labels `PREFIX2_query_name_1`, `PREFIX2_query_name_2`, etc. Query names are set as aliases in config `reader` section in [`queries`](anomaly-detection/components/reader.html#config-parameters) parameter.
|
||||
|
||||
It is possible to specify other custom label names needed.
|
||||
|
||||
@ -43,7 +43,7 @@ _Note: if you have more than one Prometheus, you need to run this query across a
|
||||
|
||||
### Churn Rate
|
||||
|
||||
The higher the Churn Rate, the more compute resources are needed for the efficient work of VictoriaMetrics. It is recommended to lower the churn rate as much as possible. The tolerable churn rate is less than 5% of the number of Active Time Series.
|
||||
The higher the Churn Rate, the more compute resources are needed for the efficient work of VictoriaMetrics. It is recommended to lower the churn rate as much as possible.
|
||||
|
||||
The high Churn Rate is commonly a result of using high-volatile labels, such as `client_id`, `url`, `checksum`, `timestamp`, etc. In Kubernetes, the pod's name is also a volatile label because it changes each time pod is redeployed. For example, a service exposes 1000 time series. If we deploy 100 replicas of the service, the total amount of Active Time Series will be 1000*100 = 100000. If we redeploy the service, each replica's pod name will change, and the number of Active Time Series will double because all the time series will update the pod's name label.
|
||||
|
||||
|
||||
16
vmanomaly.md
16
vmanomaly.md
@ -52,7 +52,7 @@ processes in parallel, each using its own config.
|
||||
Currently, vmanomaly ships with a set of built-in models:
|
||||
> For a detailed description, see [model section](/anomaly-detection/components/models)
|
||||
|
||||
1. [**ZScore**](/anomaly-detection/components/models/models.html#z-score)
|
||||
1. [**ZScore**](/anomaly-detection/components/models.html#z-score)
|
||||
|
||||
_(useful for testing)_
|
||||
|
||||
@ -60,7 +60,7 @@ Currently, vmanomaly ships with a set of built-in models:
|
||||
from time-series mean (straight line). Keeps only two model parameters internally:
|
||||
`mean` and `std` (standard deviation).
|
||||
|
||||
1. [**Prophet**](/anomaly-detection/components/models/models.html#prophet)
|
||||
1. [**Prophet**](/anomaly-detection/components/models.html#prophet)
|
||||
|
||||
_(simplest in configuration, recommended for getting starting)_
|
||||
|
||||
@ -74,36 +74,36 @@ Currently, vmanomaly ships with a set of built-in models:
|
||||
|
||||
See [Prophet documentation](https://facebook.github.io/prophet/)
|
||||
|
||||
1. [**Holt-Winters**](/anomaly-detection/components/models/models.html#holt-winters)
|
||||
1. [**Holt-Winters**](/anomaly-detection/components/models.html#holt-winters)
|
||||
|
||||
Very popular forecasting algorithm. See [statsmodels.org documentation](
|
||||
https://www.statsmodels.org/stable/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing.html)
|
||||
for Holt-Winters exponential smoothing.
|
||||
|
||||
1. [**Seasonal-Trend Decomposition**](/anomaly-detection/components/models/models.html#seasonal-trend-decomposition)
|
||||
1. [**Seasonal-Trend Decomposition**](/anomaly-detection/components/models.html#seasonal-trend-decomposition)
|
||||
|
||||
Extracts three components: season, trend, and residual, that can be plotted individually for
|
||||
easier debugging. Uses LOESS (locally estimated scatterplot smoothing).
|
||||
See [statsmodels.org documentation](https://www.statsmodels.org/dev/examples/notebooks/generated/stl_decomposition.html)
|
||||
for LOESS STD.
|
||||
|
||||
1. [**ARIMA**](/anomaly-detection/components/models/models.html#arima)
|
||||
1. [**ARIMA**](/anomaly-detection/components/models.html#arima)
|
||||
|
||||
Commonly used forecasting model. See [statsmodels.org documentation](https://www.statsmodels.org/stable/generated/statsmodels.tsa.arima.model.ARIMA.html) for ARIMA.
|
||||
|
||||
1. [**Rolling Quantile**](/anomaly-detection/components/models/models.html#rolling-quantile)
|
||||
1. [**Rolling Quantile**](/anomaly-detection/components/models.html#rolling-quantile)
|
||||
|
||||
A simple moving window of quantiles. Easy to use, easy to understand, but not as powerful as
|
||||
other models.
|
||||
|
||||
1. [**Isolation Forest**](/anomaly-detection/components/models/models.html#isolation-forest-multivariate)
|
||||
1. [**Isolation Forest**](/anomaly-detection/components/models.html#isolation-forest-multivariate)
|
||||
|
||||
Detects anomalies using binary trees. It works for both univariate and multivariate data. Be aware of [the curse of dimensionality](https://en.wikipedia.org/wiki/Curse_of_dimensionality) in the case of multivariate data - we advise against using a single model when handling multiple time series *if the number of these series significantly exceeds their average length (# of data points)*.
|
||||
|
||||
The algorithm has a linear time complexity and a low memory requirement, which works well with high-volume data. See [scikit-learn.org documentation](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.IsolationForest.html) for Isolation Forest.
|
||||
|
||||
|
||||
1. [**MAD (Median Absolute Deviation)**](anomaly-detection/components/models/models.html#mad-median-absolute-deviation)
|
||||
1. [**MAD (Median Absolute Deviation)**](anomaly-detection/components/models.html#mad-median-absolute-deviation)
|
||||
|
||||
A robust method for anomaly detection that is less sensitive to outliers in data compared to standard deviation-based models. It considers a point as an anomaly if the absolute deviation from the median is significantly large.
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user