Monitoring Anomalies with a Custom Metric

Overview

In this tutorial, you will learn how to create a custom anomaly detection metric for a specific use case.

Let's take a problem described in the previous Train & Deploy Census Income Classification Model tutorial as a use case and census income dataset as a data source. We will monitor a model that classifies whether the income of a given person exceeds $50.000 per year.

By the end of this tutorial you will know how to:

• Train a monitoring model

• Deploy of a monitoring model with SDK

• Manage сustom metrics with UI

• Upload a monitoring model with CLI

Prerequisites

For this tutorial, you need to have Hydrosphere Platform deployed and Hydrosphere CLI (hs) along with Python SDK (hydrosdk) _**_installed on your local machine. If you don't have them yet, please follow these guides first:

• Platform Installation

• CLI

• Python SDK

This tutorial is a sequel to the previous tutorial. Please complete it first to have a prepared dataset and a trained model deployed to the cluster:

• Train & Deploy Census Income Classification Model

Train a Monitoring Model

We start with the steps we used for the common model. First, let's create a directory structure for our monitoring model with an /src folder containing an inference scriptfunc_main.py:

mkdir -p monitoring_model/src
cd monitoring_model
touch src/func_main.py

As a monitoring metric, we will use IsolationForest. You can learn how it works in its documentation.

To make sure that our monitoring model will see the same data as our prediction model, we are going to apply the training data that was saved previously for our monitoring model.

import joblib
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.ensemble import IsolationForest

X_train = pd.read_csv('data/train.csv', index_col=0)

monitoring_model = IForest(contamination=0.04)

train_pred = monitoring_model.fit_predict(X_train) 

train_scores = monitoring_model.decision_function(X_train)

plt.hist(
    train_scores,
    bins=30, 
    alpha=0.5,
    density=True, 
    label="Train data outlier scores"
)

plt.vlines(monitoring_model.threshold_, 0, 1.9, label = "Threshold for marking outliers")
plt.gcf().set_size_inches(10, 5)
plt.legend()

dump(monitoring_model, "monitoring_model/monitoring_model.joblib")

This is what the distribution of our inliers looks like. By choosing a contamination parameter we can adjust a threshold that will separate inliers from outliers accordingly. You have to be thorough in choosing it to avoid critical prediction mistakes. Otherwise, you can also stay with 'auto'. To create a monitoring metric, we have to deploy that IsolationForest model as a separate model on the Hydrosphere platform. Let's save a trained model for serving.

Deploy a Monitoring Model with SDK

First, let's create a new directory where we will store our inference script with declared serving function and its definitions. Put the following code inside the src/func_main.py file:

import numpy as np
from joblib import load

monitoring_model = load('/model/files/monitoring_model.joblib')

features = ['age', 'workclass', 'fnlwgt',
            'education', 'educational-num', 'marital-status',
            'occupation', 'relationship', 'race', 'gender',
            'capital-gain', 'capital-loss', 'hours-per-week',
            'native-country']

def predict(**kwargs):
    x = np.array([kwargs[feature] for feature in features]).reshape(1, len(features))
    predicted = monitoring_model.decision_function(x)

    return {"value": predicted.item()}

Next, we need to install the necessary libraries. Create a requirements.txt and add the following libraries to it:

joblib==0.13.2
numpy==1.16.2
scikit-learn==0.23.1

Just like with common models, we can use SDK to upload our monitoring model and bind it to the trained one. The steps are almost the same, but with some slight differences. First, since we want to predict the anomaly score instead of sample class, we need to change the type of output field from 'int64' to 'float64'.

Secondly, we need to apply a couple of new methods to create a metric. MetricSpec is responsible for creating a metric for a specific model, with specific MetricSpecConfig.

from hydrosdk.monitoring import MetricSpec, MetricSpecConfig, ThresholdCmpOp

path_mon = "monitoring_model/"
payload_mon = ['src/func_main.py', 
               'monitoring_model.joblib', 'requirements.txt']

monitoring_signature = SignatureBuilder('predict') 
for i in X_train.columns:
    monitoring_signature.with_input(i, 'int64', 'scalar')
monitor_signature = monitoring_signature.with_output('value', 'float64', 'scalar').build()

monitor_contract = ModelContract(predict=monitor_signature)

monitoring_model_local = LocalModel(name="adult_monitoring_model", 
                              install_command = 'pip install -r requirements.txt',
                              contract=monitor_contract,
                              runtime=DockerImage("hydrosphere/serving-runtime-python-3.7", "2.3.2", None),
                              payload=payload_mon,
                              path=path_mon)
monitoring_upload = monitoring_model_local.upload(cluster)
monitoring_upload.lock_till_released()

metric_config = MetricSpecConfig(monitoring_upload.id, monitoring_model.threshold_, ThresholdCmpOp.LESS)
metric_spec = MetricSpec.create(cluster, "custom_metric", model_find.id, metric_config)

Anomaly scores are obtained through traffic shadowing inside the Hydrosphere's engine after making a Servable, so you don't need to perform any additional manipulations.

Managing Custom Metrics with UI

Go to the UI to observe and manage all your models. Here you will find 3 models on the left panel:

• adult_model - a model that we trained for prediction in the previous tutorial

• adult_monitoring_model - our monitoring model

• adult_model_metric - a model that was created by Automatic Outlier Detection

Click on the trained model and then on Monitoring. On the monitoring dasboard you now have two external metrics: the first one is auto_od_metric that was automatically generated by Automatic Outlier Detection, and the new one is custom_metric that we have just created. You can also change settings for existing metrics and configure the new ones in the Configure Metrics section:

During the prediction, you will get anomaly scores for each sample in the form of a chart with two lines. The curved line shows scores, while the horizontal dotted one is our threshold. When the curve intersects the threshold, it might be a sign of potential anomalousness. However, this is not always the case, since there are many factors that might affect this, so be careful about your final interpretation.

Uploading a Monitoring model with CLI

Just like in the case with all other types of models, we can define and upload a monitoring model using a resource definition. We have to pack our model with a model definition, like in the previous tutorial.

kind: Model
name: "adult_monitoring_model"
payload:
  - "src/"
  - "requirements.txt"
  - "monitoring_model.joblib"
runtime: "hydrosphere/serving-runtime-python-3.7:2.3.2"
install-command: "pip install -r requirements.txt"
contract:
  name: "predict"
  inputs:
    age:
      shape: scalar
      type: int64
    workclass:
      shape: scalar
      type: int64
    education:
      shape: scalar
      type: int64
    marital_status:
      shape: scalar
      type: int64
    occupation:
      shape: scalar
      type: int64
    relationship:
      shape: scalar
      type: int64
    race:
      shape: scalar
      type: int64
    sex:
      shape: scalar
      type: int64
    capital_gain:
      shape: scalar
      type: int64
    capital_loss:
      shape: scalar
      type: int64
    hours_per_week:
      shape: scalar
      type: int64
    country:
      shape: scalar
      type: int64
    classes:
      shape: scalar
      type: int64
  outputs:
    value:
      shape: scalar
      type: float64

Inputs of this model are the inputs of the target monitored model plus the outputs of that model. We will use the value field as an output for the monitoring model. The final directory structure should look like this:

.
├── monitoring_model.joblib
├── requirements.txt
├── serving.yaml
└── src
    └── func_main.py

From that folder, upload the model to the cluster:

hs apply -f serving.yaml

Now we have to attach the deployed Monitoring model as a custom metric. Let's create a monitoring metric for our pre-deployed classification model in the UI:

That's it. Now you have a monitored income classifier deployed on the Hydrosphere platform.