Azure AI Anomalydetector Java

1---
2name: azure-ai-anomalydetector-java
3description: Build anomaly detection applications with Azure AI Anomaly Detector SDK for Java. Use when implementing univariate/multivariate anomaly detection, time-series analysis, or AI-powered monitoring.
4package: com.azure:azure-ai-anomalydetector
5---
6 
7# Azure AI Anomaly Detector SDK for Java
8 
9Build anomaly detection applications using the Azure AI Anomaly Detector SDK for Java.
10 
11## Installation
12 
13```xml
14<dependency>
15  <groupId>com.azure</groupId>
16  <artifactId>azure-ai-anomalydetector</artifactId>
17  <version>3.0.0-beta.6</version>
18</dependency>
19```
20 
21## Client Creation
22 
23### Sync and Async Clients
24 
25```java
26import com.azure.ai.anomalydetector.AnomalyDetectorClientBuilder;
27import com.azure.ai.anomalydetector.MultivariateClient;
28import com.azure.ai.anomalydetector.UnivariateClient;
29import com.azure.core.credential.AzureKeyCredential;
30 
31String endpoint = System.getenv("AZURE_ANOMALY_DETECTOR_ENDPOINT");
32String key = System.getenv("AZURE_ANOMALY_DETECTOR_API_KEY");
33 
34// Multivariate client for multiple correlated signals
35MultivariateClient multivariateClient = new AnomalyDetectorClientBuilder()
36    .credential(new AzureKeyCredential(key))
37    .endpoint(endpoint)
38    .buildMultivariateClient();
39 
40// Univariate client for single variable analysis
41UnivariateClient univariateClient = new AnomalyDetectorClientBuilder()
42    .credential(new AzureKeyCredential(key))
43    .endpoint(endpoint)
44    .buildUnivariateClient();
45```
46 
47### With DefaultAzureCredential
48 
49```java
50import com.azure.identity.DefaultAzureCredentialBuilder;
51 
52MultivariateClient client = new AnomalyDetectorClientBuilder()
53    .credential(new DefaultAzureCredentialBuilder().build())
54    .endpoint(endpoint)
55    .buildMultivariateClient();
56```
57 
58## Key Concepts
59 
60### Univariate Anomaly Detection
61- **Batch Detection**: Analyze entire time series at once
62- **Streaming Detection**: Real-time detection on latest data point
63- **Change Point Detection**: Detect trend changes in time series
64 
65### Multivariate Anomaly Detection
66- Detect anomalies across 300+ correlated signals
67- Uses Graph Attention Network for inter-correlations
68- Three-step process: Train → Inference → Results
69 
70## Core Patterns
71 
72### Univariate Batch Detection
73 
74```java
75import com.azure.ai.anomalydetector.models.*;
76import java.time.OffsetDateTime;
77import java.util.List;
78 
79List<TimeSeriesPoint> series = List.of(
80    new TimeSeriesPoint(OffsetDateTime.parse("2023-01-01T00:00:00Z"), 1.0),
81    new TimeSeriesPoint(OffsetDateTime.parse("2023-01-02T00:00:00Z"), 2.5),
82    // ... more data points (minimum 12 points required)
83);
84 
85UnivariateDetectionOptions options = new UnivariateDetectionOptions(series)
86    .setGranularity(TimeGranularity.DAILY)
87    .setSensitivity(95);
88 
89UnivariateEntireDetectionResult result = univariateClient.detectUnivariateEntireSeries(options);
90 
91// Check for anomalies
92for (int i = 0; i < result.getIsAnomaly().size(); i++) {
93    if (result.getIsAnomaly().get(i)) {
94        System.out.printf("Anomaly detected at index %d with value %.2f%n",
95            i, series.get(i).getValue());
96    }
97}
98```
99 
100### Univariate Last Point Detection (Streaming)
101 
102```java
103UnivariateLastDetectionResult lastResult = univariateClient.detectUnivariateLastPoint(options);
104 
105if (lastResult.isAnomaly()) {
106    System.out.println("Latest point is an anomaly!");
107    System.out.printf("Expected: %.2f, Upper: %.2f, Lower: %.2f%n",
108        lastResult.getExpectedValue(),
109        lastResult.getUpperMargin(),
110        lastResult.getLowerMargin());
111}
112```
113 
114### Change Point Detection
115 
116```java
117UnivariateChangePointDetectionOptions changeOptions = 
118    new UnivariateChangePointDetectionOptions(series, TimeGranularity.DAILY);
119 
120UnivariateChangePointDetectionResult changeResult = 
121    univariateClient.detectUnivariateChangePoint(changeOptions);
122 
123for (int i = 0; i < changeResult.getIsChangePoint().size(); i++) {
124    if (changeResult.getIsChangePoint().get(i)) {
125        System.out.printf("Change point at index %d with confidence %.2f%n",
126            i, changeResult.getConfidenceScores().get(i));
127    }
128}
129```
130 
131### Multivariate Model Training
132 
133```java
134import com.azure.ai.anomalydetector.models.*;
135import com.azure.core.util.polling.SyncPoller;
136 
137// Prepare training request with blob storage data
138ModelInfo modelInfo = new ModelInfo()
139    .setDataSource("https://storage.blob.core.windows.net/container/data.zip?sasToken")
140    .setStartTime(OffsetDateTime.parse("2023-01-01T00:00:00Z"))
141    .setEndTime(OffsetDateTime.parse("2023-06-01T00:00:00Z"))
142    .setSlidingWindow(200)
143    .setDisplayName("MyMultivariateModel");
144 
145// Train model (long-running operation)
146AnomalyDetectionModel trainedModel = multivariateClient.trainMultivariateModel(modelInfo);
147 
148String modelId = trainedModel.getModelId();
149System.out.println("Model ID: " + modelId);
150 
151// Check training status
152AnomalyDetectionModel model = multivariateClient.getMultivariateModel(modelId);
153System.out.println("Status: " + model.getModelInfo().getStatus());
154```
155 
156### Multivariate Batch Inference
157 
158```java
159MultivariateBatchDetectionOptions detectionOptions = new MultivariateBatchDetectionOptions()
160    .setDataSource("https://storage.blob.core.windows.net/container/inference-data.zip?sasToken")
161    .setStartTime(OffsetDateTime.parse("2023-07-01T00:00:00Z"))
162    .setEndTime(OffsetDateTime.parse("2023-07-31T00:00:00Z"))
163    .setTopContributorCount(10);
164 
165MultivariateDetectionResult detectionResult = 
166    multivariateClient.detectMultivariateBatchAnomaly(modelId, detectionOptions);
167 
168String resultId = detectionResult.getResultId();
169 
170// Poll for results
171MultivariateDetectionResult result = multivariateClient.getBatchDetectionResult(resultId);
172for (AnomalyState state : result.getResults()) {
173    if (state.getValue().isAnomaly()) {
174        System.out.printf("Anomaly at %s, severity: %.2f%n",
175            state.getTimestamp(),
176            state.getValue().getSeverity());
177    }
178}
179```
180 
181### Multivariate Last Point Detection
182 
183```java
184MultivariateLastDetectionOptions lastOptions = new MultivariateLastDetectionOptions()
185    .setVariables(List.of(
186        new VariableValues("variable1", List.of("timestamp1"), List.of(1.0f)),
187        new VariableValues("variable2", List.of("timestamp1"), List.of(2.5f))
188    ))
189    .setTopContributorCount(5);
190 
191MultivariateLastDetectionResult lastResult = 
192    multivariateClient.detectMultivariateLastAnomaly(modelId, lastOptions);
193 
194if (lastResult.getValue().isAnomaly()) {
195    System.out.println("Anomaly detected!");
196    // Check contributing variables
197    for (AnomalyContributor contributor : lastResult.getValue().getInterpretation()) {
198        System.out.printf("Variable: %s, Contribution: %.2f%n",
199            contributor.getVariable(),
200            contributor.getContributionScore());
201    }
202}
203```
204 
205### Model Management
206 
207```java
208// List all models
209PagedIterable<AnomalyDetectionModel> models = multivariateClient.listMultivariateModels();
210for (AnomalyDetectionModel m : models) {
211    System.out.printf("Model: %s, Status: %s%n",
212        m.getModelId(),
213        m.getModelInfo().getStatus());
214}
215 
216// Delete a model
217multivariateClient.deleteMultivariateModel(modelId);
218```
219 
220## Error Handling
221 
222```java
223import com.azure.core.exception.HttpResponseException;
224 
225try {
226    univariateClient.detectUnivariateEntireSeries(options);
227} catch (HttpResponseException e) {
228    System.out.println("Status code: " + e.getResponse().getStatusCode());
229    System.out.println("Error: " + e.getMessage());
230}
231```
232 
233## Environment Variables
234 
235```bash
236AZURE_ANOMALY_DETECTOR_ENDPOINT=https://<resource>.cognitiveservices.azure.com/
237AZURE_ANOMALY_DETECTOR_API_KEY=<your-api-key>
238```
239 
240## Best Practices
241 
2421. **Minimum Data Points**: Univariate requires at least 12 points; more data improves accuracy
2432. **Granularity Alignment**: Match `TimeGranularity` to your actual data frequency
2443. **Sensitivity Tuning**: Higher values (0-99) detect more anomalies
2454. **Multivariate Training**: Use 200-1000 sliding window based on pattern complexity
2465. **Error Handling**: Always handle `HttpResponseException` for API errors
247 
248## Trigger Phrases
249 
250- "anomaly detection Java"
251- "detect anomalies time series"
252- "multivariate anomaly Java"
253- "univariate anomaly detection"
254- "streaming anomaly detection"
255- "change point detection"
256- "Azure AI Anomaly Detector"
257