Get Started. It's Free
or sign up with your email address
Rocket clouds
Anomaly Detection by Mind Map: Anomaly Detection

1. Categorizing Anomalies

1.1. Point Anomalies

1.1.1. Techniques Classification Based learning testing methods concerns Nearest Neighbor based Distance to Kth nearest neighbor relative density of each data instance advantages disadvantages Clustering Based Points that does not belong to any cluster distance to the closest cluster center anomalies are clustered together, yet smaller and with less density Statistical Techniques parametric techniques non parametric techniques Information theoritic Kolomogorv Complexity Maximzing the complexity of normal instances multidimensional data Spectral Principal Component Analysis

1.2. Contextual Anomalies

1.2.1. data contextual attributes Spacial Graph Sequential Profile behavioral attributes

1.2.2. techniques reducing to point anomaly detection identify context compute anomaly score model the structure in data

1.3. Collective anomalies

1.3.1. Not discussed in detail..

2. Modes of detection

2.1. Supervised

2.2. Semi Supervised

2.3. Unsupervised

3. Training/ test Data

3.1. Labeled Data set

3.2. Normal Labeled Dataset

3.3. labeled abnormal dataset (rare)

3.4. Unlabeled dataset assumed very low abnormalities

4. Outputs

4.1. Score

4.1.1. List sorted by abnormality

4.2. Label

4.2.1. Tell whether normal or abnomal

5. Applications

5.1. Intrusion Detection

5.1.1. Host based multi point system traces can be available sequential / collective anomalies limited alphabet point anomaly detection is not applicable

5.1.2. Network based Network data collective anomalies challenges Nature of anomalies keep changing over time adapting intruders

5.2. Fraud Detection

5.2.1. Actual users doing unauthorized things

5.2.2. Credit card fraud labeled data available point anomalies detection by-owner by-operation

5.2.3. Mobile Phone Frauds

5.2.4. Insurance claim frauds

5.2.5. insider trading

5.3. Sensor networks

5.3.1. sensor faults

5.3.2. event (i.e. intrusion) detection

5.4. Image processing

5.4.1. Contextually different Points or regions

5.5. industrial damage detection

5.5.1. mechanical components defects

5.5.2. defects in physical structures

5.6. Medical/public health

5.6.1. abnormal patient conditions

5.6.2. instrument errors

5.6.3. recording errors