Курси /
Data Anomaly Detection
Types of Anomalies
| Type of Data Anomaly | Description | Example |
|---|---|---|
| Point Anomalies (Outliers) | Individual data points or instances that exhibit extreme values or deviate significantly from the rest of the data. | In a dataset of employee salaries within a company, an individual employee's salary is significantly higher or lower than the salaries of their peers. This could be due to a data entry error or a unique compensation arrangement. |
| Contextual Anomalies (Conditional Anomalies) | Data points that are considered anomalous only within a specific context or condition, requiring consideration of additional factors or conditions for detection. | In a temperature sensor dataset, a reading of 100°F may be normal in the summer but anomalous in the winter. Contextual analysis considers the time of year to identify anomalies. |
| Collective Anomalies (Group Anomalies) | Abnormal behavior exhibited by a collection of data points or a subset of the data when analyzed together as a group. | In a network traffic dataset, individual data packets may appear normal, but collectively, a sudden surge in traffic or an unusual pattern of communication between devices may indicate a collective anomaly, such as a network intrusion. |
| Global Anomalies | Anomalies that affect the entire dataset, causing a shift in the overall data distribution or patterns. | Monitoring daily website traffic, a sudden 90% decrease in daily traffic for the entire website indicates a global anomaly, possibly due to a server outage or significant changes in user behavior. |
In this course, we will concentrate primarily on the first type of anomalies: point anomalies, also commonly referred to as outliers.
Characteristics of Point Anomalies:
- Extreme Values: Point anomalies are typically characterized by data points that have values significantly higher or lower than the values of the majority of data points in the dataset. These extreme values can disrupt the analysis and modeling processes;
- Isolation: Point anomalies are often isolated from the rest of the data. They do not follow the general patterns or trends observed in the majority of the data, making them stand out;
- Causes: Point anomalies can result from a variety of causes, including data entry errors, sensor malfunctions, equipment failures, rare and unexpected events, or even fraudulent activities.
Все було зрозуміло?
Секція 1. Розділ 2
Зміст курсу
Data Anomaly Detection
2. Statistical Methods in Anomaly Detection
Data Anomaly Detection
Types of Anomalies
| Type of Data Anomaly | Description | Example |
|---|---|---|
| Point Anomalies (Outliers) | Individual data points or instances that exhibit extreme values or deviate significantly from the rest of the data. | In a dataset of employee salaries within a company, an individual employee's salary is significantly higher or lower than the salaries of their peers. This could be due to a data entry error or a unique compensation arrangement. |
| Contextual Anomalies (Conditional Anomalies) | Data points that are considered anomalous only within a specific context or condition, requiring consideration of additional factors or conditions for detection. | In a temperature sensor dataset, a reading of 100°F may be normal in the summer but anomalous in the winter. Contextual analysis considers the time of year to identify anomalies. |
| Collective Anomalies (Group Anomalies) | Abnormal behavior exhibited by a collection of data points or a subset of the data when analyzed together as a group. | In a network traffic dataset, individual data packets may appear normal, but collectively, a sudden surge in traffic or an unusual pattern of communication between devices may indicate a collective anomaly, such as a network intrusion. |
| Global Anomalies | Anomalies that affect the entire dataset, causing a shift in the overall data distribution or patterns. | Monitoring daily website traffic, a sudden 90% decrease in daily traffic for the entire website indicates a global anomaly, possibly due to a server outage or significant changes in user behavior. |
In this course, we will concentrate primarily on the first type of anomalies: point anomalies, also commonly referred to as outliers.
Characteristics of Point Anomalies:
- Extreme Values: Point anomalies are typically characterized by data points that have values significantly higher or lower than the values of the majority of data points in the dataset. These extreme values can disrupt the analysis and modeling processes;
- Isolation: Point anomalies are often isolated from the rest of the data. They do not follow the general patterns or trends observed in the majority of the data, making them stand out;
- Causes: Point anomalies can result from a variety of causes, including data entry errors, sensor malfunctions, equipment failures, rare and unexpected events, or even fraudulent activities.
Все було зрозуміло?
Секція 1. Розділ 2
