Time series aggregations(时间序列聚合(Time series aggregations))¶
Quiver supports several time series aggregations that can operate on either one or several series at a time.
Single series aggregate¶
Single series aggregates operate on a single, continuous, time series producing one aggregated value for each specified time window.
Supported aggregation options¶
The following options are supported:
- Sum: Sum of all points in each window.
- Mean: Average of all points in each window.
- Standard deviation: Standard deviation of all points in each window.
- Max: Largest value in each window.
- Min: Smallest value in each window.
- Difference: Difference between the last point and first point in each window.
- Relative difference: Percent change between the last point and first point in each window.
- Product: Product of all points in each window.
- Count: Number of points in each window.
- First: Value of the first point in each window.
- Last: Value of the last point in each window.
Quiver cards using this aggregate¶
- Cumulative aggregate: This card applies a data point aggregation for each point in the input series, where the time window ranges from the start of the series to the current point (inclusive).
- Periodic aggregate: This card divides the time series into contiguous time windows of fixed duration and applies a data point aggregation to each. Windows with no data points are not included in the output.
- Rolling aggregate: This card applies a data point aggregation for each point in the input series where the time window ends at the current timestamp (inclusive) and spans a specified duration. As new data points are added to the window, old points fall out of it if they are outside the specified duration.
- Time series numeric aggregation: This card applies a data point aggregation to a single time window spanning the entire series.
- Time series scatter plot: This card uses a periodic aggregate.
- Bollinger bands: This card uses a rolling aggregate.
Multi-series aggregate¶
Multi-series aggregates align multiple continuous time series and combine one value from each series producing an aggregated value for each timestamp.
Supported aggregation options¶
The following options are supported:
- Sum: Sum of points across all series at each timestamp.
- Mean: Average of points across all series at each timestamp.
- Standard deviation: Standard deviation of points across all series at each timestamp.
- Max: Largest value in any series at each timestamp.
- Min: Smallest value in any series at each timestamp.
- Product: Product of points across all series at each timestamp.
Quiver cards using this aggregate¶
- Linear aggregation
- Linked series aggregation
- Time series formula: This card allows you to perform most of the operations in the above list, but with mathematical operators.
Interval aggregate¶
Interval aggregates operate on an interval time series which can be produced via the Filter time series card. They produce one aggregated value for each interval in the series. Aggregates are inclusive of points at the start timestamp of each interval and exclusive of points at the end timestamp.
Supported aggregation options¶
The following options are supported:
- Sum: Sum of all points in each interval.
- Mean: Average of all points in each interval.
- Standard deviation: Standard deviation of all points in each interval.
- Max: Largest value in each interval.
- Min: Smallest value in each interval.
- Difference: Difference between the last point and first point in each interval.
- Relative difference: Percent change between the last point and first point in each interval.
- Product: Product of all points in each interval.
- Count: Number of points in each interval.
- First: Value of the first point in each interval.
- Last: Value of the last point in each interval.
- Duration: Length of each interval in the precision that the time series is configured with (usually milliseconds).
- Rate change: Difference between the last point and first point in each interval divided by the interval's length in the precision that the time series is configured with (usually milliseconds).
Quiver cards using this aggregate¶
中文翻译¶
时间序列聚合(Time series aggregations)¶
Quiver 支持多种时间序列聚合操作,可同时对单个或多个序列进行处理。
单序列聚合(Single series aggregate)¶
单序列聚合对单个连续时间序列进行操作,为每个指定的时间窗口(time window)生成一个聚合值。
支持的聚合选项¶
支持以下选项:
- 总和(Sum): 每个窗口内所有数据点的总和。
- 平均值(Mean): 每个窗口内所有数据点的平均值。
- 标准差(Standard deviation): 每个窗口内所有数据点的标准差。
- 最大值(Max): 每个窗口内的最大值。
- 最小值(Min): 每个窗口内的最小值。
- 差值(Difference): 每个窗口内最后一个数据点与第一个数据点的差值。
- 相对差值(Relative difference): 每个窗口内最后一个数据点相对于第一个数据点的百分比变化。
- 乘积(Product): 每个窗口内所有数据点的乘积。
- 计数(Count): 每个窗口内的数据点数量。
- 首个值(First): 每个窗口内第一个数据点的值。
- 末个值(Last): 每个窗口内最后一个数据点的值。
使用此聚合的 Quiver 卡片¶
- 累积聚合(Cumulative aggregate): 该卡片对输入序列中的每个数据点应用数据点聚合,时间窗口从序列起点到当前数据点(含)。
- 周期聚合(Periodic aggregate): 该卡片将时间序列划分为固定时长的连续时间窗口,并对每个窗口应用数据点聚合。不含数据点的窗口不会出现在输出中。
- 滚动聚合(Rolling aggregate): 该卡片对输入序列中的每个数据点应用数据点聚合,时间窗口以当前时间戳(含)为终点,并跨越指定时长。当新数据点加入窗口时,超出指定时长的旧数据点将被移除。
- 时间序列数值聚合(Time series numeric aggregation): 该卡片对覆盖整个序列的单个时间窗口应用数据点聚合。
- 时间序列散点图(Time series scatter plot): 该卡片使用周期聚合。
- 布林带(Bollinger bands): 该卡片使用滚动聚合。
多序列聚合(Multi-series aggregate)¶
多序列聚合对齐多个连续时间序列,并从每个序列中提取一个值,为每个时间戳生成一个聚合值。
支持的聚合选项¶
支持以下选项:
- 总和(Sum): 每个时间戳上所有序列数据点的总和。
- 平均值(Mean): 每个时间戳上所有序列数据点的平均值。
- 标准差(Standard deviation): 每个时间戳上所有序列数据点的标准差。
- 最大值(Max): 每个时间戳上任意序列中的最大值。
- 最小值(Min): 每个时间戳上任意序列中的最小值。
- 乘积(Product): 每个时间戳上所有序列数据点的乘积。
使用此聚合的 Quiver 卡片¶
- 线性聚合(Linear aggregation)
- 链接序列聚合(Linked series aggregation)
- 时间序列公式(Time series formula): 该卡片允许您执行上述列表中的大部分操作,但使用数学运算符。
区间聚合(Interval aggregate)¶
区间聚合对区间时间序列进行操作,该序列可通过过滤时间序列(Filter time series)卡片生成。它为序列中的每个区间生成一个聚合值。聚合操作包含每个区间起始时间戳的数据点,不包含结束时间戳的数据点。
支持的聚合选项¶
支持以下选项:
- 总和(Sum): 每个区间内所有数据点的总和。
- 平均值(Mean): 每个区间内所有数据点的平均值。
- 标准差(Standard deviation): 每个区间内所有数据点的标准差。
- 最大值(Max): 每个区间内的最大值。
- 最小值(Min): 每个区间内的最小值。
- 差值(Difference): 每个区间内最后一个数据点与第一个数据点的差值。
- 相对差值(Relative difference): 每个区间内最后一个数据点相对于第一个数据点的百分比变化。
- 乘积(Product): 每个区间内所有数据点的乘积。
- 计数(Count): 每个区间内的数据点数量。
- 首个值(First): 每个区间内第一个数据点的值。
- 末个值(Last): 每个区间内最后一个数据点的值。
- 时长(Duration): 每个区间的长度,精度与时间序列配置一致(通常为毫秒)。
- 变化率(Rate change): 每个区间内最后一个数据点与第一个数据点的差值,除以区间长度,精度与时间序列配置一致(通常为毫秒)。