SLS machine learning functions - time series prediction, anomaly detection, clustering - Simple Log Service

SLS provides SQL-compatible machine learning functions for time series analysis, including smoothing, prediction, anomaly detection, decomposition, and multi-series clustering.

Features

Smoothing operations for single time series data.
Prediction, anomaly detection, change point detection, inflection point detection, and multi-period estimation for single time series data.
Decomposition of single time series data.
Clustering algorithms for multiple time series.
Pattern mining across multiple fields of numeric or text data.

Limits

Time series data must be sampled at the same interval.
The data cannot contain multiple samples from the same point in time.

Processing capacity limits:

Item	Limit
Time series processing	Maximum 150,000 consecutive data points. If exceeded, aggregate the data or reduce the sampling amount.
Density-based clustering	Up to 5,000 time series, each limited to 1,440 data points.
Hierarchical clustering	Up to 2,000 time series, each limited to 1,440 data points.

Machine learning functions

Category		Function	Description
Time series	Smooth function	ts_smooth_simple	Smooths time series data using the Holt Winters algorithm.
		ts_smooth_fir	Smooths time series data using a finite impulse response (FIR) filter.
		ts_smooth_iir	Smooths time series data using an infinite impulse response (IIR) filter.
	Multi-period estimation function	ts_period_detect	Estimates time series data by period.
	Change point detection function	ts_cp_detect	Detects intervals with differing statistical features and identifies their endpoints as change points.
	Change point detection function	ts_breakout_detect	Detects points where data experiences dramatic changes.
	Maximum value detection function	ts_find_peaks	Detects local maximum values in a specified window.
	Prediction and anomaly detection function	ts_predicate_simple	Models, predicts, and detects anomalies in time series data using default parameters.
		ts_predicate_ar	Models, predicts, and detects anomalies using an autoregressive (AR) model.
		ts_predicate_arma	Models, predicts, and detects anomalies using an autoregressive moving average (ARMA) model.
		ts_predicate_arima	Models, predicts, and detects anomalies using an autoregressive integrated moving average (ARIMA) model.
		ts_regression_predict	Predicts the long-run trend for a single periodic time series.
	Sequence decomposition function	ts_decompose	Decomposes time series data using the STL (Seasonal and Trend decomposition using Loess) algorithm.
	Kernel density estimation functions	kernel_density_estimation	Fits observed data points with a smooth peak function to approximate the probability distribution.
Pattern mining	Frequent pattern mining functions	pattern_stat	Mines representative attribute combinations from multi-attribute field samples to find frequent patterns.
	Pattern mining (tabular data drill-down analysis) function	pattern_diff	Identifies patterns that cause differences between two data collections.
	Root cause analysis functions	rca_kpi_search	Analyzes subdimension attributes that cause metric anomalies.
	Correlation analysis functions	ts_association_analysis	Identifies metrics correlated to a specified metric among multiple observed metrics.
	Correlation analysis functions	ts_similar	Identifies metrics correlated to specified time series data among multiple observed ones in the system.
	Request URL classification function	url_classify	Classifies request URLs and assigns tags with defining regular expressions.