TSOutlierDetection

class hana_ai.tools.hana_ml_tools.ts_outlier_detection_tools.TSOutlierDetection(connection_context: ConnectionContext, return_direct: bool = False)

This tool detects outliers in time series data.

Parameters:

connection_contextConnectionContext

Connection context to the HANA database.

Returns:

str

The outliers in the time series data and the statistics of the detection.

Note

args_schema is used to define the schema of the inputs as follows:

Field	Description
table_name	the name of the table. If not provided, ask the user. Do not guess.
key	the key of the dataset. If not provided, ask the user. Do not guess.
endog	the endog of the dataset. If not provided, ask the user. Do not guess.
schema_name	the schema_name of the table, it is optional
auto	whether to use auto outlier detection, it is optional
detect_intermittent_ts	whether to detect intermittent time series, it is optional
smooth_method	the smoothing method for the time series chosen from {'no', 'median', 'loess'}, it is optional
window_size	odd number, the window size for median filter, not less than 3, it is optional
loess_lag	odd number, the lag for LOESS, not less than 3, it is optional
current_value_flag	whether to take the current data point when using LOESS smoothing method, it is optional
outlier_method	the outlier detection method chosen from {'z1', 'z2', 'mad', 'iqr', 'isolationforest', 'dbscan'}, it is optional
threshold	the threshold for outlier detection, it is optional
detect_seasonality	whether to detect seasonality, it is optional
alpha	the criterion for the autocorrelation coefficient, it is optional
extrapolation	whether to extrapolate the endpoints, it is optional
periods	the number of periods for seasonality, it is optional
random_state	specifies the seed for random number generator only valid for isolationforest, it is optional
n_estimators	the number of trees in the forest only valid for isolationforest, it is optional
max_samples	specifies the number of samples to draw from input to train each tree only valid for isolationforest, it is optional
bootstrap	whether to use bootstrap samples when building trees only valid for isolationforest, it is optional
contamination	the proportion of outliers in the data set only valid for isolationforest, it is optional
minpts	the number of points in a neighborhood for a point to be considered as a core point only valid for dbscan, it is optional
eps	the maximum distance between two samples for one to be considered as in the neighborhood of the other only valid for dbscan, it is optional
distance_method	the distance method for dbscan chosen from {'manhattan', 'euclidean', 'minkowski', 'chebyshev', 'standardized_euclidean', 'cosine'}, it is optional
dbscan_normalization	whether to normalize the data before dbscan, it is optional
dbscan_outlier_from_cluster	specifies how to take outliers from DBSCAN result, it is optional
thread_ratio	the ratio of threads to use for parallel processing, it is optional
residual_usage	specifies which residual to output chosen from {'outlier_detection', 'outlier_correction'}, it is optional
voting_config	the configuration for voting, it is optional
voting_outlier_method_criterion	the criterion for voting outlier method, it is optional

name: str

Name of the tool.

description: str

Description of the tool.

connection_context: ConnectionContext

Connection context to the HANA database.