ROCKET
- class hana_ml.algorithms.pal.tsa.rocket.ROCKET(method=None, num_features=None, data_dim=None, random_seed=None)
RandOm Convolutional KErnel Transform (ROCKET) is an exceptionally efficient algorithm for time series classification. Unlike other proposed time series classification algorithms which attain excellent accuracy, ROCKET maintains its performance with a fraction of the computational expense by transforming input time series using random convolutional kernels.
- Parameters
- methodstr, optional
The options are "MiniRocket" and "MultiRocket".
Defaults to "MiniRocket".
- num_featuresint, optional
Number of transformed features for each time series.
Defaults to 9996 when
method= "MiniRocket", 49728 whenmethod= "MultiRocket".- data_dimint, optional
Dimensionality of the multivariate time series.
1 means univariate time series and others for multivariate. Cannot be smaller than 1.
Defaults to 1.
- random_seedint, optional
0 indicates using machine time as seed.
Defaults to 0.
Examples
Example 1: Univariate time series fitted and transformed by MiniRocket Input dataframe is df:
>>> df.collect() RECORD_ID VAL_1 VAL_2 VAL_3 VAL_4 VAL_5 VAL_6 ... VAL_10 VAL_11 VAL_12 VAL_13 VAL_14 VAL_15 VAL_16 0 0 1.598 1.599 1.571 1.550 1.507 1.434 ... 1.117 1.024 0.926 0.828 0.739 0.643 0.556 1 1 1.701 1.671 1.619 1.547 1.475 1.391 ... 1.070 0.985 0.899 0.816 0.733 0.658 0.581 2 2 1.722 1.695 1.657 1.606 1.512 1.414 ... 1.015 0.920 0.828 0.740 0.658 0.586 0.501 3 3 1.726 1.660 1.573 1.496 1.409 1.332 ... 0.987 0.901 0.815 0.730 0.644 0.558 0.484 4 4 1.779 1.761 1.703 1.611 1.492 1.369 ... 0.900 0.786 0.679 0.580 0.502 0.415 0.333 5 5 1.800 1.743 1.686 1.633 1.532 1.423 ... 0.979 0.872 0.767 0.664 0.561 0.453 0.355 6 6 1.749 1.727 1.659 1.560 1.457 1.355 ... 0.961 0.864 0.771 0.682 0.595 0.513 0.427 7 7 1.348 1.237 1.129 1.022 0.939 0.847 ... 0.474 0.388 0.306 0.218 0.133 0.061 0.009 8 8 1.696 1.634 1.596 1.507 1.414 1.323 ... 1.048 0.966 0.890 0.805 0.719 0.632 0.553 9 9 1.723 1.713 1.665 1.587 1.495 1.404 ... 1.041 0.955 0.870 0.787 0.706 0.622 0.547 10 10 1.614 1.574 1.557 1.521 1.460 1.406 ... 1.045 0.957 0.862 0.771 0.681 0.587 0.497 11 11 1.652 1.665 1.656 1.623 1.571 1.499 ... 1.155 1.058 0.973 0.877 0.797 0.704 0.609
Create an instance of ROCKET:
>>> ro = ROCKET(method="MiniRocket", random_seed=1)
Performing fit() on the given dataframe:
>>> ro.fit(data=df)
Model:
>>> ro.model_.collect() ID MODEL_CONTENT 0 -1 MiniRocket 1 0 {"SERIES_LENGTH":16,"NUM_CHANNELS":1,"BIAS_SIZ... 2 1 843045766098464,1.0396523357230486,2.005001093... 3 2 67794124874738,1.3764943031483486,1.1499780774... 4 3 44031,0.8485649487700363,-4.853621348269184,0.... 5 4 69780774072532,1.5195149626729508,-0.397262332... ......
Make a transformation:
>>> result = ro.transform(data=df) >>> result.collect() ID STRING_CONTENT 0 0 {"NUM_FEATURES_PER_DATA":9996,"FEATURES":[{"DA... 1 1 ,0.375,0.875,0.125,0.5,1.0,0.25,0.75,1.0,0.375... 2 2 .625,1.0,0.375,0.875,0.125,0.5,1.0,0.125,0.875... 3 3 0.5625,0.0625,0.25,0.75,0.1875,0.3125,0.875,0.... 4 4 25,0.4375,0.6875,0.875,0.4375,0.8125,0.375,0.5... ......
Example 2: Multivariate time series (with dimensionality 8) fitted and transformed by MultiRocket Input dataframe is df:
>>> df.collect() RECORD_ID VAL_1 VAL_2 VAL_3 VAL_4 VAL_5 VAL_6 ... VAL_10 VAL_11 VAL_12 VAL_13 VAL_14 VAL_15 VAL_16 0 0 1.645 1.646 1.621 1.585 1.540 1.470 ... 1.161 1.070 0.980 0.893 0.798 0.705 0.620 1 1 1.704 1.705 1.706 1.680 1.632 1.560 ... 1.186 1.090 0.994 0.895 0.799 0.702 0.605 2 2 1.699 1.666 1.621 1.538 1.454 1.357 ... 0.979 0.885 0.793 0.706 0.623 0.541 0.460 3 3 1.709 1.663 1.580 1.497 1.413 1.330 ... 0.997 0.913 0.831 0.748 0.665 0.582 0.509 4 4 1.687 1.688 1.674 1.619 1.531 1.439 ... 1.069 0.977 0.900 0.810 0.722 0.644 0.557 ...... 27 27 1.697 1.665 1.590 1.508 1.424 1.341 ... 1.009 0.926 0.844 0.760 0.678 0.595 0.513 28 28 1.406 1.320 1.234 1.148 1.063 0.978 ... 0.642 0.558 0.477 0.396 0.314 0.234 0.153 29 29 1.592 1.593 1.571 1.551 1.527 1.475 ... 1.160 1.058 0.956 0.859 0.763 0.668 0.574 30 30 1.688 1.648 1.570 1.490 1.408 1.327 ... 1.011 0.930 0.849 0.768 0.687 0.606 0.524 31 31 1.708 1.663 1.595 1.504 1.411 1.318 ... 0.951 0.861 0.794 0.704 0.614 0.529 0.446
Create an instance of ROCKET:
>>> ro = ROCKET(method="multirocket", data_dim=8, random_seed=1)
Performing fit() on the given dataframe:
>>> ro.fit(data=df)
Model:
>>> ro.model_.collect() ID MODEL_CONTENT 0 -1 MultiRocket 1 0 {"SERIES_LENGTH":16,"NUM_CHANNELS":8,"BIAS_SIZ... 2 1 HANNELS":[6]},{"ID":77,"CHANNELS":[1,4,7,6,5]}... 3 2 340878522815215,7.959895076819708,5.8147048859... ......
Make a transformation:
>>> result = ro.transform(data=df) >>> result.collect() ID STRING_CONTENT 0 0 {"NUM_FEATURES_PER_DATA":49728,"FEATURES":[{"D... 1 1 .5625,0.8125,0.125,0.6875,0.9375,0.5,0.6875,0.... 2 2 5,0.5,0.8125,0.4375,0.5625,0.9375,0.5,0.75,0.1... 3 3 0.0,1.0,0.0,1.0,1.0,0.0,1.0,1.0,0.75,1.0,0.0,1... 4 4 25,0.0625,0.3125,0.375,0.1875,0.3125,0.9375,0.... ......
- Attributes
- model_DataFrame
Trained model content.
Methods
fit(data[, key])Generates a ROCKET model with given init parameters.
transform(data[, key, thread_ratio])Transform time series based on a given ROCKET model fitted by fit().
- fit(data, key=None)
Generates a ROCKET model with given init parameters.
- Parameters
- dataDataFrame
Input data. For univariate time series, each row represents one time series, while for multivariate time series, a fixed number of consecutive rows forms one time series, and that number is designated by the parameter
data_dimwhen initialize a ROCKET instance.- keystr, optional
The ID column.
Defaults to the first column of data if the index column of data is not provided. Otherwise, defaults to the index column of data.
- transform(data, key=None, thread_ratio=None)
Transform time series based on a given ROCKET model fitted by fit(). Hence, The data should be in the exact same format as that in fit(), especially the length and dimensionality of time series. The model_ used in transform comes from fit() as well.
- Parameters
- dataDataFrame
Input data. For univariate time series, each row represents one time series, while for multivariate time series, a fixed number of consecutive rows forms one time series, and that number is designated by the parameter
data_dimwhen initialize a ROCKET instance.- keystr, optional
The ID column.
Defaults to the first column of data if the index column of data is not provided. Otherwise, defaults to the index column of data.
- thread_ratiofloat, optional
Controls the proportion of available threads to use. The ratio of available threads.
0: single thread.
0~1: percentage.
Others: heuristically determined.
Defaults to 1.0.
- Returns
- DataFrame
Features, structured as follows:
ID, type INTEGER, ROW_INDEX, indicates the ID of current row.
STRING_CONTENT, type NVARCHAR, transformed features in JSON format.
- property fit_hdbprocedure
Returns the generated hdbprocedure for fit.
- property predict_hdbprocedure
Returns the generated hdbprocedure for predict.