Synthetic error generation methods

Error generation

This example illustrates how to generate synthetic errors using the generateSyntheticErrors function from the preprocessing module. It begins by creating a pandas Series s of 10 random values from a normal distribution and then introduces “noise” errors to 50% of the data. The final output shows the initial data along with the modified time series “with_errors” and the corresponding error types, where errors are labeled as “noise.” The “noise”, “bias”, “drift”, “constant value”, “outlier”, and “missing” errors are available error types. Parameter adaptions are available for some of the error types.

>>>import numpy as np

>>>num_samples = 10
>>>s = pd.Series(np.random.normal(0, 5, num_samples), name='initial')
>>>s_err, s_errtype = TSCC.preprocessing.generateSyntheticErrors(s, error_type = ["noise"], error_rate = 0.5)
>>>s_err.name = "with_errors"
>>>print(pd.DataFrame([s, s_err]).transpose())
    initial with_errors error_type
0  5.083964   10.776748      noise
1  5.518957     6.91386      noise
2 -0.835532   -0.835532
3 -0.182044   -0.182044
4 -4.237007   -4.237007
5 -1.867304   -0.504853      noise
6  0.429116    0.429116
7  1.060643    1.060643
8 -0.209708   -0.209708
9 -2.137849   -2.137849