Feature and Targets

A module for creating advanced time series features and targets including sliding windows and polynomial transformations

---

source

FeatureAndTargetGenerator

 FeatureAndTargetGenerator (context_len:int=10, target_len:int=10,
                            poly_degree:int=1)

Transforms time series data into feature matrices suitable for machine learning models. Creates lagged features using a sliding window and optionally generates polynomial features to capture non-linear relationships between variables. It also creates a target vector for the number of timesteps to predict.

	Type	Default	Details
context_len	int	10	number of previous timesteps to use as features
target_len	int	10	number of timesteps to predict
poly_degree	int	1	degree of polynomial features

We will first open the data

DATA_PATH = Path("../testing_data")

data = pd.read_csv(
    DATA_PATH/'hydro_example.csv', 
    usecols=['time', 'smoothed_rain', 'Q_mgb', 'Q_obs'], 
    index_col='time',
    converters={"time": pd.to_datetime}
    )

Then we need to create an instance of the generator setting the context and the target length and the polynomial degree.

generator = FeatureAndTargetGenerator(context_len=1, target_len=2, poly_degree=2)

Then we can generate the feature and target matrixes as follows

---

source

FeatureAndTargetGenerator.generate

 FeatureAndTargetGenerator.generate (df:pandas.core.frame.DataFrame,
                                     x_col:list[str], y_col:list[str])

Generates a feature matrix and target vector from the input data.

x_col, y_col = ['smoothed_rain','Q_mgb'], ['Q_obs']
data_x, data_y = generator.generate(data, x_col=x_col, y_col=y_col)

The generated data will look as follows

data_x.head(3)

	0	1	2	3	4	5
time
2012-01-01	1.0	0.299868	15.22	0.089921	4.563995	231.6484
2012-01-02	1.0	0.299767	14.84	0.089860	4.448548	220.2256
2012-01-03	1.0	0.265321	14.48	0.070395	3.841846	209.6704

data_y.head(3)

	t+1	t+2
time
2012-01-01	67.500000	67.349998
2012-01-02	67.349998	66.800003
2012-01-03	66.800003	66.739998