DocsTime Series

Time Series Tools

Tools for analyzing, forecasting, and processing temporal data sequences.

Available Tools

Decomposition

Trend, seasonality, residual

Stationarity Tests

ADF, KPSS, PP tests

Autocorrelation

ACF, PACF analysis

Forecasting

ARIMA, Prophet, exponential smoothing

Anomaly Detection

Detect unusual patterns

Time Series Decomposition

Separate time series into trend, seasonal, and residual components.

import cyxwiz.timeseries as ts

# Seasonal decomposition
result = ts.seasonal_decompose(
    data,
    model='additive',  # or 'multiplicative'
    period=12
)

trend = result.trend
seasonal = result.seasonal
residual = result.residual

# STL decomposition (more robust)
stl = ts.STL(data, period=12)
result = stl.fit()

Stationarity Tests

# Augmented Dickey-Fuller test
adf_stat, p_value, lags, nobs, crit_values = ts.adfuller(data)
print(f"ADF Statistic: {adf_stat:.4f}")
print(f"p-value: {p_value:.4f}")

if p_value < 0.05:
    print("Series is stationary")
else:
    print("Series is non-stationary")

# KPSS test
kpss_stat, p_value, lags, crit_values = ts.kpss(data)

# Make series stationary
diff_data = ts.diff(data)  # First difference
diff2_data = ts.diff(data, periods=2)  # Second difference

Autocorrelation Analysis

# Autocorrelation function
acf_values = ts.acf(data, nlags=40)

# Partial autocorrelation function
pacf_values = ts.pacf(data, nlags=40)

# Plot ACF and PACF
ts.plot_acf(data, lags=40)
ts.plot_pacf(data, lags=40)

# Ljung-Box test for autocorrelation
lb_stat, p_value = ts.ljung_box(residuals, lags=20)

Forecasting Models

ARIMA

from cyxwiz.timeseries import ARIMA

model = ARIMA(data, order=(1, 1, 1))
fitted = model.fit()

# Forecast
forecast = fitted.forecast(steps=30)

# With confidence intervals
forecast, conf_int = fitted.forecast(
    steps=30,
    return_conf_int=True
)

Exponential Smoothing

from cyxwiz.timeseries import ExponentialSmoothing

model = ExponentialSmoothing(
    data,
    trend='add',
    seasonal='mul',
    seasonal_periods=12
)
fitted = model.fit()

forecast = fitted.forecast(steps=12)

Window Functions

# Rolling statistics
rolling_mean = ts.rolling_mean(data, window=7)
rolling_std = ts.rolling_std(data, window=7)

# Exponential weighted
ewm_mean = ts.ewm(data, span=7).mean()

# Lag features
lagged = ts.lag(data, periods=1)
lagged_multi = ts.lag(data, periods=[1, 7, 30])

Node Editor Integration

Node	Inputs	Outputs
Decompose	Time series, period	Trend, seasonal, residual
Diff	Time series, periods	Differenced series
ARIMA	Time series, order	Forecast
Rolling	Time series, window	Rolling statistics