Time-Series Analysis
From decomposition to auto ARIMA fitting — all built in.
Stationarity tests, cointegration, CCF, and other diagnostics in one flow. Include exogenous variables and get prediction intervals out of the box.
At a glance
Common use
Demand forecasting · KPI monitoring · Period-level impact
Engine
pmdarima auto_arima · statsmodels SARIMAX
Required columns
Time (datetime / Y·M·D) + at least one numeric series
Recommended length
≥ 30 time points; seasonal analysis needs ≥ 2 cycles
Plan
PREMIUM plan and above
Data preparation
- 1Tabular file (CSV / XLSX)
- 21 time column — datetime or separate Y · M · D columns
- 31+ numeric time-series variable
- 4Exogenous variables (holidays, events) can be added
- 5Evenly-spaced time points recommended (auto-interpolation available)
If you have separate Y · M · D columns instead of a datetime column, the preprocess step will auto-create a datetime column for you.
Workflow
- 1Recognise time column + datetime conversion
- 2Derive features (weekday · month · quarter · holiday flag)
- 3Time-series visualisation + correlation matrix
- 4Trend/seasonal decomposition
- 5Stationarity tests (ADF · KPSS) + auto-differencing
- 6Auto ARIMA / SARIMAX order selection (auto_arima)
- 7Forecast + confidence intervals + diagnostic plots
Supported analyses
Time-series visualisation
Compare multiple series simultaneously and spot patterns
Decomposition
Auto-decompose into trend · seasonality · residual
Stationarity tests
ADF / KPSS; auto-difference when non-stationary
CCF · cointegration
Lead-lag relationships + long-run equilibrium
ARIMA · SARIMAX
auto_arima finds optimal order, supports exogenous variables
Forecast + diagnostics
Forecast + 95% CI + Ljung-Box / Jarque-Bera diagnostics
Use cases
Monthly revenue forecast
5-year monthly revenue + ad spend (exogenous) → 12-month forecast via SARIMAX.
Call-centre volume forecast
Hourly call volume + holiday flag → next-week forecast for staffing.
KPI anomaly monitoring
Out-of-CI time points trigger alerts.
What you get
- Decomposition chart (trend · seasonal · residual)
- Stationarity test results + differencing log
- ARIMA order + fit stats (AIC · BIC · Log-Likelihood)
- Forecast chart + confidence intervals
- Residual diagnostics + auto-generated paper