{ "slug": "time-series-forecaster", "name": "gizza-ai/time-series-forecaster", "version": "0.1.0", "title": "Time Series Forecaster — Exponential Smoothing & Holt-Winters — gizza.ai", "description": "Forecast a time series with exponential smoothing, Holt trend, damped trend, or Holt-Winters seasonality, including prediction intervals and accuracy metrics.", "tags": [ "time series forecasting", "exponential smoothing calculator", "holt winters forecast", "demand forecasting tool", "sales forecast calculator", "prediction interval", "trend and seasonality", "ets model" ], "category": "time", "urls": { "page": "https://gizza.ai/tools/time-series-forecaster/", "markdown": "https://gizza.ai/tools/time-series-forecaster/index.md", "descriptor": "https://gizza.ai/tools/time-series-forecaster/tool.json", "deep_link_example": "https://gizza.ai/tools/time-series-forecaster/?data=month%2Csales%0AJan%2C120%0AFeb%2C132%0AMar%2C141%0AApr%2C158&model=auto&horizon=6&season_length=0&alpha=0&beta=0&gamma=0&phi=0&confidence=95&show_fitted=true&header=auto&decimals=3&format=text" }, "cli": "gizza tool time-series-forecaster \"month,sales\nJan,120\nFeb,132\nMar,141\nApr,158\"", "tool": { "description": "Forecast future values of a univariate time series using the exponential-smoothing (ETS) family: simple exponential smoothing, Holt's linear trend, damped trend, and Holt-Winters additive or multiplicative seasonality. Smoothing weights left at 0 are fitted automatically by a deterministic grid search that minimises the sum of squared one-step-ahead errors, and model=auto picks the best candidate by AICc. Returns the chosen model, its alpha/beta/gamma/phi weights, in-sample accuracy (MAE, RMSE, MAPE, MASE, residual sigma, AICc), and a forecast table with 80/90/95/99 percent prediction intervals, optionally alongside per-period fitted values and residuals, as text, CSV, or JSON. Accepts labelled rows, single-line series, currency and percent symbols; up to 10000 observations and a 240-period horizon.", "parameters": { "additionalProperties": false, "properties": { "alpha": { "default": 0.0, "description": "Level smoothing weight between 0 and 1. Leave at 0 (default) to fit it automatically by minimising the sum of squared one-step-ahead errors. Higher values react faster to recent observations.", "maximum": 1, "minimum": 0, "type": "number" }, "beta": { "default": 0.0, "description": "Trend smoothing weight between 0 and 1, used by holt, damped and both Holt-Winters models. Leave at 0 (default) to fit it automatically. Ignored by the simple model.", "maximum": 1, "minimum": 0, "type": "number" }, "confidence": { "default": "95", "description": "Prediction-interval confidence level in percent: 80, 90, 95 (default) or 99. Wider levels give wider bands around each forecast.", "enum": [ "80", "90", "95", "99" ], "type": "string" }, "data": { "description": "The historical series, oldest observation first. Paste one value per line, an optional label before the value (\"Jan,120\"), or the whole series on one comma/semicolon/space separated line. Values may carry currency symbols, percent signs or accounting-style (12.5) negatives. Minimum 3 observations, maximum 10000.", "type": "string" }, "decimals": { "default": 3, "description": "Decimal places for forecasts and accuracy metrics, 0 to 10. Default 3.", "maximum": 10, "minimum": 0, "type": "integer" }, "format": { "default": "text", "description": "Output format: text report (model, smoothing parameters, accuracy metrics, forecast table with prediction intervals), csv sections, or structured json. Default text.", "enum": [ "text", "csv", "json" ], "type": "string" }, "gamma": { "default": 0.0, "description": "Seasonal smoothing weight between 0 and 1, used by the Holt-Winters models. Leave at 0 (default) to fit it automatically. Ignored by non-seasonal models.", "maximum": 1, "minimum": 0, "type": "number" }, "header": { "default": "auto", "description": "How to treat a leading label row. auto (default) drops the first row only when its value field is not numeric; yes always drops it; no parses every row as data.", "enum": [ "auto", "yes", "no" ], "type": "string" }, "horizon": { "default": 6, "description": "How many future periods to forecast, 1 to 240. Default 6.", "maximum": 240, "minimum": 1, "type": "integer" }, "model": { "default": "auto", "description": "Exponential-smoothing model. auto (default) fits every applicable candidate and picks the best AICc; simple = simple exponential smoothing (level only); holt = additive linear trend; damped = trend damped by phi so long horizons flatten; holt-winters-additive = trend plus a constant-size seasonal cycle; holt-winters-multiplicative = trend plus a seasonal cycle that grows with the level (needs strictly positive values). Both Holt-Winters models require season_length of 2 or more.", "enum": [ "auto", "simple", "holt", "damped", "holt-winters-additive", "holt-winters-multiplicative" ], "type": "string" }, "phi": { "default": 0.0, "description": "Trend damping factor between 0 and 1 for model=damped; values near 1 keep the trend, lower values flatten it quickly. Leave at 0 (default) to fit it automatically in the range 0.6 to 0.99. Ignored by other models.", "maximum": 1, "minimum": 0, "type": "number" }, "season_length": { "default": 0, "description": "Observations per seasonal cycle: 12 for monthly data with a yearly cycle, 4 for quarterly, 7 for daily data with a weekly cycle. Use 0 (default) for a non-seasonal series. Seasonal models need at least two full cycles of history.", "maximum": 366, "minimum": 0, "type": "integer" }, "show_fitted": { "default": false, "description": "Include the per-period table of actual values, one-step-ahead fitted values and residuals (default false). Useful for checking how well the model tracked history.", "type": "boolean" } }, "required": [ "data" ], "type": "object" } } }