{ "slug": "time-series-generator", "name": "gizza-ai/time-series-generator", "version": "0.1.0", "title": "Time Series Generator — Synthetic Trend, Seasonality and Noise Data — gizza.ai", "description": "Generate deterministic synthetic time-series CSV, TSV, JSON or NDJSON with configurable trend, seasonality, noise, missing values and outliers.", "tags": [ "time series", "synthetic data", "csv", "json", "forecasting", "testing", "seasonality", "noise", "statistics", "data generation" ], "category": "data", "urls": { "page": "https://gizza.ai/tools/time-series-generator/", "markdown": "https://gizza.ai/tools/time-series-generator/index.md", "descriptor": "https://gizza.ai/tools/time-series-generator/tool.json", "deep_link_example": "https://gizza.ai/tools/time-series-generator/?start=2024-01-01&interval=1d&count=100&base=100&trend=linear&trend_strength=0.5&seasonality=sine&period=7%20or%2024%2C168&litude=10%20or%208%2C4&weekday_pattern=1.1%2C%201.05%2C%201%2C%201.05%2C%201.25%2C%200.8%2C%200.7&combine=additive&noise=gaussian&noise_level=5&noise_phi=0.7&missing_rate=0&outlier_rate=0&outlier_magnitude=3&outlier_direction=both&min_value=0&max_value=250&series=1&seed=42&decimals=2&output=csv×tamp_format=auto&header=true&labels=visits%2Csignups" }, "cli": "gizza tool time-series-generator 'start=2024-01-01'", "tool": { "description": "Generate synthetic time-series data for tests, demos and examples. The tool builds a timestamp index from start, interval and count; layers configurable trend, seasonality, seeded noise, missing values and outliers; supports one or more parallel series; clamps and rounds values; and returns CSV, TSV, JSON, NDJSON or a stats summary. Randomness is deterministic from seed so the same settings reproduce exactly across chat, CLI and browser. Runs locally and does not upload data.", "parameters": { "additionalProperties": false, "properties": { "amplitude": { "default": "10", "description": "Seasonal amplitude. May be a comma-separated list matching period; in multiplicative mode values are fractions of the level, otherwise they are additive units. Default 10.", "type": "string" }, "base": { "default": 100.0, "description": "Starting level of the series before trend, seasonality and noise are applied. Default 100.", "type": "number" }, "combine": { "default": "additive", "description": "How trend and seasonal/noise layers combine. additive sums units; multiplicative treats amplitude and noise_level as fractions of the current level. Default additive.", "enum": [ "additive", "multiplicative" ], "type": "string" }, "count": { "default": 100, "description": "Rows to generate, 1-100000. The total count x series is capped at 200000 emitted values. Default 100.", "maximum": 100000, "minimum": 1, "type": "integer" }, "decimals": { "default": 2, "description": "Digits after the decimal point, 0-10. Use 0 for integer-like counts. Default 2.", "maximum": 10, "minimum": 0, "type": "integer" }, "header": { "default": true, "description": "Include the timestamp/value header row in csv and tsv output. Ignored by json, ndjson and stats. Default true.", "type": "boolean" }, "interval": { "default": "1d", "description": "Step between rows. Use a positive number plus unit: ms, s, m, h, d, w, mo, q or y, for example 15m, 1h, 1d or 1mo. Calendar months/quarters/years keep month ends aligned. Default 1d.", "type": "string" }, "labels": { "default": "", "description": "Optional comma-separated labels for value columns. Supply one label per series; leave blank to use value, value_2 and so on.", "type": "string" }, "max_value": { "default": "", "description": "Optional numeric upper clamp applied after outliers. Leave blank for no upper bound.", "type": "string" }, "min_value": { "default": "", "description": "Optional numeric lower clamp applied after outliers. Leave blank for no lower bound.", "type": "string" }, "missing_rate": { "default": 0.0, "description": "Probability that each generated value is blank/null, from 0 to 1. CSV/TSV emit an empty cell; JSON/NDJSON emit null. Default 0.", "maximum": 1, "minimum": 0, "type": "number" }, "noise": { "default": "gaussian", "description": "Noise process. gaussian draws normal noise, uniform draws in +/- noise_level, and ar1 creates autocorrelated noise controlled by noise_phi. Default gaussian.", "enum": [ "none", "gaussian", "uniform", "ar1" ], "type": "string" }, "noise_level": { "default": 5.0, "description": "Noise scale. For gaussian it is the standard deviation, for uniform it is the half-width, and for multiplicative mode it is a fraction of the level. Default 5.", "minimum": 0, "type": "number" }, "noise_phi": { "default": 0.7, "description": "AR(1) correlation coefficient used only when noise=ar1. Must be between -0.99 and 0.99. Default 0.7.", "maximum": 0.99, "minimum": -0.99, "type": "number" }, "outlier_direction": { "default": "both", "description": "Allowed outlier direction: both, up only, or down only. Default both.", "enum": [ "both", "up", "down" ], "type": "string" }, "outlier_magnitude": { "default": 3.0, "description": "Outlier size as a multiple of the value: 3 means upward outliers become value*(1+3), downward outliers become value/(1+3). Default 3.", "minimum": 0, "type": "number" }, "outlier_rate": { "default": 0.0, "description": "Probability that each generated value receives an outlier multiplier, from 0 to 1. Default 0.", "maximum": 1, "minimum": 0, "type": "number" }, "output": { "default": "csv", "description": "Result format. csv/tsv emit rows, json emits one object with rows, ndjson emits one object per row, and stats reports achieved min/max/mean/sd plus missing/outlier counts. Default csv.", "enum": [ "csv", "tsv", "json", "ndjson", "stats" ], "type": "string" }, "period": { "default": "7", "description": "Cycle length in rows. Comma-separated lists superimpose multiple cycles, such as period='24,168' for daily and weekly hourly data. Default 7.", "type": "string" }, "seasonality": { "default": "sine", "description": "Seasonality shape. Cyclic shapes use period and amplitude; weekday uses weekday_pattern as Monday through Sunday multipliers. Default sine.", "enum": [ "none", "sine", "cosine", "square", "triangle", "sawtooth", "weekday" ], "type": "string" }, "seed": { "default": 42, "description": "Seed for the deterministic SplitMix64 random streams. The same settings and seed return identical output in chat, CLI and page. Default 42.", "minimum": 0, "type": "integer" }, "series": { "default": 1, "description": "Number of parallel value columns, 1-20. Columns share the trend/seasonal signal but have independent seeded noise, missingness and outliers. Default 1.", "maximum": 20, "minimum": 1, "type": "integer" }, "start": { "default": "2024-01-01", "description": "First timestamp. Accepts YYYY-MM-DD, RFC3339/ISO datetimes, or a Unix epoch seconds value. Default 2024-01-01.", "type": "string" }, "timestamp_format": { "default": "auto", "description": "Timestamp output format. auto uses dates for day-or-larger intervals and ISO datetimes for sub-day intervals; epoch emits seconds; index emits row numbers. Default auto.", "enum": [ "auto", "iso", "date", "epoch", "index" ], "type": "string" }, "trend": { "default": "linear", "description": "Trend layer. none keeps the base flat; linear adds trend_strength per step; exponential compounds by trend_strength percent per step; logistic makes an S-curve across the requested count; random-walk adds seeded increments. Default linear.", "enum": [ "none", "linear", "exponential", "logistic", "random-walk" ], "type": "string" }, "trend_strength": { "default": 0.5, "description": "Trend amount. For linear/random-walk it is units per step, for exponential it is percent per step, and for logistic it is the total rise from first to last row. Default 0.5.", "type": "number" }, "weekday_pattern": { "default": "1.1, 1.05, 1, 1.05, 1.25, 0.8, 0.7", "description": "Seven Monday-through-Sunday values used when seasonality=weekday. The pattern is mean-centred and scaled by amplitude so it works with additive or multiplicative combine modes.", "type": "string" } }, "type": "object" } } }